Imported: 17 Feb '17 | Published: 23 Sep '14

USPTO - Utility Patents

Disclosed is a transmission scheme for transmitting a first modulated signal and a second modulated signal over the same frequency at the same time. According to the transmission scheme, a precoding weight multiplying unit multiplies a baseband signal after a first mapping and a baseband signal after a second mapping by a precoding weight and outputs the first modulated signal and the second modulated signal. In the precoding weight multiplying unit, precoding weights are regularly hopped.

The present invention relates to a precoding scheme, a precoding device, a transmission scheme, a transmission device, a reception scheme, and a reception device that in particular perform communication using a multi-antenna.

This application is based on Japanese Patent Applications No. 2010-138532, No. 2010-152503, No. 2011-177310, No. 2011-250331, No. 2011-275165, and No. 2011-276456 filed in Japan, the contents of which are hereby incorporated by reference.

Multiple-Input Multiple-Output (MIMO) is a conventional example of a communication scheme using a multi-antenna. In multi-antenna communication, of which MIMO is representative, multiple transmission signals are each modulated, and each modulated signal is transmitted from a different antenna simultaneously in order to increase the transmission speed of data.

FIG. 28 shows an example of the structure of a transmission and reception device when the number of transmit antennas is two, the number of receive antennas is two, and the number of modulated signals for transmission (transmission streams) is two. In the transmission device, encoded data is interleaved, the interleaved data is modulated, and frequency conversion and the like is performed to generate transmission signals, and the transmission signals are transmitted from antennas. In this case, the scheme for simultaneously transmitting different modulated signals from different transmit antennas at the same time and at the same frequency is a spatial multiplexing MIMO system.

In this context, it has been suggested in Patent Literature 1 to use a transmission device provided with a different interleave pattern for each transmit antenna. In other words, the transmission device in FIG. 28 would have two different interleave patterns with respective interleaves (πa, πb). As shown in Non-Patent Literature 1 and Non-Patent Literature 2, reception quality is improved in the reception device by iterative performance of a detection scheme that uses soft values (the MIMO detector in FIG. 28).

Models of actual propagation environments in wireless communications include non-line of sight (NLOS), of which a Rayleigh fading environment is representative, and line of sight (LOS), of which a Rician fading environment is representative. When the transmission device transmits a single modulated signal, and the reception device performs maximal ratio combining on the signals received by a plurality of antennas and then demodulates and decodes the signal resulting from maximal ratio combining, excellent reception quality can be achieved in an LOS environment, in particular in an environment where the Rician factor is large, which indicates the ratio of the received power of direct waves versus the received power of scattered waves. However, depending on the transmission system (for example, spatial multiplexing MIMO system), a problem occurs in that the reception quality deteriorates as the Rician factor increases (see Non-Patent Literature 3).

FIGS. 29A and 29B show an example of simulation results of the Bit Error Rate (BER) characteristics (vertical axis:BER, horizontal axis:signal-to-noise power ratio (SNR)) for data encoded with low-density parity-check (LDPC) code and transmitted over a 2×2 (two transmit antennas, two receive antennas) spatial multiplexing MIMO system in a Rayleigh fading environment and in a Rician fading environment with Rician factors of K=3, 10, and 16 dB. FIG. 29A shows the BER characteristics of Max-log A Posteriori Probability (APP) without iterative detection (see Non-Patent Literature 1 and Non-Patent Literature 2), and FIG. 29B shows the BER characteristics of Max-log-APP with iterative detection (see Non-Patent Literature 1 and Non-Patent Literature 2) (number of iterations: five). As is clear from FIGS. 29A and 29B, regardless of whether iterative detection is performed, reception quality degrades in the spatial multiplexing MIMO system as the Rician factor increases. It is thus clear that the unique problem of “degradation of reception quality upon stabilization of the propagation environment in the spatial multiplexing MIMO system”, which does not exist in a conventional single modulation signal transmission system, occurs in the spatial multiplexing MIMO system.

Broadcast or multicast communication is a service directed towards line-of-sight users. The radio wave propagation environment between the broadcasting station and the reception devices belonging to the users is often an LOS environment. When using a spatial multiplexing MIMO system having the above problem for broadcast or multicast communication, a situation may occur in which the received electric field strength is high at the reception device, but degradation in reception quality makes it impossible to receive the service. In other words, in order to use a spatial multiplexing MIMO system in broadcast or multicast communication in both an NLOS environment and an LOS environment, there is a desire for development of a MIMO system that offers a certain degree of reception quality.

Non-Patent Literature 8 describes a scheme to select a codebook used in precoding (i.e. a precoding matrix, also referred to as a precoding weight matrix) based on feedback information from a communication partner. Non-Patent Literature 8 does not at all disclose, however, a scheme for precoding in an environment in which feedback information cannot be acquired from the communication partner, such as in the above broadcast or multicast communication.

On the other hand, Non-Patent Literature 4 discloses a scheme for hopping the precoding matrix over time. This scheme can be applied even when no feedback information is available. Non-Patent Literature 4 discloses using a unitary matrix as the matrix for precoding and hopping the unitary matrix at random but does not at all disclose a scheme applicable to degradation of reception quality in the above-described LOS environment. Non-Patent Literature 4 simply recites hopping between precoding matrices at random. Obviously, Non-Patent Literature 4 makes no mention whatsoever of a precoding scheme, or a structure of a precoding matrix, for remedying degradation of reception quality in an LOS environment.

- Patent Literature 1
- WO 2005/050885

- Non-Patent Literature 1
- “Achieving near-capacity on a multiple-antenna channel”, IEEE Transaction on Communications, vol. 51, no. 3, pp. 389-399, March 2003.
- Non-Patent Literature 2
- “Performance analysis and design optimization of LDPC-coded MIMO OFDM systems”, IEEE Trans. Signal Processing, vol. 52, no. 2, pp. 348-361, February 2004.
- Non-Patent Literature 3
- “BER performance evaluation in 2×2 MIMO spatial multiplexing systems under Rician fading channels”, IEICE Trans. Fundamentals, vol. E91-A, no. 10, pp. 2798-2807, October 2008.
- Non-Patent Literature 4
- “Turbo space-time codes with time varying linear transformations”, IEEE Trans. Wireless communications, vol. 6, no. 2, pp. 486-493, February 2007.
- Non-Patent Literature 5
- “Likelihood function for QR-MLD suitable for soft-decision turbo decoding and its performance”, IEICE Trans. Commun., vol. E88-B, no. 1, pp. 47-57, January 2004.
- Non-Patent Literature 6
- “A tutorial on ‘parallel concatenated (Turbo) coding’, ‘Turbo (iterative) decoding’ and related topics”, The Institute of Electronics, Information, and Communication Engineers, Technical Report IT 98-51.
- Non-Patent Literature 7
- “Advanced signal processing for PLCs: Wavelet-OFDM”, Proc. of IEEE International symposium on ISPLC 2008, pp. 187-192, 2008.
- Non-Patent Literature 8
- D. J. Love, and R. W. Heath, Jr., “Limited feedback unitary precoding for spatial multiplexing systems”, IEEE Trans. Inf. Theory, vol. 51, no. 8, pp. 2967-2976, August 2005.
- Non-Patent Literature 9
- DVB Document A122, Framing structure, channel coding and modulation for a second generation digital terrestrial television broadcasting system, (DVB-T2), June 2008.
- Non-Patent Literature 10
- L. Vangelista, N. Benvenuto, and S. Tomasin, “Key technologies for next-generation terrestrial digital television standard DVB-T2”, IEEE Commun. Magazine, vol. 47, no. 10, pp. 146-153, October 2009.
- Non-Patent Literature 11
- T. Ohgane, T. Nishimura, and Y. Ogawa, “Application of space division multiplexing and those performance in a MIMO channel”, IEICE Trans. Commun., vol. 88-B, no. 5, pp. 1843-1851, May 2005.
- Non-Patent Literature 12
- R. G. Gallager, “Low-density parity-check codes”, IRE Trans. Inform. Theory, IT-8, pp. 21-28, 1962.
- Non-Patent Literature 13
- D. J. C. Mackay, “Good error-correcting codes based on very sparse matrices”, IEEE Trans. Inform. Theory, vol. 45, no. 2, pp. 399-431, March 1999.
- Non-Patent Literature 14
- ETSI EN 302 307, “Second generation framing structure, channel coding and modulation systems for broadcasting, interactive services, news gathering and other broadband satellite applications”, v. 1.1.2, June 2006.
- Non-Patent Literature 15
- Y.-L. Ueng, and C.-C. Cheng, “A fast-convergence decoding method and memory-efficient VLSI decoder architecture for irregular LDPC codes in the IEEE 802.16e standards”, IEEE VTC-2007 Fall, pp. 1255-1259.

It is an object of the present invention to provide a MIMO system that improves reception quality in an LOS environment.

To solve the above problem, the present invention provides a precoding method for generating, from a plurality of signals which are based on a selected modulation scheme and represented by in-phase components and quadrature components, a plurality of precoded signals that are transmitted in the same frequency bandwidth at the same time and transmitting the generated precoded signals, the precoding method comprising: selecting one precoding weight matrix from among a plurality of precoding weight matrices by regularly hopping between the matrices; and generating the plurality of precoded signals by multiplying the selected precoding weight matrix by the plurality of signals which are based on the selected modulation scheme, the plurality of precoding weight matrices being nine matrices expressed, using a positive real number α, as Equations 339 through 347 (details are described below).

According to each aspect of the above invention, precoded signals, which are generated by precoding signals by using one precoding weight matrix selected from among a plurality of precoding weight matrices by regularly hopping between the matrices, are transmitted and received. Thus the precoding weight matrix used in the precoding is any of a plurality of precoding weight matrices that have been predetermined. This makes it possible to improve the reception quality in an LOS environment based on the design of the plurality of precoding weight matrices.

With the above structure, the present invention provides a precoding method, a precoding device, a transmission method, a reception method, a transmission device, and a reception device that remedy degradation of reception quality in an LOS environment, thereby providing high-quality service to LOS users during broadcast or multicast communication.

The following describes embodiments of the present invention with reference to the drawings.

The following describes the transmission scheme, transmission device, reception scheme, and reception device of the present embodiment.

Prior to describing the present embodiment, an overview is provided of a transmission scheme and decoding scheme in a conventional spatial multiplexing MIMO system.

FIG. 1 shows the structure of an N_{t}×N_{r }spatial multiplexing MIMO system. An information vector z is encoded and interleaved. As output of the interleaving, an encoded bit vector u=(u_{1}, . . . , u_{Nt}) is acquired. Note that u_{i}=(u_{i1}, . . . , u_{iM}) (where M is the number of transmission bits per symbol). Letting the transmission vector s=(s_{1}, . . . , s_{Nt})^{T }and the transmission signal from transmit antenna #**1** be represented as s_{i}=map(u_{i}), the normalized transmission energy is represented as E{|s_{i}|^{2}}=Es/Nt (E_{s }being the total energy per channel). Furthermore, letting the received vector be y=(y_{1}, . . . , y_{Nr})^{T}, the received vector is represented as in Equation 1.

In this Equation, H_{NtNr }is the channel matrix, n=(n_{1}, . . . , n_{Nr})^{T }is the noise vector, and n_{i }is the i.i.d. complex Gaussian random noise with an average value 0 and variance σ^{2}. From the relationship between transmission symbols and reception symbols that is induced at the reception device, the probability for the received vector may be provided as a multi-dimensional Gaussian distribution, as in Equation 2.

Here, a reception device that performs iterative decoding composed of an outer soft-in/soft-out decoder and a MIMO detector, as in FIG. 1, is considered. The vector of a log-likelihood ratio (L-value) in FIG. 1 is represented as in Equations 3-5.

<Iterative Detection Scheme>

The following describes iterative detection of MIMO signals in the N_{t}×N_{r }spatial multiplexing MIMO system.

The log-likelihood ratio of u_{mn }is defined as in Equation 6.

From Bayes' theorem, Equation 6 can be expressed as Equation 7.

Let U_{mn,±1}={u|u_{mn}=±1}. When approximating ln Σa_{j}˜max ln a_{j}, an approximation of Equation 7 can be sought as Equation 8. Note that the above symbol “˜” indicates approximation.

P(u|u_{mn}) and ln P(u|u_{mn}) in Equation 8 are represented as follows.

Incidentally, the logarithmic probability of the equation defined in Equation 2 is represented in Equation 12.

Accordingly, from Equations 7 and 13, in MAP or A Posteriori Probability (APP), the a posteriori L-value is represented as follows.

Hereinafter, this is referred to as iterative APP decoding. From Equations 8 and 12, in the log-likelihood ratio utilizing Max-Log approximation (Max-Log APP), the a posteriori L-value is represented as follows.

Hereinafter, this is referred to as iterative Max-log APP decoding. The extrinsic information required in an iterative decoding system can be sought by subtracting prior inputs from Equations 13 and 14.

<System Model>

FIG. 28 shows the basic structure of the system that is related to the subsequent description. This system is a 2×2 spatial multiplexing MIMO system. There is an outer encoder for each of streams A and B. The two outer encoders are identical LDPC encoders. (Here, a structure using LDPC encoders as the outer encoders is described as an example, but the error correction coding used by the outer encoder is not limited to LDPC coding. The present invention may similarly be embodied using other error correction coding such as turbo coding, convolutional coding, LDPC convolutional coding, and the like. Furthermore, each outer encoder is described as having a transmit antenna, but the outer encoders are not limited to this structure. A plurality of transmit antennas may be used, and the number of outer encoders may be one. Also, a greater number of outer encoders may be used than the number of transmit antennas.) The streams A and B respectively have interleavers (π_{a}, π_{b}). Here, the modulation scheme is 2^{h}-QAM (with h bits transmitted in one symbol).

The reception device performs iterative detection on the above MIMO signals (iterative APP (or iterative Max-log APP) decoding). Decoding of LDPC codes is performed by, for example, sum-product decoding.

FIG. 2 shows a frame structure and lists the order of symbols after interleaving. In this case, (i_{a}, j_{a}), (i_{b}, j_{b}) are represented by the following Equations.

Math 16

(*i*_{a}*,j*_{a})=π_{a}(Ω_{ia,ja}^{a}) Equation 16

Math 17

(*i*_{b}*,j*_{b})=π_{b}(Ω_{ib,jb}^{a}) Equation 17

In this case, i^{a}, i^{b }indicate the order of symbols after interleaving, j^{a}, j^{b }indicate the bit positions (j^{a}, j^{b}=1, . . . , h) in the modulation scheme, π^{a}, π^{b }indicate the interleavers for the streams A and B, and Ω_{ia,ja}^{a}, Ω_{ib,jb}^{b }indicate the order of data in streams A and B before interleaving. Note that FIG. 2 shows the frame structure for i_{a}=i_{b}.

<Iterative Decoding>

The following is a detailed description of the algorithms for sum-product decoding used in decoding of LDPC codes and for iterative detection of MIMO signals in the reception device.

Sum-Product Decoding

Let a two-dimensional M×N matrix H={H_{mn}} be the check matrix for LDPC codes that are targeted for decoding. Subsets A(m), B(n) of the set [1, N]={1, 2, . . . , N} are defined by the following Equations.

Math 18

*A*(*m*)≡{*n:H*_{mn}=1} Equation 18

Math 19

*B*(*n*)≡{*m:H*_{mn}=1} Equation 19

In these Equations, A(m) represents the set of column indices of 1's in the m^{th }column of the check matrix H, and B(n) represents the set of row indices of 1's in the n^{th }row of the check matrix H. The algorithm for sum-product decoding is as follows.

Step A•1 (initialization): let a priori value log-likelihood ratio β_{mn}=0 for all combinations (m, n) satisfying H_{mn}=1. Assume that the loop variable (the number of iterations) l_{sum}=1 and the maximum number of loops is set to l_{sum, max}.

Step A•2 (row processing): the extrinsic value log-likelihood ratio α_{mn }is updated for all combinations (m, n) satisfying H_{mn}=1 in the order of m=1, 2, . . . , M, using the following updating Equations.

In these Equations, f represents a Gallager function. Furthermore, the scheme of seeking λ_{n }is described in detail later.

Step A•3 (column processing): the extrinsic value log-likelihood ratio β_{mn }is updated for all combinations (m, n) satisfying H_{mn}=1 in the order of n=1, 2, . . . , N, using the following updating Equation.

Step A•4 (calculating a log-likelihood ratio): the log-likelihood ratio L_{n }is sought for nε[1, N] by the following Equation.

Step A•5 (count of the number of iterations): if l_{sum}<l_{sum, max}, then l_{sum }is incremented, and processing returns to step A•2. If l_{sum}=L_{sum, max}, the sum-product decoding in this round is finished.

The operations in one sum-product decoding have been described. Subsequently, iterative MIMO signal detection is performed. In the variables m, n, α_{mn}, β_{mn}, λ_{n}, and L_{n}, used in the above description of the operations of sum-product decoding, the variables in stream A are m_{a}, n_{a}, α_{mana}^{a}, β_{mana}^{a}, λ_{na}, and L_{na}, and the variables in stream B are m_{b}, n_{b}, α_{mbnb}^{b}, β_{mbnb}^{b}, λ_{nb}, and L_{nb}.

<Iterative MIMO Signal Detection>

The following describes the scheme of seeking λ_{n }in iterative MIMO signal detection in detail.

The following Equation holds from Equation 1.

The following Equations are defined from the frame structures of FIG. 2 and from Equations 16 and 17.

Math 26

*n*_{a}=Ω_{ia,ja}^{a} Equation 26

Math 27

*n*_{b}=Ω_{ib,jb}^{b} Equation 27

In this case, n_{a},n_{b}ε[1, N]. Hereinafter, λ_{na}, L_{na}, λ_{nb}, and L_{nb}, where the number of iterations of iterative MIMO signal detection is k, are represented as λ_{k, na}, L_{k, na}, λ_{k, nb}, and L_{k, nb}.

Step B•1 (initial detection; k=0): λ_{0, na }and λ_{0, nb }are sought as follows in the case of initial detection.

In iterative APP decoding:

In iterative Max-log APP decoding:

Here, let X=a, b. Then, assume that the number of iterations of iterative MIMO signal detection is l_{mimo}=0 and the maximum number of iterations is set to l_{mimo, max}.

Step B•2 (iterative detection; the number of iterations k): λ_{k, na }and λ_{k, nb}, where the number of iterations is k, are represented as in Equations 31-34, from Equations 11, 13-15, 16, and 17. Let (X, Y)=(a, b)(b, a).

In iterative APP decoding:

In iterative Max-log APP decoding:

Step B•3 (counting the number of iterations and estimating a codeword): increment l_{mimo }if l_{mimo}<l_{mimo, max}, and return to step B•2. Assuming that l_{mimo}=l_{mimo, max}, the estimated codeword is sought as in the following Equation.

Here, let X=a, b.

FIG. 3 is an example of the structure of a transmission device **300** in the present embodiment. An encoder **302**A receives information (data) **301**A and a frame structure signal **313** as inputs and, in accordance with the frame structure signal **313**, performs error correction coding such as convolutional coding, LDPC coding, turbo coding, or the like, outputting encoded data **303**A. (The frame structure signal **313** includes information such as the error correction scheme used for error correction coding of data, the coding rate, the block length, and the like. The encoder **302**A uses the error correction scheme indicated by the frame structure signal **313**. Furthermore, the error correction scheme may be hopped.)

An interleaver **304**A receives the encoded data **303**A and the frame structure signal **313** as inputs and performs interleaving, i.e. changing the order of the data, to output interleaved data **305**A. (The scheme of interleaving may be hopped based on the frame structure signal **313**.)

A mapping unit **306**A receives the interleaved data **305**A and the frame structure signal **313** as inputs, performs modulation such as Quadrature Phase Shift Keying (QPSK), 16 Quadrature Amplitude Modulation (16QAM), 64 Quadrature Amplitude Modulation (64QAM), or the like, and outputs a resulting baseband signal **307**A. (The modulation scheme may be hopped based on the frame structure signal **313**.)

FIGS. 24A and 24B are an example of a mapping scheme over an I-Q plane, having an in-phase component I and a quadrature component Q, to form a baseband signal in QPSK modulation. For example, as shown in FIG. 24A, if the input data is “00”, the output is I=1.0, Q=1.0. Similarly, for input data of “01”, the output is I=1.0, Q=1.0, and so forth. FIG. 24B is an example of a different scheme of mapping in an I-Q plane for QPSK modulation than FIG. 24A. The difference between FIG. 24B and FIG. 24A is that the signal points in FIG. 24A have been rotated around the origin to yield the signal points of FIG. 24B. Non-Patent Literature 9 and Non-Patent Literature 10 describe such a constellation rotation scheme, and the Cyclic Q Delay described in Non-Patent Literature 9 and Non-Patent Literature 10 may also be adopted. As another example apart from FIGS. 24A and 24B, FIGS. 25A and 25B show signal point layout in the I-Q plane for 16QAM. The example corresponding to FIG. 24A is shown in FIG. 25A, and the example corresponding to FIG. 24B is shown in FIG. 25B.

An encoder **302**B receives information (data) **301**B and the frame structure signal **313** as inputs and, in accordance with the frame structure signal **313**, performs error correction coding such as convolutional coding, LDPC coding, turbo coding, or the like, outputting encoded data **303**B. (The frame structure signal **313** includes information such as the error correction scheme used, the coding rate, the block length, and the like. The error correction scheme indicated by the frame structure signal **313** is used. Furthermore, the error correction scheme may be hopped.)

An interleaver **304**B receives the encoded data **303**B and the frame structure signal **313** as inputs and performs interleaving, i.e. changing the order of the data, to output interleaved data **305**B. (The scheme of interleaving may be hopped based on the frame structure signal **313**.)

A mapping unit **306**B receives the interleaved data **305**B and the frame structure signal **313** as inputs, performs modulation such as Quadrature Phase Shift Keying (QPSK), 16 Quadrature Amplitude Modulation (16QAM), 64 Quadrature Amplitude Modulation (64QAM), or the like, and outputs a resulting baseband signal **307**B. (The modulation scheme may be hopped based on the frame structure signal **313**.)

A weighting information generating unit **314** receives the frame structure signal **313** as an input and outputs information **315** regarding a weighting scheme based on the frame structure signal **313**. The weighting scheme is characterized by regular hopping between weights.

A weighting unit **308**A receives the baseband signal **307**A, the baseband signal **307**B, and the information **315** regarding the weighting scheme, and based on the information **315** regarding the weighting scheme, performs weighting on the baseband signal **307**A and the baseband signal **307**B and outputs a signal **309**A resulting from the weighting. Details on the weighting scheme are provided later.

A wireless unit **310**A receives the signal **309**A resulting from the weighting as an input and performs processing such as orthogonal modulation, band limiting, frequency conversion, amplification, and the like, outputting a transmission signal **311**A. A transmission signal **511**A is output as a radio wave from an antenna **312**A.

A weighting unit **308**B receives the baseband signal **307**A, the baseband signal **307**B, and the information **315** regarding the weighting scheme, and based on the information **315** regarding the weighting scheme, performs weighting on the baseband signal **307**A and the baseband signal **307**B and outputs a signal **309**B resulting from the weighting.

FIG. 26 shows the structure of a weighting unit. The baseband signal **307**A is multiplied by w**11**(*t*), yielding w**11**(*t*)s**1**(*t*), and is multiplied by w**21**(*t*), yielding w**21**(*t*)s**1**(*t*). Similarly, the baseband signal **307**B is multiplied by w**12**(*t*) to generate w**12**(*t*)s**2**(*t*) and is multiplied by w**22**(*t*) to generate w**22**(*t*)s**2**(*t*). Next, z**1**(*t*)=w**11**(*t*)s**1**(*t*)+w**12**(*t*) s**2**(*t*) and z**2**(*t*)=w**21**(*t*)s**1**(*t*)+w**22**(*t*)s**2**(*t*) are obtained.

Details on the weighting scheme are provided later.

A wireless unit **310**B receives the signal **309**B resulting from the weighting as an input and performs processing such as orthogonal modulation, band limiting, frequency conversion, amplification, and the like, outputting a transmission signal **311**B. A transmission signal **511**B is output as a radio wave from an antenna **312**B.

FIG. 4 shows an example of the structure of a transmission device **400** that differs from FIG. 3. The differences in FIG. 4 from FIG. 3 are described.

An encoder **402** receives information (data) **401** and the frame structure signal **313** as inputs and, in accordance with the frame structure signal **313**, performs error correction coding and outputs encoded data **402**.

A distribution unit **404** receives the encoded data **403** as an input, distributes the data **403**, and outputs data **405**A and data **405**B. Note that in FIG. 4, one encoder is shown, but the number of encoders is not limited in this way. The present invention may similarly be embodied when the number of encoders is m (where m is an integer greater than or equal to one) and the distribution unit divides encoded data generated by each encoder into two parts and outputs the divided data.

FIG. 5 shows an example of a frame structure in the time domain for a transmission device according to the present embodiment. A symbol **500**_**1** is a symbol for notifying the reception device of the transmission scheme. For example, the symbol **500**_**1** conveys information such as the error correction scheme used for transmitting data symbols, the coding rate, and the modulation scheme used for transmitting data symbols.

The symbol **501**_**1** is for estimating channel fluctuation for the modulated signal z**1**(*t*) (where t is time) transmitted by the transmission device. The symbol **502**_**1** is the data symbol transmitted as symbol number u (in the time domain) by the modulated signal z**1**(*t*), and the symbol **503**_**1** is the data symbol transmitted as symbol number u+1 by the modulated signal z**1**(*t*).

The symbol **501**_**2** is for estimating channel fluctuation for the modulated signal z**2**(*t*) (where t is time) transmitted by the transmission device. The symbol **502**_**2** is the data symbol transmitted as symbol number u by the modulated signal z**2**(*t*), and the symbol **503**_**2** is the data symbol transmitted as symbol number u+1 by the modulated signal z**2**(*t*).

The following describes the relationships between the modulated signals z**1**(*t*) and z**2**(*t*) transmitted by the transmission device and the received signals r**1**(*t*) and r**2**(*t*) received by the reception device.

In FIG. 5, **504**#**1** and **504**#**2** indicate transmit antennas in the transmission device, and **505**#**1** and **505**#**2** indicate receive antennas in the reception device. The transmission device transmits the modulated signal z**1**(*t*) from transmit antenna **504**#**1** and transmits the modulated signal z**2**(*t*) from transmit antenna **504**#**2**. In this case, the modulated signal z**1**(*t*) and the modulated signal z**2**(*t*) are assumed to occupy the same (a shared/common) frequency (bandwidth). Letting the channel fluctuation for the transmit antennas of the transmission device and the antennas of the reception device be h_{11}(t), h_{12}(t), h_{21}(t), and h_{22}(t), the signal received by the receive antenna **505**#**1** of the reception device be r**1**(*t*), and the signal received by the receive antenna **505**#**2** of the reception device be r**2**(*t*), the following relationship holds.

FIG. 6 relates to the weighting scheme (precoding scheme) in the present embodiment. A weighting unit **600** integrates the weighting units **308**A and **308**B in FIG. 3. As shown in FIG. 6, a stream s**1**(*t*) and a stream s**2**(*t*) correspond to the baseband signals **307**A and **307**B in FIG. 3. In other words, the streams s**1**(*t*) and s**2**(*t*) are the baseband signal in-phase components I and quadrature components Q when mapped according to a modulation scheme such as QPSK, 16QAM, 64QAM, or the like. As indicated by the frame structure of FIG. 6, the stream s**1**(*t*) is represented as s**1**(*u*) at symbol number u, as s**1**(*u+*1) at symbol number u+1, and so forth. Similarly, the stream s**2**(*t*) is represented as s**2**(*u*) at symbol number u, as s**2**(*u+*1) at symbol number u+1, and so forth. The weighting unit **600** receives the baseband signals **307**A (s**1**(*t*)) and **307**B (s**2**(*t*)) and the information **315** regarding weighting information in FIG. 3 as inputs, performs weighting in accordance with the information **315** regarding weighting, and outputs the signals **309**A (z**1**(*t*)) and **309**B (z**2**(*t*)) after weighting in FIG. 3. In this case, z**1**(*t*) and z**2**(*t*) are represented as follows.

For symbol number 4i (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number 4i+1:

For symbol number 4i+2:

For symbol number 4i+3:

In this way, the weighting unit in FIG. 6 regularly hops between precoding weights over a four-slot period (cycle). (While precoding weights have been described as being hopped between regularly over four slots, the number of slots for regular hopping is not limited to four.)

Incidentally, Non-Patent Literature 4 describes hopping the precoding weights for each slot. This hopping of precoding weights is characterized by being random. On the other hand, in the present embodiment, a certain period (cycle) is provided, and the precoding weights are hopped between regularly. Furthermore, in each 2×2 precoding weight matrix composed of four precoding weights, the absolute value of each of the four precoding weights is equivalent to (1/sqrt(2)), and hopping is regularly performed between precoding weight matrices having this characteristic.

In an LOS environment, if a special precoding matrix is used, reception quality may greatly improve, yet the special precoding matrix differs depending on the conditions of direct waves. In an LOS environment, however, a certain tendency exists, and if precoding matrices are hopped between regularly in accordance with this tendency, the reception quality of data greatly improves. On the other hand, when precoding matrices are hopped between at random, a precoding matrix other than the above-described special precoding matrix may exist, and the possibility of performing precoding only with biased precoding matrices that are not suitable for the LOS environment also exists. Therefore, in an LOS environment, excellent reception quality may not always be obtained. Accordingly, there is a need for a precoding hopping scheme suitable for an LOS environment. The present invention proposes such a precoding scheme.

FIG. 7 is an example of the structure of a reception device **700** in the present embodiment. A wireless unit **703**_X receives, as an input, a received signal **702**_X received by an antenna **701**_X, performs processing such as frequency conversion, quadrature demodulation, and the like, and outputs a baseband signal **704**_X.

A channel fluctuation estimating unit **705**_**1** for the modulated signal z**1** transmitted by the transmission device receives the baseband signal **704**_X as an input, extracts a reference symbol **501**_**1** for channel estimation as in FIG. 5, estimates a value corresponding to h_{11 }in Equation 36, and outputs a channel estimation signal **706**_**1**.

A channel fluctuation estimating unit **705**_**2** for the modulated signal z**2** transmitted by the transmission device receives the baseband signal **704**_X as an input, extracts a reference symbol **501**_**2** for channel estimation as in FIG. 5, estimates a value corresponding to h_{12 }in Equation 36, and outputs a channel estimation signal **706**_**2**.

A wireless unit **703**_Y receives, as input, a received signal **702**_Y received by an antenna **701**_Y, performs processing such as frequency conversion, quadrature demodulation, and the like, and outputs a baseband signal **704**_Y.

A channel fluctuation estimating unit **707**_**1** for the modulated signal z**1** transmitted by the transmission device receives the baseband signal **704**_Y as an input, extracts a reference symbol **501**_**1** for channel estimation as in FIG. 5, estimates a value corresponding to h_{21 }in Equation 36, and outputs a channel estimation signal **708**_**1**.

A channel fluctuation estimating unit **707**_**2** for the modulated signal z**2** transmitted by the transmission device receives the baseband signal **704**_Y as an input, extracts a reference symbol **501**_**2** for channel estimation as in FIG. 5, estimates a value corresponding to h_{22 }in Equation 36, and outputs a channel estimation signal **708**_**2**.

A control information decoding unit **709** receives the baseband signal **704**_X and the baseband signal **704**_Y as inputs, detects the symbol **500**_**1** that indicates the transmission scheme as in FIG. 5, and outputs a signal **710** regarding information on the transmission scheme indicated by the transmission device.

A signal processing unit **711** receives, as inputs, the baseband signals **704**_X and **704**_Y, the channel estimation signals **706**_**1**, **706**_**2**, **708**_**1**, and **708**_**2**, and the signal **710** regarding information on the transmission scheme indicated by the transmission device, performs detection and decoding, and outputs received data **712**_**1** and **712**_**2**.

Next, operations by the signal processing unit **711** in FIG. 7 are described in detail. FIG. 8 is an example of the structure of the signal processing unit **711** in the present embodiment. FIG. 8 shows an INNER MIMO detector, a soft-in/soft-out decoder, and a weighting coefficient generating unit as the main elements. Non-Patent Literature 2 and Non-Patent Literature 3 describe the scheme of iterative decoding with this structure. The MIMO system described in Non-Patent Literature 2 and Non-Patent Literature 3 is a spatial multiplexing MIMO system, whereas the present embodiment differs from Non-Patent Literature 2 and Non-Patent Literature 3 by describing a MIMO system that changes precoding weights with time. Letting the (channel) matrix in Equation 36 be H(t), the precoding weight matrix in FIG. 6 be W(t) (where the precoding weight matrix changes over t), the received vector be R(t)=(r**1**(*t*),r**2**(*t*))^{T}, and the stream vector be S(t)=(s**1**(*t*),s**2**(*t*))^{T}, the following Equation holds.

Math 41

*R*(*t*)=*H*(*t*)*W*(*t*)*S*(*t*) Equation 41

In this case, the reception device can apply the decoding scheme in Non-Patent Literature 2 and Non-Patent Literature 3 to the received vector R(t) by considering H(t)W(t) as the channel matrix.

Therefore, a weighting coefficient generating unit **819** in FIG. 8 receives, as input, a signal **818** regarding information on the transmission scheme indicated by the transmission device (corresponding to **710** in FIG. 7) and outputs a signal **820** regarding information on weighting coefficients.

An INNER MIMO detector **803** receives the signal **820** regarding information on weighting coefficients as input and, using the signal **820**, performs the calculation in Equation 41. Iterative detection and decoding is thus performed. The following describes operations thereof.

In the signal processing unit in FIG. 8, a processing scheme such as that shown in FIG. 10 is necessary for iterative decoding (iterative detection). First, one codeword (or one frame) of the modulated signal (stream) s**1** and one codeword (or one frame) of the modulated signal (stream) s**2** are decoded. As a result, the Log-Likelihood Ratio (LLR) of each bit of the one codeword (or one frame) of the modulated signal (stream) s**1** and of the one codeword (or one frame) of the modulated signal (stream) s**2** is obtained from the soft-in/soft-out decoder. Detection and decoding is performed again using the LLR. These operations are performed multiple times (these operations being referred to as iterative decoding (iterative detection)). Hereinafter, description focuses on the scheme of generating the log-likelihood ratio (LLR) of a symbol at a particular time in one frame.

In FIG. 8, a storage unit **815** receives, as inputs, a baseband signal **801**X (corresponding to the baseband signal **704**_X in FIG. 7), a channel estimation signal group **802**X (corresponding to the channel estimation signals **706**_**1** and **706**_**2** in FIG. 7), a baseband signal **801**Y (corresponding to the baseband signal **704**_Y in FIG. 7), and a channel estimation signal group **802**Y (corresponding to the channel estimation signals **708**_**1** and **708**_**2** in FIG. 7). In order to achieve iterative decoding (iterative detection), the storage unit **815** calculates H(t)W(t) in Equation 41 and stores the calculated matrix as a transformed channel signal group. The storage unit **815** outputs the above signals when necessary as a baseband signal **816**X, a transformed channel estimation signal group **817**X, a baseband signal **816**Y, and a transformed channel estimation signal group **817**Y.

Subsequent operations are described separately for initial detection and for iterative decoding (iterative detection).

<Initial Detection>

The INNER MIMO detector **803** receives, as inputs, the baseband signal **801**X, the channel estimation signal group **802**X, the baseband signal **801**Y, and the channel estimation signal group **802**Y. Here, the modulation scheme for the modulated signal (stream) s**1** and the modulated signal (stream) s**2** is described as 16QAM.

The INNER MIMO detector **803** first calculates H(t)W(t) from the channel estimation signal group **802**X and the channel estimation signal group **802**Y to seek candidate signal points corresponding to the baseband signal **801**X. FIG. 11 shows such calculation. In FIG. 11, each black dot (•) is a candidate signal point in the I-Q plane. Since the modulation scheme is 16QAM, there are 256 candidate signal points. (Since FIG. 11 is only for illustration, not all 256 candidate signal points are shown.) Here, letting the four bits transferred by modulated signal s**1** be b**0**, b**1**, b**2**, and b**3**, and the four bits transferred by modulated signal s**2** be b**4**, b**5**, b**6**, and b**7**, candidate signal points corresponding to (b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**) in FIG. 11 exist. The squared Euclidian distance is sought between a received signal point **1101** (corresponding to the baseband signal **801**X) and each candidate signal point. Each squared Euclidian distance is divided by the noise variance σ^{2}. Accordingly, E_{X}(b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**), i.e. the value of the squared Euclidian distance between a candidate signal point corresponding to (b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**) and a received signal point, divided by the noise variance, is sought. Note that the baseband signals and the modulated signals s**1** and s**2** are each complex signals.

Similarly, H(t)W(t) is calculated from the channel estimation signal group **802**X and the channel estimation signal group **802**Y, candidate signal points corresponding to the baseband signal **801**Y are sought, the squared Euclidian distance for the received signal point (corresponding to the baseband signal **801**Y) is sought, and the squared Euclidian distance is divided by the noise variance σ^{2}. Accordingly, E_{y}(b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**), i.e. the value of the squared Euclidian distance between a candidate signal point corresponding to (b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**) and a received signal point, divided by the noise variance, is sought.

Then E_{X}(b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**)+E_{Y}(b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**)=E(b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**) is sought.

The INNER MIMO detector **803** outputs E(b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**) as a signal **804**.

A log-likelihood calculating unit **805**A receives the signal **804** as input, calculates the log likelihood for bits b**0**, b**1**, b**2**, and b**3**, and outputs a log-likelihood signal **806**A. Note that during calculation of the log likelihood, the log likelihood for “1” and the log likelihood for “0” are calculated. The calculation scheme is as shown in Equations 28, 29, and 30. Details can be found in Non-Patent Literature 2 and Non-Patent Literature 3.

Similarly, a log-likelihood calculating unit **805**B receives the signal **804** as input, calculates the log likelihood for bits b**4**, b**5**, b**6**, and b**7**, and outputs a log-likelihood signal **806**B.

A deinterleaver (**807**A) receives the log-likelihood signal **806**A as an input, performs deinterleaving corresponding to the interleaver (the interleaver (**304**A) in FIG. 3), and outputs a deinterleaved log-likelihood signal **808**A.

Similarly, a deinterleaver (**807**B) receives the log-likelihood signal **806**B as an input, performs deinterleaving corresponding to the interleaver (the interleaver (**304**B) in FIG. 3), and outputs a deinterleaved log-likelihood signal **808**B.

A log-likelihood ratio calculating unit **809**A receives the interleaved log-likelihood signal **808**A as an input, calculates the log-likelihood ratio (LLR) of the bits encoded by the encoder **302**A in FIG. 3, and outputs a log-likelihood ratio signal **810**A.

Similarly, a log-likelihood ratio calculating unit **809**B receives the interleaved log-likelihood signal **808**B as an input, calculates the log-likelihood ratio (LLR) of the bits encoded by the encoder **302**B in FIG. 3, and outputs a log-likelihood ratio signal **810**B.

A soft-in/soft-out decoder **811**A receives the log-likelihood ratio signal **810**A as an input, performs decoding, and outputs a decoded log-likelihood ratio **812**A.

Similarly, a soft-in/soft-out decoder **811**B receives the log-likelihood ratio signal **810**B as an input, performs decoding, and outputs a decoded log-likelihood ratio **812**B.

<Iterative Decoding (Iterative Detection), Number of Iterations k>

An interleaver (**813**A) receives the log-likelihood ratio **812**A decoded by the soft-in/soft-out decoder in the (k−1)^{th }iteration as an input, performs interleaving, and outputs an interleaved log-likelihood ratio **814**A. The interleaving pattern in the interleaver (**813**A) is similar to the interleaving pattern in the interleaver (**304**A) in FIG. 3.

An interleaver (**813**B) receives the log-likelihood ratio **812**B decoded by the soft-in/soft-out decoder in the (k−1)^{th }iteration as an input, performs interleaving, and outputs an interleaved log-likelihood ratio **814**B. The interleaving pattern in the interleaver (**813**B) is similar to the interleaving pattern in the interleaver (**304**B) in FIG. 3.

The INNER MIMO detector **803** receives, as inputs, the baseband signal **816**X, the transformed channel estimation signal group **817**X, the baseband signal **816**Y, the transformed channel estimation signal group **817**Y, the interleaved log-likelihood ratio **814**A, and the interleaved log-likelihood ratio **814**B. The reason for using the baseband signal **816**X, the transformed channel estimation signal group **817**X, the baseband signal **816**Y, and the transformed channel estimation signal group **817**Y instead of the baseband signal **801**X, the channel estimation signal group **802**X, the baseband signal **801**Y, and the channel estimation signal group **802**Y is because a delay occurs due to iterative decoding.

The difference between operations by the INNER MIMO detector **803** for iterative decoding and for initial detection is the use of the interleaved log-likelihood ratio **814**A and the interleaved log-likelihood ratio **814**B during signal processing. The INNER MIMO detector **803** first seeks E(b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**), as during initial detection. Additionally, coefficients corresponding to Equations 11 and **32** are sought from the interleaved log-likelihood ratio **814**A and the interleaved log-likelihood ratio **914**B. The value E(b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**) is adjusted using the sought coefficients, and the resulting value E′(b**0**, b**1**, b**2**, b**3**, b**4**, b**5**, b**6**, b**7**) is output as the signal **804**.

The log-likelihood calculating unit **805**A receives the signal **804** as input, calculates the log likelihood for bits b**0**, b**1**, b**2**, and b**3**, and outputs the log-likelihood signal **806**A. Note that during calculation of the log likelihood, the log likelihood for “1” and the log likelihood for “0” are calculated. The calculation scheme is as shown in Equations 31, 32, 33, 34, and 35. Details can be found in Non-Patent Literature 2 and Non-Patent Literature 3.

Similarly, the log-likelihood calculating unit **805**B receives the signal **804** as input, calculates the log likelihood for bits b**4**, b**5**, b**6**, and b**7**, and outputs the log-likelihood signal **806**B. Operations by the deinterleaver onwards are similar to initial detection.

Note that while FIG. 8 shows the structure of the signal processing unit when performing iterative detection, iterative detection is not always essential for obtaining excellent reception quality, and a structure not including the interleavers **813**A and **813**B, which are necessary only for iterative detection, is possible. In such a case, the INNER MIMO detector **803** does not perform iterative detection.

The main part of the present embodiment is calculation of H(t)W(t). Note that as shown in Non-Patent Literature 5 and the like, QR decomposition may be used to perform initial detection and iterative detection.

Furthermore, as shown in Non-Patent Literature 11, based on H(t)W(t), linear operation of the Minimum Mean Squared Error (MMSE) and Zero Forcing (ZF) may be performed in order to perform initial detection.

FIG. 9 is the structure of a different signal processing unit than FIG. 8 and is for the modulated signal transmitted by the transmission device in FIG. 4. The difference with FIG. 8 is the number of soft-in/soft-out decoders. A soft-in/soft-out decoder **901** receives, as inputs, the log-likelihood ratio signals **810**A and **810**B, performs decoding, and outputs a decoded log-likelihood ratio **902**. A distribution unit **903** receives the decoded log-likelihood ratio **902** as an input and distributes the log-likelihood ratio **902**. Other operations are similar to FIG. 8.

FIGS. 12A and 12B show BER characteristics for a transmission scheme using the precoding weights of the present embodiment under similar conditions to FIGS. 29A and 29B. FIG. 12A shows the BER characteristics of Max-log A Posteriori Probability (APP) without iterative detection (see Non-Patent Literature 1 and Non-Patent Literature 2), and FIG. 12B shows the BER characteristics of Max-log-APP with iterative detection (see Non-Patent Literature 1 and Non-Patent Literature 2) (number of iterations: five). Comparing FIGS. 12A, **12**B, **29**A, and **29**B shows how if the transmission scheme of the present embodiment is used, the BER characteristics when the Rician factor is large greatly improve over the BER characteristics when using spatial multiplexing MIMO system, thereby confirming the usefulness of the scheme in the present embodiment.

As described above, when a transmission device transmits a plurality of modulated signals from a plurality of antennas in a MIMO system, the advantageous effect of improved transmission quality, as compared to conventional spatial multiplexing MIMO system, is achieved in an LOS environment in which direct waves dominate by hopping between precoding weights regularly over time, as in the present embodiment.

In the present embodiment, and in particular with regards to the structure of the reception device, operations have been described for a limited number of antennas, but the present invention may be embodied in the same way even if the number of antennas increases. In other words, the number of antennas in the reception device does not affect the operations or advantageous effects of the present embodiment. Furthermore, in the present embodiment, the example of LDPC coding has particularly been explained, but the present invention is not limited to LDPC coding. Furthermore, with regards to the decoding scheme, the soft-in/soft-out decoders are not limited to the example of sum-product decoding. Another soft-in/soft-out decoding scheme may be used, such as a BCJR algorithm, a SOYA algorithm, a Max-log-MAP algorithm, and the like. Details are provided in Non-Patent Literature 6.

Additionally, in the present embodiment, the example of a single carrier scheme has been described, but the present invention is not limited in this way and may be similarly embodied for multi-carrier transmission. Accordingly, when using a scheme such as spread spectrum communication, Orthogonal Frequency-Division Multiplexing (OFDM), Single Carrier Frequency Division Multiple Access (SC-FDMA), Single Carrier Orthogonal Frequency-Division Multiplexing (SC-OFDM), or wavelet OFDM as described in Non-Patent Literature 7 and the like, for example, the present invention may be similarly embodied. Furthermore, in the present embodiment, symbols other than data symbols, such as pilot symbols (preamble, unique word, and the like), symbols for transmission of control information, and the like, may be arranged in the frame in any way.

The following describes an example of using OFDM as an example of a multi-carrier scheme.

FIG. 13 shows the structure of a transmission device when using OFDM. In FIG. 13, elements that operate in a similar way to FIG. 3 bear the same reference signs.

An OFDM related processor **1301**A receives, as input, the weighted signal **309**A, performs processing related to OFDM, and outputs a transmission signal **1302**A. Similarly, an OFDM related processor **1301**B receives, as input, the weighted signal **309**B, performs processing related to OFDM, and outputs a transmission signal **1302**B.

FIG. 14 shows an example of a structure from the OFDM related processors **1301**A and **1301**B in FIG. 13 onwards. The part from **1401**A to **1410**A is related to the part from **1301**A to **312**A in FIG. 13, and the part from **1401**B to **1410**B is related to the part from **1301**B to **312**B in FIG. 13.

A serial/parallel converter **1402**A performs serial/parallel conversion on a weighted signal **1401**A (corresponding to the weighted signal **309**A in FIG. 13) and outputs a parallel signal **1403**A.

A reordering unit **1404**A receives a parallel signal **1403**A as input, performs reordering, and outputs a reordered signal **1405**A. Reordering is described in detail later.

An inverse fast Fourier transformer **1406**A receives the reordered signal **1405**A as an input, performs a fast Fourier transform, and outputs a fast Fourier transformed signal **1407**A.

A wireless unit **1408**A receives the fast Fourier transformed signal **1407**A as an input, performs processing such as frequency conversion, amplification, and the like, and outputs a modulated signal **1409**A. The modulated signal **1409**A is output as a radio wave from an antenna **1410**A.

A serial/parallel converter **1402**B performs serial/parallel conversion on a weighted signal **1401**B (corresponding to the weighted signal **309**B in FIG. 13) and outputs a parallel signal **1403**B.

A reordering unit **1404**B receives a parallel signal **1403**B as input, performs reordering, and outputs a reordered signal **1405**B. Reordering is described in detail later.

An inverse fast Fourier transformer **1406**B receives the reordered signal **1405**B as an input, performs a fast Fourier transform, and outputs a fast Fourier transformed signal **1407**B.

A wireless unit **1408**B receives the fast Fourier transformed signal **1407**B as an input, performs processing such as frequency conversion, amplification, and the like, and outputs a modulated signal **1409**B. The modulated signal **1409**B is output as a radio wave from an antenna **1410**B.

In the transmission device of FIG. 3, since the transmission scheme does not use multi-carrier, precoding hops to form a four-slot period (cycle), as shown in FIG. 6, and the precoded symbols are arranged in the time domain. When using a multi-carrier transmission scheme as in the OFDM scheme shown in FIG. 13, it is of course possible to arrange the precoded symbols in the time domain as in FIG. 3 for each (sub)carrier. In the case of a multi-carrier transmission scheme, however, it is possible to arrange symbols in the frequency domain, or in both the frequency and time domains. The following describes these arrangements.

FIGS. 15A and 15B show an example of a scheme of reordering symbols by reordering units **1401**A and **1401**B in FIG. 14, the horizontal axis representing frequency, and the vertical axis representing time. The frequency domain runs from (sub)carrier **0** through (sub)carrier **9**. The modulated signals z**1** and z**2** use the same frequency bandwidth at the same time. FIG. 15A shows the reordering scheme for symbols of the modulated signal z**1**, and FIG. 15B shows the reordering scheme for symbols of the modulated signal z**2**. Numbers #**1**, #**2**, #**3**, #**4**, . . . are assigned to in order to the symbols of the weighted signal **1401**A which is input into the serial/parallel converter **1402**A. At this point, symbols are assigned regularly, as shown in FIG. 15A. The symbols #**1**, #**2**, #**3**, #**4**, . . . are arranged in order starting from carrier **0**. The symbols #**1** through #**9** are assigned to time $**1**, and subsequently, the symbols #**10** through #**19** are assigned to time $**2**.

Similarly, numbers #**1**, #**2**, #**3**, #**4**, . . . are assigned in order to the symbols of the weighted signal **1401**B which is input into the serial/parallel converter **1402**B. At this point, symbols are assigned regularly, as shown in FIG. 15B. The symbols #**1**, #**2**, #**3**, #**4**, . . . are arranged in order starting from carrier **0**. The symbols #**1** through #**9** are assigned to time $**1**, and subsequently, the symbols #**10** through #**19** are assigned to time $**2**. Note that the modulated signals z**1** and z**2** are complex signals.

The symbol group **1501** and the symbol group **1502** shown in FIGS. 15A and 15B are the symbols for one period (cycle) when using the precoding weight hopping scheme shown in FIG. 6. Symbol #**0** is the symbol when using the precoding weight of slot 4i in FIG. 6. Symbol #**1** is the symbol when using the precoding weight of slot 4i+1 in FIG. 6. Symbol #**2** is the symbol when using the precoding weight of slot 4i+2 in FIG. 6. Symbol #**3** is the symbol when using the precoding weight of slot 4i+3 in FIG. 6. Accordingly, symbol #x is as follows. When x mod 4 is 0, the symbol #x is the symbol when using the precoding weight of slot 4i in FIG. **6**. When x mod 4 is 1, the symbol #x is the symbol when using the precoding weight of slot 4i+1 in FIG. 6. When x mod 4 is 2, the symbol #x is the symbol when using the precoding weight of slot 4i+2 in FIG. 6. When x mod 4 is 3, the symbol #x is the symbol when using the precoding weight of slot 4i+3 in FIG. 6.

In this way, when using a multi-carrier transmission scheme such as OFDM, unlike during single carrier transmission, symbols can be arranged in the frequency domain. Furthermore, the ordering of symbols is not limited to the ordering shown in FIGS. 15A and 15B. Other examples are described with reference to FIGS. 16A, **16**B, **17**A, and **17**B.

FIGS. 16A and 16B show an example of a scheme of reordering symbols by the reordering units **1404**A and **1404**B in FIG. 14, the horizontal axis representing frequency, and the vertical axis representing time, that differs from FIGS. 15A and 15B. FIG. 16A shows the reordering scheme for symbols of the modulated signal z**1**, and FIG. 16B shows the reordering scheme for symbols of the modulated signal z**2**. The difference in FIGS. 16A and 16B as compared to FIGS. 15A and 15B is that the reordering scheme of the symbols of the modulated signal z**1** differs from the reordering scheme of the symbols of the modulated signal z**2**. In FIG. 16B, symbols #**0** through #**5** are assigned to carriers **4** through **9**, and symbols #**6** through #**9** are assigned to carriers **0** through **3**. Subsequently, symbols #**10** through #**19** are assigned regularly in the same way. At this point, as in FIGS. 15A and 15B, the symbol group **1601** and the symbol group **1602** shown in FIGS. 16A and 16B are the symbols for one period (cycle) when using the precoding weight hopping scheme shown in FIG. 6.

FIGS. 17A and 17B show an example of a scheme of reordering symbols by the reordering units **1404**A and **1404**B in FIG. 14, the horizontal axis representing frequency, and the vertical axis representing time, that differs from FIGS. 15A and 15B. FIG. 17A shows the reordering scheme for symbols of the modulated signal z**1**, and FIG. 17B shows the reordering scheme for symbols of the modulated signal z**2**. The difference in FIGS. 17A and 17B as compared to FIGS. 15A and 15B is that whereas the symbols are arranged in order by carrier in FIGS. 15A and 15B, the symbols are not arranged in order by carrier in FIGS. 17A and 17B. It is obvious that, in FIGS. 17A and 17B, the reordering scheme of the symbols of the modulated signal z**1** may differ from the reordering scheme of the symbols of the modulated signal z**2**, as in FIGS. 16A and 16B.

FIGS. 18A and 18B show an example of a scheme of reordering symbols by the reordering units **1404**A and **1404**B in FIG. 14, the horizontal axis representing frequency, and the vertical axis representing time, that differs from FIGS. 15A through 17B. FIG. 18A shows the reordering scheme for symbols of the modulated signal z**1**, and FIG. 18B shows the reordering scheme for symbols of the modulated signal z**2**. In FIGS. 15A through 17B, symbols are arranged in the frequency domain, whereas in FIGS. 18A and 18B, symbols are arranged in both the frequency and time domains.

In FIG. 6, an example has been described of hopping between precoding weights over four slots. Here, however, an example of hopping over eight slots is described. The symbol groups **1801** and **1802** shown in FIGS. 18A and 18B are the symbols for one period (cycle) when using the precoding weight hopping scheme (and are therefore eight-symbol groups). Symbol #**0** is the symbol when using the precoding weight of slot 8i. Symbol #**1** is the symbol when using the precoding weight of slot 8i+1. Symbol #**2** is the symbol when using the precoding weight of slot 8i+2. Symbol #**3** is the symbol when using the precoding weight of slot 8i+3. Symbol #**4** is the symbol when using the precoding weight of slot 8i+4. Symbol #**5** is the symbol when using the precoding weight of slot 8i+5. Symbol #**6** is the symbol when using the precoding weight of slot 8i+6. Symbol #**7** is the symbol when using the precoding weight of slot 8i+7. Accordingly, symbol #x is as follows. When x mod 8 is 0, the symbol #x is the symbol when using the precoding weight of slot 8i. When x mod 8 is 1, the symbol #x is the symbol when using the precoding weight of slot 8i+1. When x mod 8 is 2, the symbol #x is the symbol when using the precoding weight of slot 8i+2. When x mod 8 is 3, the symbol #x is the symbol when using the precoding weight of slot 8i+3. When x mod 8 is 4, the symbol #x is the symbol when using the precoding weight of slot 8i+4. When x mod 8 is 5, the symbol #x is the symbol when using the precoding weight of slot 8i+5. When x mod 8 is 6, the symbol #x is the symbol when using the precoding weight of slot 8i+6. When x mod 8 is 7, the symbol #x is the symbol when using the precoding weight of slot 8i+7. In the symbol ordering in FIGS. 18A and 18B, four slots in the time domain and two slots in the frequency domain for a total of 4×2=8 slots are used to arrange symbols for one period (cycle). In this case, letting the number of symbols in one period (cycle) be m×n symbols (in other words, m×n precoding weights exist), the number of slots (the number of carriers) in the frequency domain used to arrange symbols in one period (cycle) be n, and the number of slots used in the time domain be m, then m>n should be satisfied. This is because the phase of direct waves fluctuates more slowly in the time domain than in the frequency domain. Therefore, since the precoding weights are changed in the present embodiment to minimize the influence of steady direct waves, it is preferable to reduce the fluctuation in direct waves in the period (cycle) for changing the precoding weights. Accordingly, m>n should be satisfied. Furthermore, considering the above points, rather than reordering symbols only in the frequency domain or only in the time domain, direct waves are more likely to become stable when symbols are reordered in both the frequency and the time domains as in FIGS. 18A and 18B, thereby making it easier to achieve the advantageous effects of the present invention. When symbols are ordered in the frequency domain, however, fluctuations in the frequency domain are abrupt, leading to the possibility of yielding diversity gain. Therefore, reordering in both the frequency and the time domains is not necessarily always the best scheme.

FIGS. 19A and 19B show an example of a scheme of reordering symbols by the reordering units **1404**A and **1404**B in FIG. 14, the horizontal axis representing frequency, and the vertical axis representing time, that differs from FIGS. 18A and 18B. FIG. 19A shows the reordering scheme for symbols of the modulated signal z**1**, and FIG. 19B shows the reordering scheme for symbols of the modulated signal z**2**. As in FIGS. 18A and 18B, FIGS. 19A and 19B show arrangement of symbols using both the frequency and the time axes. The difference as compared to FIGS. 18A and 18B is that, whereas symbols are arranged first in the frequency domain and then in the time domain in FIGS. 18A and 18B, symbols are arranged first in the time domain and then in the frequency domain in FIGS. 19A and 19B. In FIGS. 19A and 19B, the symbol group **1901** and the symbol group **1902** are the symbols for one period (cycle) when using the precoding hopping scheme.

Note that in FIGS. 18A, **18**B, **19**A, and **19**B, as in FIGS. 16A and 16B, the present invention may be similarly embodied, and the advantageous effect of high reception quality achieved, with the symbol arranging scheme of the modulated signal z**1** differing from the symbol arranging scheme of the modulated signal z**2**. Furthermore, in FIGS. 18A, **18**B, **19**A, and **19**B, as in FIGS. 17A and 17B, the present invention may be similarly embodied, and the advantageous effect of high reception quality achieved, without arranging the symbols in order.

FIG. 27 shows an example of a scheme of reordering symbols by the reordering units **1404**A and **1404**B in FIG. 14, the horizontal axis representing frequency, and the vertical axis representing time, that differs from the above examples. The case of hopping between precoding matrices regularly over four slots, as in Equations 37-40, is considered. The characteristic feature of FIG. 27 is that symbols are arranged in order in the frequency domain, but when progressing in the time domain, symbols are cyclically shifted by n symbols (in the example in FIG. 27, n=1). In the four symbols shown in the symbol group **2710** in the frequency domain in FIG. 27, precoding hops between the precoding matrices of Equations 37-40.

In this case, symbol #**0** is precoded using the precoding matrix in Equation 37, symbol #**1** is precoded using the precoding matrix in Equation 38, symbol #**2** is precoded using the precoding matrix in Equation 39, and symbol #**3** is precoded using the precoding matrix in Equation 40.

Similarly, for the symbol group **2720** in the frequency domain, symbol #**4** is precoded using the precoding matrix in Equation 37, symbol #**5** is precoded using the precoding matrix in Equation 38, symbol #**6** is precoded using the precoding matrix in Equation 39, and symbol #**7** is precoded using the precoding matrix in Equation 40.

For the symbols at time $**1**, precoding hops between the above precoding matrices, but in the time domain, symbols are cyclically shifted. Therefore, precoding hops between precoding matrices for the symbol groups **2701**, **2702**, **2703**, and **2704** as follows.

In the symbol group **2701** in the time domain, symbol #**0** is precoded using the precoding matrix in Equation 37, symbol #**9** is precoded using the precoding matrix in Equation 38, symbol #**18** is precoded using the precoding matrix in Equation 39, and symbol #**27** is precoded using the precoding matrix in Equation 40.

In the symbol group **2702** in the time domain, symbol #**28** is precoded using the precoding matrix in Equation 37, symbol #**1** is precoded using the precoding matrix in Equation 38, symbol #**10** is precoded using the precoding matrix in Equation 39, and symbol #**19** is precoded using the precoding matrix in Equation 40.

In the symbol group **2703** in the time domain, symbol #**20** is precoded using the precoding matrix in Equation 37, symbol #**29** is precoded using the precoding matrix in Equation 38, symbol #**2** is precoded using the precoding matrix in Equation 39, and symbol #**11** is precoded using the precoding matrix in Equation 40.

In the symbol group **2704** in the time domain, symbol #**12** is precoded using the precoding matrix in Equation 37, symbol #**21** is precoded using the precoding matrix in Equation 38, symbol #**30** is precoded using the precoding matrix in Equation 39, and symbol #**3** is precoded using the precoding matrix in Equation 40.

The characteristic of FIG. 27 is that, for example focusing on symbol #**11**, the symbols on either side in the frequency domain at the same time (symbols #**10** and #**12**) are both precoded with a different precoding matrix than symbol #**11**, and the symbols on either side in the time domain in the same carrier (symbols #**2** and #**20**) are both precoded with a different precoding matrix than symbol #**11**. This is true not only for symbol #**11**. Any symbol having symbols on either side in the frequency domain and the time domain is characterized in the same way as symbol #**11**. As a result, precoding matrices are effectively hopped between, and since the influence on stable conditions of direct waves is reduced, the possibility of improved reception quality of data increases.

In FIG. 27, the case of n=1 has been described, but n is not limited in this way. The present invention may be similarly embodied with n=3. Furthermore, in FIG. 27, when symbols are arranged in the frequency domain and time progresses in the time domain, the above characteristic is achieved by cyclically shifting the number of the arranged symbol, but the above characteristic may also be achieved by randomly (or regularly) arranging the symbols.

In Embodiment 1, regular hopping of the precoding weights as shown in FIG. 6 has been described. In the present embodiment, a scheme for designing specific precoding weights that differ from the precoding weights in FIG. 6 is described.

In FIG. 6, the scheme for hopping between the precoding weights in Equations 37-40 has been described. By generalizing this scheme, the precoding weights may be changed as follows. (The hopping period (cycle) for the precoding weights has four slots, and Equations are listed similarly to Equations 37-40.)

For symbol number 4i (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number 4i+1:

For symbol number 4i+2:

For symbol number 4i+3:

From Equations 36 and 41, the received vector R(t)=(r**1**(*t*), r**2**(*t*))^{T }can be represented as follows.

For symbol number 4i:

For symbol number 4i+1:

For symbol number 4i+2:

For symbol number 4i+3:

In this case, it is assumed that only components of direct waves exist in the channel elements h_{11}(t), h_{12}(t), h_{21}(t), and h_{22}(t), that the amplitude components of the direct waves are all equal, and that fluctuations do not occur over time. With these assumptions, Equations 46-49 can be represented as follows.

For symbol number 4i:

For symbol number 4i+1:

For symbol number 4i+2:

For symbol number 4i+3:

In Equations 50-53, let A be a positive real number and q be a complex number. The values of A and q are determined in accordance with the positional relationship between the transmission device and the reception device. Equations 50-53 can be represented as follows.

For symbol number 4i:

For symbol number 4i+1:

For symbol number 4i+2:

For symbol number 4i+3:

As a result, when q is represented as follows, a signal component based on one of s**1** and s**2** is no longer included in r**1** and r**2**, and therefore one of the signals s**1** and s**2** can no longer be obtained.

For symbol number 4i:

Math 58

*q=−A*_{e}^{j(θ}^{11}^{(4i)−θ}^{21}^{(4i))}*,−A*_{e}^{j(θ}^{11}^{(4i)−θ}^{21}^{(4i)−δ)} Equation 58

For symbol number 4i+1:

Math 59

*q=−A*_{e}^{j(θ}^{11}^{(4i+1)−θ}^{21}^{(4i+1))}*,−A*_{e}^{j(θ}^{11}^{(4i+1)−θ}^{21}^{(4i+1)−δ)} Equation 59

For symbol number 4i+2:

Math 60

*q=−A*_{e}^{j(θ}^{11}^{(4i+2)−θ}^{21}^{(4i+2))}*,−A*_{e}^{j(θ}^{11}^{(4i+2)−θ}^{21}^{(4i+2)−δ)} Equation 60

For symbol number 4i+3:

Math 61

*q=−A*_{e}^{j(θ}^{11}^{(4i+3)−θ}^{21}^{(4i+3))}*,−A*_{e}^{j(θ}^{11}^{(4i+3)−θ}^{21}^{(4i+3)−δ)} Equation 61

In this case, if q has the same solution in symbol numbers 4i, 4i+1, 4i+2, and 4i+3, then the channel elements of the direct waves do not greatly fluctuate. Therefore, a reception device having channel elements in which the value of q is equivalent to the same solution can no longer obtain excellent reception quality for any of the symbol numbers. Therefore, it is difficult to achieve the ability to correct errors, even if error correction codes are introduced. Accordingly, for q not to have the same solution, the following condition is necessary from Equations 58-61 when focusing on one of two solutions of q which does not include δ.

Math 62

*e*^{j(θ}^{11}^{(4i+x)−θ}^{21}^{(4i+x))}*≠e*^{j(θ}^{11}^{(4i+y)−θ}^{21}^{(4i+y)) }for *∀x, ∀y *(*x≠y; x, y=*0, 1, 2*, . . . , N−*2*, N−*1) Condition #1

(x is 0, 1, 2, 3; y is 0, 1, 2, 3; and x≠y.)

In an example fulfilling Condition #1, values are set as follows:

(1) θ_{11}(4i)=θ_{11}(4i+1)=θ_{11}(4i+2)=θ_{11}(4i+3)=0 radians,

(2) θ_{21}(4i)=0 radians,

(3) θ_{21}(4i+1)=π/2 radians,

(4) θ_{21}(4i+2)=π radians, and

(5) θ_{21}(4i+3)=3π/2 radians.

(The above is an example. It suffices for one each of zero radians, π/2 radians, π radians, and 3π/2 radians to exist for the set (θ_{21}(4i), θ_{21}(4i+1), θ_{21}(4i+2), θ_{21}(4i+3)).) In this case, in particular under condition (1), there is no need to perform signal processing (rotation processing) on the baseband signal S1(*t*), which therefore offers the advantage of a reduction in circuit size. Another example is to set values as follows.

(6) θ_{11}(4i)=0 radians,

(7) θ_{11}(4i+1)=π/2 radians,

(8) θ_{11}(4i+2)=π radians,

(9) θ_{11}(4i+3)=3π/2 radians, and

(10) θ_{21}(4i)=θ_{21}(4i+1)=θ_{21}(4i+2)=θ_{21}(4i+3)=0 radians.

(The above is an example. It suffices for one each of zero radians, π/2 radians, π radians, and 3π/2 radians to exist for the set (θ_{11}(4i), θ_{11}(4i+1), θ_{11}(4i+2), θ_{11}(4i+3)).) In this case, in particular under condition (6), there is no need to perform signal processing (rotation processing) on the baseband signal S2(*t*), which therefore offers the advantage of a reduction in circuit size. Yet another example is as follows.

(11) θ_{11}(4i)=θ_{11}(4i+1)=θ_{11}(4i+2)=θ_{11}(4i+3)=0 radians,

(12) θ_{21}(4i)=0 radians,

(13) θ_{21}(4i+1)=π/4 radians,

(14) θ_{21}(4i+2)=π/2 radians, and

(15) θ_{21}(4i+3)=3π/4 radians.

(The above is an example. It suffices for one each of zero radians, π/4 radians, π/2 radians, and 3π/4 radians to exist for the set (θ_{21}(4i), θ_{21}(4i+1), θ_{21}(4i+2), θ_{21}(4i+3)).)

(16) θ_{11}(4i)=0 radians,

(17) θ_{11}(4i+1)=π/4 radians,

(18) θ_{11}(4i+2)=π/2 radians,

(19) θ_{11}(4i+3)=3π/4 radians, and

(20) θ_{21}(4i)=θ_{21}(4i+1)=θ_{21}(4i+2)=θ_{21}(4i+3)=0 radians.

(The above is an example. It suffices for one each of zero radians, π/4 radians, π/2 radians, and 3π/4 radians to exist for the set (θ_{11}(4i), θ_{11}(4i+1), θ_{11}(4i+2), θ_{11}(4i+3)).)

While four examples have been shown, the scheme of satisfying Condition #1 is not limited to these examples.

Next, design requirements for not only θ_{11 }and θ_{12}, but also for λ and δ are described. It suffices to set λ to a certain value; it is then necessary to establish requirements for δ. The following describes the design scheme for δ when λ is set to zero radians.

In this case, by defining δ so that π/2 radians ≦|δ|≦π radians, excellent reception quality is achieved, particularly in an LOS environment.

Incidentally, for each of the symbol numbers 4i, 4i+1, 4i+2, and 4i+3, two points q exist where reception quality becomes poor. Therefore, a total of 2×4=8 such points exist. In an LOS environment, in order to prevent reception quality from degrading in a specific reception terminal, these eight points should each have a different solution. In this case, in addition to Condition #1, Condition #2 is necessary.

Math 63

*e*^{j(θ}^{11}^{(4i+x)−θ}^{21}^{(4i+x))}*≠e*^{j(θ}^{11}^{(4i+y)−θ}^{21}^{(4i+y)−δ) }for *∀x, ∀y *(*x, y=*0, 1, 2, 3)

and

*e*^{j(θ}^{11}^{(4i+x)−θ}^{21}^{(4i+x)−δ)}*≠e*^{j(θ}^{11}^{(4i+y)−θ}^{21}^{(4i+y)−δ) }for *∀x, ∀y *(*x≠y; x, y=*0, 1, 2, 3) Condition #2

Additionally, the phase of these eight points should be evenly distributed (since the phase of a direct wave is considered to have a high probability of even distribution). The following describes the design scheme for δ to satisfy this requirement.

In the case of example #1 and example #2, the phase becomes even at the points at which reception quality is poor by setting δ to ±3π/4 radians. For example, letting δ be 3π/4 radians in example #1 (and letting A be a positive real number), then each of the four slots, points at which reception quality becomes poor exist once, as shown in FIG. 20. In the case of example #3 and example #4, the phase becomes even at the points at which reception quality is poor by setting δ to ±π radians. For example, letting δ be π radians in example #3, then in each of the four slots, points at which reception quality becomes poor exist once, as shown in FIG. 21. (If the element q in the channel matrix H exists at the points shown in FIGS. 20 and 21, reception quality degrades.)

With the above structure, excellent reception quality is achieved in an LOS environment. Above, an example of changing precoding weights in a four-slot period (cycle) is described, but below, changing precoding weights in an N-slot period (cycle) is described. Making the same considerations as in Embodiment 1 and in the above description, processing represented as below is performed on each symbol number.

For symbol number Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number Ni+1:

When generalized, this equation is as follows.

For symbol number Ni+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number Ni+N−1:

Accordingly, r**1** and r**2** are represented as follows.

For symbol number Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number Ni+1:

When generalized, this equation is as follows.

For symbol number Ni+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number Ni+N−1:

In this case, it is assumed that only components of direct waves exist in the channel elements h_{11}(t), h_{12}(t), h_{21}(t), and h_{22}(t), that the amplitude components of the direct waves are all equal, and that fluctuations do not occur over time. With these assumptions, Equations 66-69 can be represented as follows.

For symbol number Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number Ni+1:

When generalized, this equation is as follows.

For symbol number Ni+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number Ni+N−1:

In Equations 70-73, let A be a real number and q be a complex number. The values of A and q are determined in accordance with the positional relationship between the transmission device and the reception device. Equations 70-73 can be represented as follows.

For symbol number Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number Ni+1:

When generalized, this equation is as follows.

For symbol number Ni+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number Ni+N−1:

As a result, when q is represented as follows, a signal component based on one of s**1** and s**2** is no longer included in r**1** and r**2**, and therefore one of the signals s**1** and s**2** can no longer be obtained.

For symbol number Ni (where i is an integer greater than or equal to zero):

Math 80

*q=−A*_{e}^{j(θ}^{11}^{(Ni)−θ}^{21}^{(Ni))}*,−A*_{e}^{j(θ}^{11}^{(Ni)−θ}^{21}^{(Ni)−δ)} Equation 78

For symbol number Ni+1:

Math 81

*q=−A*_{e}^{j(θ}^{11}^{(Ni+1)−θ}^{21}^{(Ni+1))}*,−A*_{e}^{j(θ}^{11}^{(Ni+1)−θ}^{21}^{(Ni+1)−δ)} Equation 79

When generalized, this equation is as follows.

For symbol number Ni+k (k=0, 1, . . . , N−1):

Math 82

*q=−A*_{e}^{j(θ}^{11}^{(Ni+k)−θ}^{21}^{(Ni+k))}*,−A*_{e}^{j(θ}^{11}^{(Ni+k)−θ}^{21}^{(Ni+k)−δ)} Equation 80

Furthermore, for symbol number Ni+N−1:

Math 83

*q=−A*_{e}^{j(θ}^{11}^{(Ni+N−1)−θ}^{21}^{(Ni+N−1))}*,−A*_{e}^{j(θ}^{11}^{(Ni+N−1)−θ}^{21}^{(Ni+N−1)−δ)} Equation 81

In this case, if q has the same solution in symbol numbers Ni through Ni+N−1, then since the channel elements of the direct waves do not greatly fluctuate, a reception device having channel elements in which the value of q is equivalent to this same solution can no longer obtain excellent reception quality for any of the symbol numbers. Therefore, it is difficult to achieve the ability to correct errors, even if error correction codes are introduced. Accordingly, for q not to have the same solution, the following condition is necessary from Equations 78-81 when focusing on one of two solutions of q which does not include 6.

Math 84

*e*^{j(θ}^{11}^{(Ni+x)−θ}^{21}^{(Ni+x))}*≠e*^{j(θ}^{11}^{(Ni+y)−θ}^{21}^{(Ni+y)) }for ∀*x, ∀y *(*x≠y; x, y=*0, 1, 2*, . . . , N−*2*, N−*1) Condition #3

x is 0, 1, 2, . . . , N−2, N−1; y is 0, 1, 2, . . . , N−2, N−1; and x≠y.)

Next, design requirements for not only θ_{11 }and θ_{12}, but also for λ and δ are described. It suffices to set λ to a certain value; it is then necessary to establish requirements for δ. The following describes the design scheme for δ when λ is set to zero radians.

In this case, similar to the scheme of changing the precoding weights in a four-slot period (cycle), by defining δ so that π/2 radians ≦|δ|≦π radians, excellent reception quality is achieved, particularly in an LOS environment.

In each symbol number Ni through Ni+N−1, two points labeled q exist where reception quality becomes poor, and therefore 2N such points exist. In an LOS environment, in order to achieve excellent characteristics, these 2N points should each have a different solution. In this case, in addition to Condition #3, Condition #4 is necessary.

Math 85

*e*^{j(θ}^{11}^{(Ni+x)−θ}^{21}^{(Ni+x))}*≠e*^{j(θ}^{11}^{(Ni+y)−θ}^{21}^{(Ni+y)−δ) }for *∀x, ∀y *(*x≠y; x, y=*0, 1, 2*, . . . , N−*2*, N−*1)

and

*e*^{j(θ}^{11}^{(Ni+x)−θ}^{21}^{(Ni+x)−δ)}*≠e*^{j(θ}^{11}^{(Ni+y)−θ}^{21}^{(Ni+y)−δ) }for *∀x, ∀y *(*x≠y; x, y=*0, 1, 2*, . . . , N−*2*, N−*1) Condition #4

Additionally, the phase of these 2N points should be evenly distributed (since the phase of a direct wave at each reception device is considered to have a high probability of even distribution).

As described above, when a transmission device transmits a plurality of modulated signals from a plurality of antennas in a MIMO system, the advantageous effect of improved transmission quality, as compared to conventional spatial multiplexing MIMO system, is achieved in an LOS environment in which direct waves dominate by hopping between precoding weights regularly over time.

In the present embodiment, the structure of the reception device is as described in Embodiment 1, and in particular with regards to the structure of the reception device, operations have been described for a limited number of antennas, but the present invention may be embodied in the same way even if the number of antennas increases. In other words, the number of antennas in the reception device does not affect the operations or advantageous effects of the present embodiment. Furthermore, in the present embodiment, similar to Embodiment 1, the error correction codes are not limited.

In the present embodiment, in contrast with Embodiment 1, the scheme of changing the precoding weights in the time domain has been described. As described in Embodiment 1, however, the present invention may be similarly embodied by changing the precoding weights by using a multi-carrier transmission scheme and arranging symbols in the frequency domain and the frequency-time domain. Furthermore, in the present embodiment, symbols other than data symbols, such as pilot symbols (preamble, unique word, and the like), symbols for control information, and the like, may be arranged in the frame in any way.

In Embodiment 1 and Embodiment 2, the scheme of regularly hopping between precoding weights has been described for the case where the amplitude of each element in the precoding weight matrix is equivalent. In the present embodiment, however, an example that does not satisfy this condition is described.

For the sake of contrast with Embodiment 2, the case of changing precoding weights over an N-slot period (cycle) is described. Making the same considerations as in Embodiment 1 and Embodiment 2, processing represented as below is performed on each symbol number. Let β be a positive real number, and β≠1,

For symbol number Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number Ni+1:

When generalized, this equation is as follows.

For symbol number Ni+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number Ni+N−1:

Accordingly, r**1** and r**2** are represented as follows.

For symbol number Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number Ni+1:

When generalized, this equation is as follows.

For symbol number Ni+k (k=0, 1, . . . , N−1):

When generalized, this equation is as follows.

For symbol number Ni+N−1:

In this case, it is assumed that only components of direct waves exist in the channel elements h_{11}(t), h_{12}(t), h_{21}(t), and h_{22}(t), that the amplitude components of the direct waves are all equal, and that fluctuations do not occur over time. With these assumptions, Equations 86-89 can be represented as follows.

For symbol number Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number Ni+1:

When generalized, this equation is as follows.

For symbol number Ni+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number Ni+N−1:

In Equations 90-93, let A be a real number and q be a complex number. Equations 90-93 can be represented as follows.

For symbol number Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number Ni+1:

When generalized, this equation is as follows.

For symbol number Ni+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number Ni+N−1:

As a result, when q is represented as follows, one of the signals s**1** and s**2** can no longer be obtained.

For symbol number Ni (where i is an integer greater than or equal to zero):

For symbol number Ni+1:

When generalized, this equation is as follows.

For symbol number Ni+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number Ni+N−1:

In this case, if q has the same solution in symbol numbers Ni through Ni+N−1, then since the channel elements of the direct waves do not greatly fluctuate, excellent reception quality can no longer be obtained for any of the symbol numbers. Therefore, it is difficult to achieve the ability to correct errors, even if error correction codes are introduced. Accordingly, for q not to have the same solution, the following condition is necessary from Equations 98-101 when focusing on one of two solutions of q which does not include δ.

Math 106

*e*^{j(θ}^{11}^{(Ni+x)−θ}^{21}^{(Ni+x))}*≠e*^{j(θ}^{11}^{(Ni+y)−θ}^{21}^{(Ni+y)) }for ∀*x, ∀y *(*x≠y; x, y=*0, 1, 2*, . . . , N−*2*, N−*1) Condition #5

(x is 0, 1, 2, . . . , N−2, N−1; y is 0, 1, 2, . . . , N−2, N−1; and x≠y.)

Next, design requirements for not only θ_{11 }and θ_{12}, but also for λ and δ are described. It suffices to set λ to a certain value; it is then necessary to establish requirements for δ. The following describes the design scheme for δ when λ is set to zero radians.

In this case, similar to the scheme of changing the precoding weights in a four-slot period (cycle), by defining δ so that π/2 radians ≦|δ|≦π radians, excellent reception quality is achieved, particularly in an LOS environment.

In each of symbol numbers Ni through Ni+N−1, two points q exist where reception quality becomes poor, and therefore 2N such points exist. In an LOS environment, in order to achieve excellent characteristics, these 2N points should each have a different solution. In this case, in addition to Condition #5, considering that β is a positive real number, and β≠1, Condition #6 is necessary.

Math 107

*e*^{j(θ}^{11}^{(Ni+x)−θ}^{21}^{(Ni+x)−δ)}*≠e*^{j(θ}^{11}^{(Ni+y)−θ}^{21}^{(Ni+y)−δ) }for ∀*x, ∀y *(*x≠y; x, y=*0, 1, 2*, . . . , N−*2*, N−*1) Condition #6

As described above, when a transmission device transmits a plurality of modulated signals from a plurality of antennas in a MIMO system, the advantageous effect of improved transmission quality, as compared to conventional spatial multiplexing MIMO system, is achieved in an LOS environment in which direct waves dominate by hopping between precoding weights regularly over time.

In the present embodiment, the structure of the reception device is as described in Embodiment 1, and in particular with regards to the structure of the reception device, operations have been described for a limited number of antennas, but the present invention may be embodied in the same way even if the number of antennas increases. In other words, the number of antennas in the reception device does not affect the operations or advantageous effects of the present embodiment. Furthermore, in the present embodiment, similar to Embodiment 1, the error correction codes are not limited.

In the present embodiment, in contrast with Embodiment 1, the scheme of changing the precoding weights in the time domain has been described. As described in Embodiment 1, however, the present invention may be similarly embodied by changing the precoding weights by using a multi-carrier transmission scheme and arranging symbols in the frequency domain and the frequency-time domain. Furthermore, in the present embodiment, symbols other than data symbols, such as pilot symbols (preamble, unique word, and the like), symbols for control information, and the like, may be arranged in the frame in any way.

In Embodiment 3, the scheme of regularly hopping between precoding weights has been described for the example of two types of amplitudes for each element in the precoding weight matrix, 1 and 3.

In this case, the following is ignored.

Next, the example of changing the value of β by slot is described. For the sake of contrast with Embodiment 3, the case of changing precoding weights over a 2×N-slot period (cycle) is described.

Making the same considerations as in Embodiment 1, Embodiment 2, and Embodiment 3, processing represented as below is performed on symbol numbers. Let β be a positive real number, and β≠1. Furthermore, let α be a positive real number, and α≠β.

For symbol number 2Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number 2Ni+1:

When generalized, this equation is as follows.

For symbol number 2Ni+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number 2Ni+N−1:

For symbol number 2Ni+N (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number 2Ni+N+1:

When generalized, this equation is as follows.

For symbol number 2Ni+N+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number 2Ni+2N−1:

Accordingly, r**1** and r**2** are represented as follows.

For symbol number 2Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number 2Ni+1:

When generalized, this equation is as follows.

For symbol number 2Ni+k (k=0, 1, . . . , N−1):

Furthermore, for symbol number 2Ni+N−1:

For symbol number 2Ni+N (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit.

For symbol number 2Ni+N+1:

When generalized, this equation is as follows.

For symbol number 2Ni+N+k (k=0, 1, . . . , N−1):

When generalized, this equation is as follows.

For symbol number 2Ni+2N−1:

In this case, it is assumed that only components of direct waves exist in the channel elements h_{11}(t), h_{12}(t), h_{21}(t), and h_{22}(t), that the amplitude components of the direct waves are all equal, and that fluctuations do not occur over time. With these assumptions, Equations 110-117 can be represented as follows.

For symbol number 2Ni (where i is an integer greater than or equal to zero):

Here, j is an imaginary unit,

For symbol number 2Ni+1:

When generalized, this equation is as follows.

For symbol number 2Ni+k (k=0, 1, . . . , N−1):