JP6195490B2 - MIMO receiver - Google Patents

Description

  The present invention relates to a demodulation technique in a MIMO receiver using a multiple-input multiple-output (hereinafter referred to as “MIMO (Multiple Input Multiple Output)”) multiplex transmission scheme using a plurality of transmission antennas and a plurality of reception antennas.

  When transmitting program materials such as news videos and live video of events from a news gathering site to a broadcast studio or relay station, it is effective to use a video wireless transmission system that transmits video signals wirelessly. Typical devices used in this video wireless transmission system include an FPU (Field Pick-up Unit) device and a wireless camera.

  2. Description of the Related Art Conventionally, development of a new video wireless transmission system for realizing a wireless camera that wirelessly transmits a high-definition (registered trademark) television signal with low delay and high line reliability has attracted attention. In this new video wireless transmission system, it is considered to use a MIMO-OFDM transmission system that transmits a plurality of OFDM (Orthogonal Frequency Division Multiplexing) signals on the same frequency using a plurality of transmitting and receiving antennas. Yes.

  This MIMO-OFDM transmission scheme realizes space division multiplex transmission by transmitting a plurality of OFDM signals on the same frequency. As a result, the transmission speed can be increased to several times the number of transmission antennas (the number of reception antennas is required to be equal to or greater than the number of transmission antennas), and the transmission quality is improved and the required C / N is reduced due to diversity effect. be able to.

  In the development of the above-described wireless camera, high-capacity transmission is required to deliver high-quality captured images to viewers, but uninterrupted video transmission is particularly important due to the nature of broadcasting. For this reason, application of an MLD (Maximum Likelihood Detection) method having the best transmission characteristics has been studied in the demodulation technique of the MIMO multiplex transmission method capable of simultaneously transmitting a plurality of transmission signals on the same frequency.

  However, in the MLD scheme, the computation scale necessary for demodulation increases exponentially as the number of transmission signals to be multiplexed and the number of modulation multilevels increase. In a transmitter / receiver using such a MIMO multiplex transmission system, it has been difficult to implement a mechanism for transmitting information of a transmission signal number and a modulation multi-level number exceeding a certain level.

  Therefore, many MLD schemes that reduce the amount of computation and reduce the amount of computation that do not degrade the transmission characteristics have been studied. Among them, the QRM-MLD method is widely known as a method that can effectively reduce the amount of calculation. In the QRM-MLD system, QR decomposition is performed on a matrix of channel responses between transmitting and receiving antennas, and based on the obtained upper triangular matrix R, the modulation candidate points of each transmission signal are divided into stages corresponding to the number of transmitting antennas. It asks in order.

[MIMO-OFDM transmission system]
First, an outline of a MIMO-OFDM transmission system that performs demodulation using the MLD scheme or the QRM-MLD scheme will be described. FIG. 7 is a schematic diagram showing an overall configuration of a MIMO-OFDM transmission system. This MIMO-OFDM transmission system 300 includes one terminal device (hereinafter referred to as a transmission device (MIMO transmission device)) 100 having four transmission antennas # 1 to # 4 and four receptions. This is a wireless camera system that performs MIMO-OFDM transmission with a base station apparatus (hereinafter referred to as a receiving apparatus (MIMO receiving apparatus)) 200 provided with antennas # 1 to # 4. The format of the OFDM signal transmitted from the transmission apparatus 100 to the reception apparatus 200 conforms to the provisions of ARIB STD-B43. Transmitting antennas # 1 to # 4 and the reception antenna # 1 to # MIMO channel between the 4 (channel responses _{h 11, h 21, h 31} , h 41, h 12, ···, h 44) are Is formed.

  The transmission device 100 is a terminal device that can move freely, and transmits different OFDM signals at the same frequency from the four transmission antennas # 1 to # 4. Note that the transmission apparatus 100 arranges pilot signals at predetermined intervals on the frequency axis and continuously arranges them on the time axis so that the reception unit of the reception apparatus 200 can estimate the propagation path response.

[MIMO transmitter / terminal side]
The terminal-side transmission apparatus 100 includes four transmission antennas # 1 to # 4, an S / P (serial / parallel) conversion unit 101, a MIMO-OFDM modulation unit 102, and the like. When the transmission apparatus 100 inputs a TS signal obtained by encoding a captured video, the transmission apparatus 100 adds an error correction code or the like to the TS signal. The S / P converter 101 receives a TS signal to which an error correction code or the like has been added, and converts the input serial TS signal into four parallel TS signals. The MIMO-OFDM modulation unit 102 receives the four systems of TS signals converted by the S / P conversion unit 101 and applies predetermined MIMO-OFDM modulation to the four systems of TS signals. The four OFDM signals that are MIMO-OFDM modulated by the MIMO-OFDM modulation section 102 are transmitted from the transmission antennas # 1 to # 4. In this case, an orthogonal code or the like is assigned to the pilot signal of each OFDM signal. Thereby, the receiving apparatus 200 on the base station side can separate the pilot signal for each MIMO propagation path between the respective transmission / reception antennas and estimate each propagation path response (see Patent Document 1).

[MIMO receiver / base station side]
The receiving apparatus 200 on the base station side includes four receiving antennas # 1 to # 4, a MIMO-OFDM demodulator 201, a P / S (parallel / serial) converter 202, and the like. The receiving apparatus 200 receives the four systems of OFDM signals transmitted from the transmitting apparatus 100 using the four receiving antennas # 1 to # 4. MIMO-OFDM demodulation section 201 estimates a channel response between each transmitting / receiving antenna based on a pilot signal included in the received OFDM signal. Also, the MIMO-OFDM demodulation unit 201 performs MIMO-OFDM demodulation on the received OFDM signal using the estimated propagation path response, and restores the original four systems of TS signals. The P / S conversion unit 202 inputs the four systems of TS signals demodulated by the MIMO-OFDM by the MIMO-OFDM demodulation unit 201, and converts the input four systems of parallel TS signals into serial TS signals. Then, error correction code decoding and decoding are performed on the TS signal converted by the P / S conversion unit 202 to restore the original captured video.

ここで、Ａ^Ｔは任意の行列Ａに対する転置行列を示す。Ｙ^Ｔ＝［ｙ₁，ｙ₂，ｙ₃，ｙ₄］は受信信号ベクトル、Ｘ^Ｔ＝［ｘ₁，ｘ₂，ｘ₃，ｘ₄］は送信信号ベクトル、Ｈは伝搬路応答の行列、Ｎ^Ｔ＝［ｎ₁，ｎ₂，ｎ₃，ｎ₄］は雑音ベクトルを示す。ｙ₁は受信アンテナ＃１にて受信したＯＦＤＭ信号、ｙ_2〜4はそれぞれ受信アンテナ＃２〜＃４にて受信したＯＦＤＭ信号である。ｘ₁は送信アンテナ＃１から送信されたＯＦＤＭ信号、ｘ_2〜4はそれぞれ送信アンテナ＃２〜＃４から送信されたＯＦＤＭ信号である。また、例えばｈ₁₁は送信アンテナ＃１と受信アンテナ＃１との間の伝搬路応答、ｈ₂₁は送信アンテナ＃１と受信アンテナ＃２との間の伝搬路応答を示す。 Here, the relationship between the OFDM signal received by the receiving apparatus 200, the estimated propagation path response, and the OFDM signal transmitted by the transmitting apparatus 100 is expressed by Expression (1).

Here, ^AT represents a transposed matrix for an arbitrary matrix A. Y ^T = [y ₁ , y ₂ , y ₃ , y ₄ ] is a received signal vector, X ^T = [x ₁ , x ₂ , x ₃ , x ₄ ] is a transmitted signal vector, H is a channel response matrix, N ^T = [n ₁ , n ₂ , n ₃ , n ₄ ] represents a noise vector. y ₁ is an OFDM signal received by the receiving antenna # 1, y _{2 to 4} are OFDM signal received by the receiving antenna # 2 to # 4, respectively. x ₁ is an OFDM signal transmitted from the transmission antenna # 1, and x _{2 to 4} are OFDM signals transmitted from the transmission antennas # 2 to # 4, respectively. For example, h ₁₁ represents a propagation path response between the transmission antenna # 1 and the reception antenna # 1, and h ₂₁ represents a propagation path response between the transmission antenna # 1 and the reception antenna # 2.

  In the equation (1), since the reception signal and the propagation path response can be detected by the reception apparatus 200, the unknown values are the transmission signal vector X and the noise vector N. The noise vector N causes a transmission error. When the power of the received signal is sufficiently larger than the noise power, a desired signal can be transmitted without a transmission error. Therefore, if the noise vector N is ignored, an expression in which the unknown value is only the transmission signal vector X is obtained.

  As described above, the MLD scheme and the QRM-MLD scheme are known as the MIMO-OFDM demodulation processing scheme by the receiving apparatus 200. The QRM-MLD method is a reduction in the amount of calculation of the MLD method. Hereinafter, each of the MLD method and the QRM-MLD method will be described.

[MLD method]
First, the MLD method will be described. The MLD method is known as a method having the most excellent transmission characteristics. In the MLD scheme, a replica signal (received replica signal) of the received signal vector Y is generated for modulation candidate points of all patterns that can be taken by each element (x ₁ , x ₂ , x ₃ , x ₄ ) of the transmitted signal vector X. . The received replica signal is expressed by Expression (2) using the right side of Expression (1).

Of the received replica signals generated from the combinations of all transmitted signal vectors X, the received replica signal is generated for the received replica signal closest to the received signal vector Y ^T = [y ₁ , y ₂ , y ₃ , y ₄ ]. The modulation candidate point of the transmission signal vector X to be selected as the original transmission signal is the demodulation result of the MLD hard decision. At this time, as the metric for determining the difference between the received replica signal and the received signal vector Y ^T, the square metric of formula (3) indicating the theoretically best transmission characteristics are used.

  In the square metric of Equation (3), the difference between the received signal and the received replica signal is obtained, and the sum of the square of the real part and the square of the imaginary part is obtained.

  In the MLD scheme, a transmission signal vector X that generates a reception replica signal that takes the smallest value with respect to the square metric E obtained by the equation (3) is selected as a demodulation result. However, since the combinations of transmission signal vectors X that may be selected as demodulation results are combinations in the factorial of the number of multiplexed signals with respect to the number of modulation multilevels, the number of transmission signals that are the number of modulation multilevels or the number of multiplexed signals Increases, the number of received replica signals obtained by the equation (2) to be calculated and the number of square metric operations obtained by the equation (3) increase, and the amount of computation increases. This makes it difficult to implement the device.

[QRM-MLD method]
Therefore, in order to reduce the amount of calculation of the MLD method, the QRM-MLD method is known (see Patent Document 2). FIG. 8 is a block diagram showing a configuration of a conventional receiving apparatus. The receiving device 200 is a device that performs demodulation processing using a conventional QRM-MLD method, and includes four receiving antennas # 1 to # 4 (not shown), an FFT processing unit 21, a propagation path estimation unit 22, and a rearrangement QR. A decomposing unit 23, a received signal converting unit 24, and a likelihood information / demodulation result generating unit 25 are provided. Note that only the components directly related to the present invention are shown in the receiving apparatus 200 of FIG. 8, and other components such as an error correction unit are omitted. Hereinafter, the QRM-MLD method will be described with reference to FIG.

Four receiving antennas # 1 to # 4 (not shown) provided in the receiving apparatus 200 are MIMO propagation paths (propagation path responses h ₁₁ ,. The OFDM signal is received via h ₂₁ , h ₃₁ , h ₄₁ , h ₁₂ ,..., h ₄₄ ). That is, OFDM signals transmitted from the four transmission antennas # 1 to # 4 and interfering via the MIMO propagation path are received by the respective reception antennas # 1 to # 4.

  The FFT processing unit 21 inputs OFDM signals obtained by orthogonally demodulating the signals received by the receiving antennas # 1 to # 4 and detecting the symbol timing, removes the GI signal from the OFDM signal, performs FFT, and performs a time axis. Convert data to frequency axis data. Then, the FFT processing unit 21 outputs the pilot signal in the frequency axis data to the propagation path estimation unit 22 and outputs the data signal to the reception signal conversion unit 24.

The propagation path estimation unit 22 receives the pilot signal from the FFT processing unit 21 and uses the pilot signal to perform propagation path responses h ₁₁ and h ₂₁ between the transmission antennas # 1 to # 4 and the reception antennas # 1 to # 4. , H ₃₁ , h ₄₁ , h ₁₂ ,..., H ₄₄ , a propagation path response H (a propagation path response matrix H) is generated and output to the rearrangement QR decomposition unit 23.

The rearrangement QR decomposition unit 23 receives the channel response H from the channel estimation unit 22 and a plurality of column vectors h _i (= [h _1i , h _2i , h _3i , h _4i ]) that configure the channel response H. ^T ) (i = 1, 2, 3, 4), the norm is calculated, and the column vectors are rearranged from left to right in ascending order of the norm to generate the propagation path response H ′. Norm of column vectors h _{i ^{is, | h i | 2 = |}} h 1i | 2 + | h 2i | 2 + | h 3i | 2 + | formula of ² | h _4i. In this way, the rearrangement QR decomposition unit 23 calculates a norm for each of i = 1, 2, 3, and 4 using the above equation, and rearranges them from the left in ascending order of the norm, thereby creating a new propagation path. A matrix H ′ is generated.

  Thereby, in the channel response H ′, the column vectors are arranged in ascending order of the norm from the left column to the right column, the column vector having the smallest norm is arranged in the leftmost column, and the rightmost column is arranged in the rightmost column. Is the column vector with the largest norm.

The rearrangement QR decomposition unit 23 performs QR decomposition on the generated propagation path response H ′ (H ′ = QR) to obtain a unitary matrix Q and an upper triangular matrix R. Then, rearrangement QR decomposition section 23 outputs unitary matrix Q to reception signal conversion section 24 and outputs upper triangular matrix R to likelihood information / demodulation result generation section 25. In FIG. 8, R (4) is a non-zero element r ₄₄ in the fourth row (bottom row) in the upper triangular matrix R shown in equation (7) described later, and R (k) is the upper triangular matrix R. Are all elements in the k-th row (k = 1, 2, 3).

As a result, in the upper triangular matrix R, as in the channel response H ′, column vectors are arranged in ascending order of norms from the left column to the right column, and the column vector having the smallest norm in the leftmost column. Are arranged, and the column vector having the maximum norm is arranged in the rightmost column. The plurality of column vectors constituting the upper triangular matrix R are as shown in Equation (7) described later when the number of columns is 4 and the number of rows is 4, and the rightmost column vector h ₄ having the maximum norm is an element. It consists of r ₁₄ ~r _44. In a likelihood information / demodulation result generation unit 25 to be described later, processing for each row from the lowest row (fourth row) to the first row of the upper triangular matrix R is performed.

In this case, for example, in the process of the lowermost row (fourth row) of the upper triangular matrix R, the elements r ₁₄ to r ₄₄ constituting the rightmost column vector h ₄ having the maximum norm with respect to the column vector of the propagation path matrix H ′. Part r ₄₄ is used. This element r ₄₄ has a size equal to or larger than the element r ₄₄ obtained from the channel matrix H that is not rearranged. For this reason, the energy of the element r ₄₄ of the channel matrix H ′ that has been rearranged becomes larger than that obtained from the channel matrix H that has not been rearranged. signal detection accuracy of the transmitted signal to calculate improves with r _44. Note that the method of rearrangement QR decomposition is known, and refer to Non-Patent Document 1 for details.

The received signal conversion unit 24 receives the data signal from the FFT processing unit 21 and also receives the unitary matrix Q from the rearrangement QR decomposition unit 23, converts the received signal Y, which is a data signal, by the unitary matrix Q, The subsequent reception signal Y ′ is output to the likelihood information / demodulation result generation unit 25. In FIG. 8, y ′ ₄ is an element of the fourth row (lowermost row) in the received signal Y ′ shown in Expression (7) described later, and y ′ _k is the kth row (k = 3,2,1).

ｎは雑音成分である。以下、ｙを受信信号ベクトルとし、ｘを送信信号ベクトルとする。 When general QR decomposition is performed on the propagation path response H of the equation (1), the propagation path response H can be decomposed into a unitary matrix Q and an upper triangular matrix R as represented by the expression (4). .

n is a noise component. Hereinafter, y is a received signal vector and x is a transmitted signal vector.

Further, since the unitary matrix Q satisfies Q ^H Q = I, Expression (5) is obtained. Matrix I represents a unit matrix.

Here, when Q ^H y = y ′ and Q ^H n = n ′, Expression (6) is obtained.

If the said Formula (6) is described for every element, it will become Formula (7).

Here, when the noise component N ′ is ignored, the equation (7) becomes the equation (8).

  The likelihood information / demodulation result generation unit 25 receives the reception signal Y ′ from the reception signal conversion unit 24 and receives the upper triangular matrix R from the rearrangement QR decomposition unit 23, and receives the reception signal Y ′ and the upper triangular matrix R. All possible modulation candidate points are substituted for each stage of each row from the bottom row (fourth row) to the first row (for each stage of each element from the bottom row to the first row in the received signal Y ′). The square metric is obtained, a modulation candidate point is selected from the square metric, the demodulation result by the hard decision is obtained, and the likelihood information by the soft decision is obtained.

  The likelihood information / demodulation result generation unit 25 includes a modulation point recording memory 26, a metric calculation unit 27, a candidate point selection unit 28, an interference component calculation unit 29, a subtraction unit 30, a metric calculation unit 31, and a candidate point selection unit 32. ing.

In the QRM-MLD method, the lowest row is the first stage, and operations are performed in each stage in order from the bottom. First, the metric calculation unit 27 obtains the square metric of Expression (9) for the bottom row in the first stage.

At this time, all possible modulation candidate points are substituted for x _{4 in the} equation (9) to obtain the square metric E ₁ . Then, the candidate point selection unit 28, the square metric E ₁ is the smallest modulation candidate points by selecting the (x ₄ modulation candidate points), or squared metric E ₁ is smaller sequentially from the M ₁ single modulation candidate point (M ₁ x ₄ modulation candidate point) is selected.

Next, in the second stage, the metric calculation unit 31 obtains the square metric E ₂ of Expression (10) for the row one row above the lowest row.

Here, E ₁ (x ₄ ) is a square metric corresponding to the value of x ₄ obtained by the equation (9). At this time, the modulation candidate point selected in the first stage is substituted for x _{4 in the} equation (10), and all possible modulation candidate points are substituted for x ₃ to obtain the square metric E ₂ . . Then, the candidate point selector 32, the square metric E ₂ is the smallest modulation candidate point (x _3, x ₄ modulation candidate point) to either select or square metric E ₂ is smaller sequentially from the M ₂ pieces of modulation candidates A combination of points (M ₂ (x ₃ , x ₄ ) modulation candidate points) is selected.

The modulation point recording memory 26 records all the modulation candidate points of the transmission signals x _{4 to} x ₁ , and all the modulation candidate points of the transmission signal x ₄ are stored in the metric calculation unit 27 and the candidate point selection unit 28. And outputs all the modulation candidate points of the transmission signals x _{3 to} x ₁ to the metric calculation unit 31 and the candidate point selection unit 32. The interference component calculation unit 29 inputs elements of the k-th row (k = 3, 2, 1) in the upper triangular matrix R from the rearrangement QR decomposition unit 23 corresponding to each stage, and the candidate point selection unit 28 inputting modulated candidate points C _M of the transmission signal x _4, which is selected, and the transmission signal x _3, x _2, x ₁ modulation candidate points C _M selected from the candidate point selection unit 32 from. Then, the interference component calculation unit 29 calculates the interference component and outputs it to the subtraction unit 30. The interference component, when the formula of k = 3 (3 row) in (8), a r ₃₄ x _4, if k = 2 in the (second row), be _{r 23 x 3 + r 24 x} 4 , K = 1 (first line), r ₁₂ x ₂ + r ₁₃ x ₃ + r ₁₄ x ₄ . The subtraction unit 30 subtracts the interference component from the reception signal y ′ _k from the reception signal conversion unit 24 corresponding to each stage, and outputs the subtraction result to the metric calculation unit 31.

Next, in the third stage, the metric calculation unit 31 obtains the square metric E ₃ of Equation (11) for the row two rows above the lowest row.

Here, E ₂ (x ₃ , x ₄ ) is a square metric corresponding to the values of x ₃ and x ₄ obtained by the equation (10). At this time, M ₂ modulation candidate points selected in the second stage are substituted for x ₃ and x _{4 in} the equation (11), and all possible modulation candidate points are substituted for x _2. Find the square metric E ₃ . Then, the candidate point selector 32, the square metric E ₃ is the smallest modulation candidate points _{(x 2, x 3, x} 4 modulation candidate point) to either select or _three M from ascending order square metric E ₃ Modulation candidate point combinations (M ₃ (x ₂ , x ₃ , x ₄ ) modulation candidate points) are selected.

Next, in the fourth stage, the metric calculation unit 31 obtains the square metric E ₄ of Expression (12) for the top row.

Here, E ₃ (x ₂ , x ₃ , x ₄ ) is a square metric corresponding to the values of x ₂ , x _3, and x ₄ obtained by the equation (11). At this time, M ₃ modulation candidate points selected in the third stage are substituted for x ₂ , x ₃ and x _{4 in the} equation (12), and all possible modulation candidate points for x ₁ are obtained. And square metric E ₄ is obtained. Then, the candidate point selection unit 32 selects a modulation candidate point (modulation candidate point of x ₁ , x ₂ , x ₃ , x ₄ ) having the smallest square metric E ₄ .

In the hard decision, the candidate point selection unit 32 outputs the modulation candidate points (x ₁ , x ₂ , x ₃ , x ₄ modulation candidate points) selected in the fourth stage as the hard decision demodulation result. On the other hand, in the soft decision, the metric calculation unit 31 outputs E ₄ obtained by adding the square metrics calculated in each stage as likelihood information.

  As described above, by using the QRM-MLD scheme, the number of reception replica signals to be generated and the number of square metric calculations can be significantly reduced as compared with the MLD scheme. Since the QRM-MLD system has little deterioration in transmission characteristics, its application is being studied in a wide range of fields as a computational complexity reduction type MLD system.

  In this QRM-MLD scheme, even when the number of modulation levels of transmission signals or the number of transmission signals to be multiplexed increases, the processing load can be reduced as compared with the MLD scheme as described above. However, the number of reception replica signals and the number of square metric calculations increase as the number of modulation levels of transmission signals or the number of transmission signals to be multiplexed increases.

In order to solve this problem, the number of multiplications can be reduced by using the Manhattan metric shown in Expression (13) instead of the square metric.

  In the Manhattan metric of the equation (13), the difference between the received signal and the received replica signal is obtained, and the sum of the absolute value of the real part and the absolute value of the imaginary part is obtained. Since the Manhattan metric does not require multiplication, the amount of calculation can be reduced compared to the square metric shown in the equation (3). In particular, in an arithmetic integrated circuit using an FPGA (Field Programmable Gate Array) or the like, the multiplication circuit consumes a lot of resources compared to the addition and subtraction circuit, and therefore a metric that does not use a multiplication circuit such as a Manhattan metric is It becomes a very effective metric calculation method. However, since the Manhattan metric is not a theoretically optimal metric, transmission characteristics deteriorate.

JP 2005-124125 A Japanese Patent No. 4290660

D. Wubben, J. Rinas, R. Bohnke, V. Kuhn, and KDKammeyer, “Efficient algorithm for detecting layered Space-time code”, Proc. 4th International ITG Conference on Source and Channel Coding, pp.399-405 ( Jan. 2002)

  In general, when MIMO demodulation is performed on a propagation path response having a large correlation between the MIMO propagation paths between the transmitting and receiving antennas, transmission characteristics are significantly deteriorated compared to when MIMO demodulation is performed on a propagation path response having a low correlation. Further, when performing general QR decomposition on a channel response having a large correlation of the MIMO channel, the QR decomposition element approaches 0, so that the modulation candidate point narrowing accuracy (selection accuracy) is greatly degraded. . For this reason, by performing the above-described rearrangement QR decomposition, it is possible to suppress deterioration in the accuracy of narrowing down modulation candidate points.

  However, there has been a problem that the above-described rearrangement QR decomposition cannot be directly applied to block QR decomposition described later. In the block QR decomposition, the channel response H is subjected to QR decomposition in units of blocks, and MIMO demodulation processing is performed in units of blocks. For this reason, when the correlation of a plurality of elements in a block is large, as described above, the accuracy of narrowing down modulation candidate points deteriorates.

  Therefore, the present invention has been made to solve the above-described problems, and the object of the present invention is to perform transmission characteristics even when the correlation of the MIMO channel is large when performing MIMO demodulation using block QR decomposition. An object of the present invention is to provide a MIMO receiving apparatus capable of suppressing degradation of noise.

In order to achieve the above object, a MIMO receiving apparatus according to the present invention receives signals transmitted from a plurality of transmitting antennas via a plurality of receiving antennas, and between the plurality of transmitting antennas and the plurality of receiving antennas. In the MIMO receiver that estimates the channel response of the channel, QR-decomposes the matrix of the channel response, and performs MIMO demodulation, the column vectors constituting the channel response matrix are rearranged, and the rearranged channel response A block QR decomposition to obtain a unitary matrix and a block triangular matrix, and a received signal received by the plurality of receiving antennas based on the unitary matrix generated by the rearrangement block QR decomposition unit A received signal conversion unit that converts the matrix of the received signal, and a matrix of the received signal converted by the received signal conversion unit, the rearrangement block The block triangular matrix generated by the R decomposition unit is multiplied by a matrix of transmission signals transmitted from the plurality of transmission antennas, and the block triangular matrix is obtained by a plurality of block matrices obtained by decomposing the block triangular matrix. When configured, in the bottom row from the bottom row to the top row corresponding to the block matrix constituting the transformed reception signal matrix and the block triangular matrix, the transformed reception signal matrix All modulation candidate points for the bottom row element corresponding to the block matrix, the bottom row element corresponding to the block matrix of the block triangular matrix, and the bottom row element corresponding to the block matrix of the transmission signal matrix A metric is obtained based on the metric, a predetermined number of modulation candidate points having a small metric value are selected, and the lowest row is selected from the row following the bottom row. In rows up to a row, row elements corresponding to the block matrix of the transformed received signal matrix, row elements corresponding to the block matrix of the block triangular matrix, and the block matrix of the transmission signal matrix The metric is obtained based on all modulation candidate points for the element corresponding to, the predetermined number of modulation candidate points having a small metric value is selected, and the predetermined number selected in each row from the bottom row to the top row And a demodulating unit that demodulates the transmission signal based on a metric at a modulation candidate point, and the rearrangement block QR decomposition unit detects a norm for each of the column vectors constituting the channel response matrix A norm calculation unit that selects a column vector having the smallest norm, and a column vector selected by the norm calculation unit. A correlation detection unit configured to set a leftmost column in the answer matrix and detect a correlation value between the column vector other than the selected column vector and the selected column vector; and the correlation detection unit By rearranging column vectors other than the selected column vector in ascending order of detected correlation values, and setting them to the remaining columns excluding the left end in the channel response matrix, the channel responses are rearranged. A channel matrix conversion unit for converting to a subsequent channel response, and a rearranged channel response converted by the channel matrix conversion unit are subjected to block QR decomposition by block house holder conversion, and a unitary matrix and a block triangular matrix are converted into And a block house holder conversion unit to be obtained.

  The MIMO receiving apparatus according to the present invention is characterized in that the metric is a square metric or a Manhattan metric.

  Further, in the MIMO receiving apparatus according to the present invention, the demodulator may be configured such that, in the bottom row from the bottom row to the top row corresponding to the block matrix constituting the converted reception signal matrix and the block triangular matrix, The bottom row element corresponding to the block matrix of the transformed received signal matrix, the bottom row element corresponding to the block matrix of the block triangular matrix, and the bottom row corresponding to the block matrix of the transmission signal matrix Obtaining a Manhattan metric based on all the modulation candidate points for the element, selecting a predetermined number of modulation candidate points having a small value of the Manhattan metric, obtaining a square metric at the selected predetermined number of modulation candidate points, Corresponds to the block matrix of the transformed received signal matrix in the row from the next row to the top row of the bottom row Obtaining the Manhattan metric based on all modulation candidate points for the row elements corresponding to the block matrix of the block triangular matrix and the elements corresponding to the block matrix of the transmit signal matrix; Select a predetermined number of modulation candidate points having a small Manhattan metric value, obtain a square metric at the selected predetermined number of modulation candidate points, and select a predetermined number of modulation candidates selected in each row from the bottom row to the top row Based on a square metric at a point, likelihood information by soft decision is generated.

  As described above, according to the present invention, when MIMO demodulation is performed using block QR decomposition, it is possible to suppress degradation of transmission characteristics even when the correlation of the MIMO channel is large.

It is a block diagram which shows the structure of the receiver by embodiment of this invention. It is a block diagram which shows the structure of the rearrangement block QR decomposition | disassembly part. FIG. 11 is a diagram for describing processing of generating a channel response H ″ by rearranging column vectors constituting the channel response H in the rearrangement block QR decomposition unit. It is a block diagram which shows the structure of a likelihood information / demodulation result production | generation part. It is a flowchart which shows the process of a likelihood information / demodulation result production | generation part. It is a figure which shows a simulation result. It is the schematic which shows the whole structure of a MIMO-OFDM transmission system. It is a block diagram which shows the structure of the conventional receiver.

  Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings. The present invention proposes a new scheme capable of improving the transmission characteristics in the MIMO demodulation processing by the MIMO-OFDM transmission system using a plurality of transmitting antennas and a plurality of receiving antennas as compared with the conventional MIMO demodulation scheme. In particular, by applying the MIMO demodulation method of the present invention to a propagation environment where the correlation of the MIMO propagation path is large, deterioration of transmission characteristics is suppressed.

  The MIMO receiving apparatus according to the embodiment of the present invention, when generating a block triangular matrix by performing block QR decomposition on the propagation path response, reduces the correlation between elements in the block matrix constituting the block triangular matrix. Before performing decomposition, the column vectors constituting the channel response are rearranged. By using the MIMO demodulation method in units of this block matrix, the MIMO demodulation performance of each block matrix can be improved, and the transmission characteristics can be improved as compared with MIMO demodulation without rearrangement.

[MIMO receiver]
First, a receiving apparatus according to an embodiment of the present invention will be described. FIG. 1 is a block diagram showing a configuration of a receiving apparatus according to an embodiment of the present invention. This receiving apparatus (MIMO receiving apparatus) 1 includes four receiving antennas # 1 to # 4 (not shown), an FFT processing unit 21, a propagation path estimation unit 22, a rearrangement block QR decomposition unit 2, a received signal conversion unit 3, and a likelihood. Degree information / demodulation result generation unit (demodulation unit) 4. Note that only the components directly related to the present invention are shown in the receiving apparatus 1 of FIG. 1, and other components such as an error correction unit are omitted.

  The FFT processing unit 21 and the propagation path estimation unit 22 are the same as the components shown in FIG.

The rearrangement block QR decomposition unit 2 receives the propagation path response H from the propagation path estimation unit 22, and each of a plurality of column vectors h _i (i = 1, 2, 3, 4) constituting the propagation path response H. The total norm from each of the transmission antennas # 1 to # 4 is detected, and the smallest column vector having the smallest norm is selected from the plurality of column vectors. The rearrangement block QR decomposition unit 2 detects the correlation with respect to the column vector having the smallest norm for the column vectors other than the smallest column vector having the smallest norm, and rearranges them in order of increasing correlation. The rearrangement block QR decomposition unit 2 generates a channel response H ″ including the selected minimum column vector and the rearranged column vector.

The rearrangement block QR decomposition unit 2 performs block QR decomposition (H ′ = Q _b R _b ) on the generated propagation path response H ′ by block householder transformation to obtain a unitary matrix Q _b and a block triangular matrix R _b . The rearrangement block QR decomposition unit 2 outputs the unitary matrix Q _b to the received signal conversion unit 3, and converts the block triangular matrix R _b (elements R ₂₂ , R ₁₂ and R ₁₁ described later) into likelihood information / demodulation result generation unit. 4 is output.

  FIG. 2 is a block diagram showing a configuration of the rearrangement block QR decomposition unit 2. FIG. 3 shows a rearrangement block QR decomposition unit 2 that rearranges the column vectors constituting the propagation path response H to change the propagation path response H. It is a figure explaining the process which produces | generates ''. The rearrangement block QR decomposition unit 2 includes a norm calculation unit 41, a correlation detection unit 42, a channel matrix conversion unit 43, and a block house holder conversion unit 44.

The norm calculation unit 41 of the rearrangement block QR decomposition unit 2 includes, for example, four column vectors h _i (i = 1, 2, 3, 4) shown in the equation (1) from the propagation path estimation unit 22. Is input, the norm is obtained for each of the four column vectors h _i , and the smallest column vector h _min having the smallest column vector norm is selected as shown in the following equation. Then, the norm calculation unit 41 outputs the selected minimum column vector h _min and propagation path response H to the correlation detection unit 42. Here, as shown in FIG. 3, it is assumed that h _min = h ₃ .

ｈ’_minは、ｈ_minの複素共役転置を示す。 Correlation detector 42 receives the minimum column vector h _min (= h ₃₎ and channel response H from the norm calculating unit 41, a minimum column vector in the far left column of the channel response H h _min (= h ₃ ) And other three column vectors h ₁ , h ₂ , h ₄ other than the minimum column vector h _min are arranged in order to generate a new propagation path response H ′. Correlation detection unit 42, the minimum column vector h _min (= h ₃₎ other three columns except vectors _{h 1, h 2, h 4} , according to the following equation for the minimum column vector h _min (= h ₃₎ Correlation value Corr is detected. Then, the correlation detector 42 outputs the correlation values Corr and the channel response H ′ of the column vectors h ₁ , h ₂ , h _{4 to} the channel matrix converter 43. Here, as shown in FIG. 3, it is assumed that the correlation value Corr is small in the order of h ₂ , h ₁ , and h ₄ .

h ′ _min indicates a complex conjugate transpose of h _min .

The channel matrix conversion unit 43 receives the correlation values Corr and the channel response H ′ of the column vectors h ₁ , h ₂ , h ₄ from the correlation detection unit 42, and the leftmost minimum column vector in the channel response H ′. For the other three column vectors h ₁ , h ₂ , and h ₄ other than h _min (= h ₃ ), the channel response H ″ is generated by rearranging the correlation values Corr in ascending order. Then, the propagation path matrix conversion unit 43 outputs the propagation path response H ″ to the block house holder conversion unit 44. As shown in FIG. 3, the propagation path response H ″ is a matrix in which column vectors h ₃ , h ₂ , h ₁ , and h ₄ are set from left to right.

  As a result, the channel response H ″ has a plurality of column vectors constituting the channel response H rearranged, the column vector having the minimum norm is arranged in the leftmost column, and the leftmost column toward the rightmost column. Column vectors are arranged in ascending order of correlation with the column vector having the minimum norm. Thereby, the correlation between the elements of each block matrix becomes small.

The block house holder converting unit 44 receives the channel response H ″ from the channel matrix converting unit 43, and performs block QR decomposition on the channel response H ″ by block house holder conversion (H ″ = Q _b R _b ), A unitary matrix Q _b and a block triangular matrix R _b are obtained, the unitary matrix Q _b is output to the reception signal converting unit 3, and the block triangular matrix R _b (elements R ₂₂ , R ₁₂ and R ₁₁ described later) is a likelihood. Output to the information / demodulation result generation unit 4.

As a result, in the block triangular matrix R _b , the column vector having the minimum norm is arranged in the leftmost column in the same manner as the propagation path response H ″, and the column vector having the leftmost minimum norm is arranged toward the rightmost column. Column vectors are arranged in ascending order of correlation between.

When the block triangular matrix R _b is composed of 4 columns and 4 rows, the respective column vectors constituting the block triangular matrix R _b are elements r ₁₁ and r as shown in equation (19) described later. leftmost column vector of _21, a column vector whose elements r ₁₂ and r leftmost of right column vector of ₂₂ elements r ₁₃ in the column next to it, r _23, r ₃₃ and r _43, and the element It is constituted by a rightmost column vector consisting of r ₁₄ , r ₂₄ , r ₃₄ and r ₄₄ .

In this case, since the two column vectors adjacent to each other in block triangular matrix R _b is small correlation, in block triangular matrix R _b, elements r _11, r _21, r ₁₂ and r ₂₂ constituting the block matrix R ₁₁ to be described later Correlation is also reduced. Similarly, the correlation between elements r ₁₃ , r ₂₃ , r ₁₄ and r ₂₄ constituting a block matrix R ₁₂ which will be described later is also reduced, and elements r ₃₃ , r ₄₃ , r ₃₄ and r constituting a block matrix R ₂₂ which will be described later will be reduced. _{The 44} correlation is also reduced.

  Therefore, in the case of the block QR decomposition with rearrangement shown in FIG. 2 and FIG. 3, the correlation between the elements constituting the block matrix is smaller than in the case of the block QR decomposition without rearrangement. In a likelihood information / demodulation result generation unit 4 (described later) that performs processing in units of matrices, the selection accuracy of modulation candidate points to be selected in each block matrix is improved, and the accuracy of metrics used to generate likelihood information is also improved. .

Returning to FIG. 1, the received signal conversion unit 3 receives the data signal (received signal y) from the FFT processing unit 21 and also receives the unitary matrix Q _b from the rearrangement block QR decomposition unit 2, and the unitary matrix Q _b Is used to convert the received signal y into the received signal y ′ (y ′ = Q _b ^H y), and the received signal y ′ (elements y ′ ₃₄ and y ′ ₁₂ described later) is converted into a likelihood information / demodulation result generation unit 4. Output to.

The block house holder conversion formula can be expressed by the following formula.

The matrix I ₄ indicates a unit matrix of size 4, and the matrix V is a matrix extracted from the propagation path response H, and indicates a matrix having the same number of columns as the columns defined by the block triangular matrix R _b . The matrix H _b is a matrix obtained by converting the matrix V, and is calculated by performing an inverse matrix operation and an eigenvalue operation using the matrix V. In addition, the block house holder conversion is known, and the details are referred to the following documents.
“Rotella F, Zambettakis I,“ Block Householder Transformation for Parallel QR Factorization ”, Applied Mathematics Letters, Volume 12, No. 4, May. 1999, pp. 29-34”

The propagation path response H ″ is decomposed into a unitary matrix Q _b and a block triangular matrix R _b by using the equation (16) (H ″ = Q _b R _b ). Hereinafter, block QR decomposition that decomposes a channel response H ″ obtained by rearranging the column vectors of the channel response H of the transmission 4 and the reception 4 into a plurality of block matrices having two rows and two columns is taken as an example. explain.

ｎは雑音成分である。この場合、受信信号ｙ、送信信号ｘ及び雑音成分ｎの各要素は、並び替えた伝搬路応答Ｈ’’の要素に対応して並び替えられているものとする。 When block QR decomposition is performed on the propagation path response H ″ obtained by rearranging the propagation path response H in the equation (1), the propagation path response H ″ is expressed as a unitary matrix Q _b as shown in the following expression. And can be decomposed into a block triangular matrix R _b .

n is a noise component. In this case, it is assumed that the elements of the reception signal y, the transmission signal x, and the noise component n are rearranged corresponding to the elements of the rearranged propagation path response H ″.

Since the unitary matrix Q _b satisfies Q _b ^H Q _b = I, the following equation can be obtained. Matrix I represents a unit matrix.

Here, in the above formula (18), when each element is also described as Q _b ^H y = y ′ and Q _b ^H n = n ′, the following formula is obtained.

とし、雑音成分ｎ’を無視すると以下の式が得られる。

here,

When the noise component n ′ is ignored, the following equation is obtained.

As a result, the propagation path response H ″ having 4 rows and 4 columns is decomposed into a plurality of block matrices R ₁₁ , R ₁₂ and R ₂₂ having 2 rows and 2 columns. In other words, the block matrix R ₁₁ , R ₁₂ , R ₂₂ with 2 rows and 2 columns has four elements that make up the block triangular matrix R _b with 4 rows and 4 columns (one of which is It is a zero matrix.)

Such block QR decomposition is performed in units of the number of columns of the block matrix with respect to the propagation path response H ″, and therefore the decomposition process into the block matrix is repeatedly performed. The number of repetitions, number of columns and number of rows N _b of the block matrix, when the number of columns of the channel response H '' is N, the (N / N _b -1) times. When the channel response H ″ of the transmission 4 and the reception 4 is decomposed into a block matrix having 2 rows and 2 columns, one block house holder conversion is required.

  Here, the number of multiplications necessary to perform general QR decomposition on the propagation path response H ″ with 4 rows and 4 columns is the propagation response H ″ with 2 rows and 2 columns. Compared to the number of multiplications necessary for performing the block QR decomposition to be decomposed into block matrices, the amount of calculation can be suppressed to about 3/4 times.

In FIG. 1, likelihood information / demodulation result generation unit 4 receives reception signal y ′ (elements y ′ ₃₄ and y ′ _{12 of} reception signal y ′) from reception signal conversion unit 3, and rearrangement block QR decomposition. part 2 type of block triangular matrix R _b (block triangular matrix R _b elements (block matrix) R _22, R ₁₂ and R _11), the bottom line (the second line of the received signal y 'and block triangular matrix R _b ) To the top row (first row) for each block (each block of each element from the bottom row to the top row in the received signal y ′). Specifically, the likelihood information / demodulation result generation unit 4 performs all the modulations that can be performed on the transmission signal x in units of blocks in order from the bottom row to the top row of the received signal y ′ and the block triangular matrix R _b. Obtain a Manhattan metric with candidate points assigned, select a modulation candidate point from the Manhattan metric, obtain a demodulation result by hard decision, and obtain likelihood information by soft decision using a square metric only for the selected modulation candidate point .

  FIG. 4 is a block diagram showing a configuration of the likelihood information / demodulation result generation unit 4, and FIG. 5 is a flowchart showing processing of the likelihood information / demodulation result generation unit 4. The likelihood information / demodulation result generation unit 4 includes a modulation point recording memory 5, a MIMO demodulation unit 6, a candidate point selection unit 7, an interference component calculation unit 8, a subtraction unit 9, a MIMO demodulation unit 10, a candidate point selection unit 11, and Square metric calculation units 13 and 14 are provided.

The likelihood information / demodulation result generation unit 4 inputs the reception signal y ′ (elements y ′ ₃₄ and y ′ _{12 of the} reception signal y ′) from the reception signal conversion unit 3 and blocks from the rearrangement block QR decomposition unit 2 The triangular matrix R _b (elements R ₂₂ , R ₁₂ and R _{11 of the} block triangular matrix R _b ) is input (step S501).

First, the MIMO demodulator 6, the candidate point selector 7, and the square metric calculator 13 are modulation candidates as processing of the first block for the received signal y ′ and the second row that is the bottom row of the block triangular matrix R _b. A Manhattan metric is calculated to select points, and a square metric is calculated to generate likelihood information. That is, the processing of the first block includes the received signal y ′ (column vector of elements y ′ ₁₂ and y ′ ₃₄ ) and the block triangular matrix R _b (elements of R ₁₁ , R ₁₂ and R) in the above equation (20). ₂₂ matrix), the processing is based on the bottom row element (formula (21) described later) corresponding to the block matrix R ₂₂ which is the bottom row element.

このように、ブロック行列Ｒ₂₂は行数２及び列数２の行列であるため、行数４及び列数４の行列よりも少ない演算量でＭＩＭＯ復調を行うことできる。 The block in the second row, which is the bottom row of the received signal y ′ and the block triangular matrix R _b , is expressed by the following equation.

Thus, since the block matrix R ₂₂ is a matrix having 2 rows and 2 columns, MIMO demodulation can be performed with a smaller amount of computation than a matrix having 4 rows and 4 columns.

The MIMO demodulator 6 receives the received signal y ′ ₃₄ that is an element of the lowermost row (second row) of the received signal y ′ from the received signal converter 3, and also receives a block triangular matrix R from the rearrangement block QR decomposition unit 2. The element R ₂₂ in the lowermost row (second row) of _b is input, and all the modulation candidate points C of the transmission signal x ₃₄ are input from the modulation point recording memory 5. Then, the MIMO demodulator 6 obtains the elements r ₃₃ , r ₃₄ , r ₄₃ and r ₄₄ constituting the element R ₂₂ of the block triangular matrix R _b and the transmission signals x ₃ and x ₄ constituting the transmission signal x _34. The right side of the equation (21) is calculated for all modulation candidate points C that the transmission signals x ₃ and x ₄ can take, and r ₃₃ x ₃ , r ₃₄ x ₄ , r ₄₃ x ₃ and r ₄₄ x Ask for ₄ . The MIMO demodulator 6 then calculates all modulation candidate points C that can be taken by the transmission signals x ₃ and x ₄ for r ₃₃ x ₃ , r ₃₄ x ₄ , r ₄₃ x _3, and r ₄₄ x _{4 according} to the following equations. The received replica signal is generated by adding the combinations of the received signal y ′ ₃₄ and the received signals y ′ ₃ and y ′ _{4 and the} received replica signal that constitute the received signal y ′ ₃₄ , and all the transmission signals x ₃ and x ₄ can take. The Manhattan metric E ₃₄ is obtained for the modulation candidate point C (step S502).

  In the equation (22), real (X) represents the real part of the complex number X, and imag (X) represents the imaginary part of the complex number X. The same applies to equation (26) described later. This Manhattan metric is a value obtained by adding the absolute values of each real part and imaginary part in the difference between the received signal y 'and the received replica signal.

Candidate point selection section 7, of the Manhattan metric E ₃₄ for all modulation candidate points C of the transmission signal x ₃ and x ₄ obtained by the MIMO demodulator 6 may take, _{the 34} M from ascending order of Manhattan metric E ₃₄ And a combination of transmission signals (x ₃ , x ₄ ) in M ₃₄ Manhattan metrics E ₃₄ (modulation candidate points of transmission signals x ₃ and x ₄ ) is selected (step S503). _M34 is a preset value.

The square metric calculation unit 13 squares the combination of transmission signals (x ₃ , x ₄ ) in the selected M ₃₄ Manhattan metrics E ₃₄ (modulation candidate points of the transmission signals x ₃ and x ₄ ) according to the following equation. A metric E ′ ₃₄ is obtained (step S504).

  The square metric is a value obtained by calculating the sum of the square of the real part and the square of the imaginary part in the difference between the received signal y 'and the received replica signal.

Next, the interference component calculation unit 8, the subtraction unit 9, the MIMO demodulation unit 10, the candidate point selection unit 11, and the square metric calculation unit 14 perform the reception signal y ′ and the first row that is the top row of the block triangular matrix R _b. As a process of the second block, a Manhattan metric is calculated to select a modulation candidate point, and a square metric is calculated to generate likelihood information. That is, the processing of the second block includes the received signal y ′ (column vector of elements y ′ ₁₂ and y ′ ₃₄ ) and the block triangular matrix R _b (elements of R ₁₁ , R ₁₂ and R) in the above equation (20). ₂₂ matrix), the process is based on the top row element (formula (24) described later) corresponding to the block matrix R ₁₁ , R ₁₂ which is the top row element.

The block of the first row that is the top row of the received signal y ′ and the block triangular matrix R _b is expressed by the following equation. Using the combination of M ₃₄ transmission signals (x ₃ , x ₄ ) selected in ascending order of Manhattan metric, the following equation is calculated.

The interference component calculation unit 8 receives the element R ₁₂ of the uppermost row (first row) of the block triangular matrix R _b from the rearrangement block QR decomposition unit 2 and M ₃₄ transmission signals (from the candidate point selection unit 7). A combination of x ₃ , x ₄ ) is input, and an interference component is obtained by multiplying element R ₁₂ by a combination of M ₃₄ transmission signals (x ₃ , x ₄ ).

The subtraction unit 9 receives the reception signal y ′ ₁₂ that is the element of the uppermost row (first row) of the reception signal y ′ from the reception signal conversion unit 3 and receives the interference component from the interference component calculation unit 8 to receive The interference component is subtracted from the signal y ′ ₁₂ and the subtraction result y ″ is output to the MIMO demodulator 10. The subtraction result y ″ (column vector whose elements are y ^″ ₁ and y ^″ ₂ ) is expressed by the following equation.

The MIMO demodulator 10 receives the element R ₁₁ of the uppermost row (first row) of the block triangular matrix R _b from the rearrangement block QR decomposition unit 2 and the subtraction result y ″ from the subtractor 9. Then, all the modulation candidate points C of the transmission signal x ₁₂ are input from the modulation point recording memory 5, and the Manhattan metric E ₃₄ at the selection candidate point of the selected M ₃₄ transmission signals x ₃₄ is input from the candidate point selection unit 7. To do. Then, the MIMO demodulator 10 obtains the elements r ₁₁ , r ₁₂ , r ₂₁ and r ₂₂ constituting the element R ₁₁ of the block triangular matrix R _b and the transmission signals x ₁ and x ₂ constituting the transmission signal x _12. The right side of the equation (25) is calculated for all modulation candidate points C that the transmission signals x ₁ and x ₂ can take, and r ₁₁ x ₁ , r ₁₂ x ₂ , r ₂₁ x ₁ and r ₂₂ x Ask for ₂ . The MIMO demodulator 10 then calculates all the modulation candidate points C that can be taken by the transmission signals x ₁ and x ₂ for r ₁₁ x ₁ , r ₁₂ x ₂ , r ₂₁ x ₁ and r ₂₂ x _{2 according} to the following equations. adding together the combinations as to generate a reception replica signal of the subtraction result y ^'with' ₁ and y ^'' ₂ and reception replica signal 'received signal y constituting ^the' take the transmission signals x ₁ and x ₂ Manhattan metrics are obtained for all obtained modulation candidate points C, and Manhattan metrics E ₃₄ (selection of M ₃₄ transmission signals x ₃₄ ) in the transmission signals x ₃ and x ₄ corresponding to the reception signals y ^″ ₁ and y ^″ _2. One of M ₃₄ Manhattan metrics E ₃₄ corresponding to the candidate points is added to obtain Manhattan metric E ₁₂ (step S505).

Candidate point selection unit 11, of the Manhattan metric E ₁₂ obtained by the MIMO demodulator 10, to select the ₁₂ M from ascending order of Manhattan metric E _12, a transmission signal in M ₁₂ Manhattan metric E ₁₂ (x ₁ , X ₂ , x ₃ , x ₄ ) (modulation candidate points for transmission signals x ₁ and x ₂ ) are selected. M ₁₂ is a preset value. That is, the candidate point selection unit 11 combines the transmission signals (x ₁ , x ₂ ) corresponding to M ₁₂ Manhattan metrics E ₁₂ (modulation candidate points of the transmission signals x ₁ and x ₂ ) and the transmission signal (x ₃ , X ₄ ) (modulation candidate points for transmission signals x ₃ and x ₄ ) are selected (step S506).

The square metric calculation unit 14 combines the transmission signals (x ₁ , x ₂ ) corresponding to the selected M ₁₂ Manhattan metrics E ₁₂ (modulation candidate points of the transmission signals x ₁ and x ₂ ) and the transmission signal (x ₃ , X ₄ ) (square modulation metric E ′ ₁₂ is obtained by the following equation for the modulation candidate points of transmission signals x ₃ and x ₄ ) (step S507).

The candidate point selection unit 11 specifies the smallest Manhattan metric E among the Manhattan metrics E ₁₂ obtained by the MIMO demodulation unit 10 and combines the transmission signals (x ₁ , x ₂ , x ₃ , x ₄ ) in the smallest E (Modulation candidate points of transmission signals x ₁ , x ₂ , x _3, and x ₄ ) are obtained and output as hard decision demodulation results (step S 508).

The square metric calculator 14 outputs the M ₁₂ square metrics E ′ ₁₂ obtained in step S507 as likelihood information for soft decision (step S509). As the likelihood information, for example, bit likelihood or log likelihood ratio (LLR) is obtained. By using the likelihood information for error correction decoding, a demodulation result by soft decision can be obtained.

(simulation result)
Next, simulation results in the new scheme (block QRM-MLD scheme with rearrangement) and the block QRM-MLD scheme without rearrangement according to the embodiment of the present invention will be described. FIG. 6 is a diagram showing the simulation result. The vertical axis represents the bit error rate, and the horizontal axis represents the received CNR [dB]. a shows a simulation result of the number M (M ₃₄ , M ₁₂ ) of selection of modulation candidate points in the new system according to the embodiment of the present invention, and b shows a modulation candidate in the block QRM-MLD system without rearrangement. The simulation result of the point selection number M = 2 is shown, c is the simulation result of the modulation candidate point selection number M (M ₃₄ , M ₁₂ ) = 4 in the new scheme according to the embodiment of the present invention, and d is The simulation result of the selection number M = 4 of modulation candidate points in the block QRM-MLD system without rearrangement is shown, and e shows the simulation result of the conventional MLD system (using a square metric). Each of a to e indicates a case where the transmission correlation and the reception correlation of the MIMO propagation path are 0.8. In order to generate likelihood information, the block QRM-MLD scheme without rearrangement performs block QR decomposition without rearranging the channel responses H, calculates a Manhattan metric to select a modulation candidate point, and generates likelihood information. This is a method for calculating a square metric.

  FIG. 6 shows a comparison between the new scheme (block QRM-MLD scheme with rearrangement) and the block QRM-MLD scheme without rearrangement when the same selection number M is used according to the embodiment of the present invention. It can be seen that the transmission characteristics of the new scheme that performs the replacement are improved by about 0.5 dB compared to the scheme that does not perform the rearrangement.

As described above, according to the receiving device 1 according to the embodiment of the present invention, the rearrangement block QR decomposition unit 2 detects the norm of the column vector constituting the channel response H and selects the column vector having the smallest norm. Then, with respect to the column vectors other than the selected column vector, correlations with respect to the column vector having the smallest norm are detected and rearranged in ascending order of correlation. Then, the rearrangement block QR decomposition unit 2 generates a channel response H ″ composed of the selected column vector and the rearranged column vector, and performs block QR decomposition on the channel response H ″ by block house holder transformation ( H ″ = Q _b R _b ), unitary matrix Q _b and block triangular matrix R _b are obtained. Further, the reception signal conversion unit 3 converts the reception signal y into the reception signal y ′ using the unitary matrix Q _b . Then, the likelihood information / demodulation result generation unit 4 sequentially applies all the modulation candidate points C that can be taken for the transmission signal x in units of blocks in order from the bottom row to the top row of the received signal y ′ and the block triangular matrix R _b. To obtain a Manhattan metric substituting, select a modulation candidate point from the Manhattan metric, obtain a demodulation result by hard decision, and obtain a likelihood information by soft decision using a square metric only for the selected modulation candidate point did.

As a result, the propagation path response H ″ obtained by performing the rearrangement by the rearrangement block QR decomposition unit 2 has a smaller correlation between the adjacent column vectors, so that the two adjacent column vectors in the block triangular matrix R _b Correlation is also reduced. That is, in a plurality of block matrices constituting the block triangular matrix R _b , the correlation between elements in the block matrix becomes small. Therefore, in the likelihood information / demodulation result generation unit 4 that performs processing in units of block matrices (in units of the number of columns of the block matrix), the accuracy of selection of modulation candidate points is improved, and the accuracy of metrics used for generating likelihood information Will be improved.

  In addition, according to the receiving device 1 according to the embodiment of the present invention, as shown in the simulation result of FIG. Can do. Therefore, when performing MIMO demodulation using block QR decomposition, it is possible to suppress degradation of transmission characteristics even when the correlation of the MIMO propagation path is large.

  The present invention has been described with reference to the embodiment. However, the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the technical idea thereof. For example, in the above-described embodiment, in the transmission-four and reception-four MIMO-OFDM transmission system, the block response QR decomposition for decomposing the channel response H with 4 rows and 4 columns and the block matrix with 2 rows and 2 columns. However, the present invention is also applicable to the channel response H having a larger number of rows and columns, and is also applicable to the case where the size of the block matrix is changed. There is.

  In the above-described embodiment, the Manhattan metric is used for selecting the modulation candidate point. However, for example, another metric having a lower processing load than the square metric is used, such as the Manhattan-Schebychev metric. It may be. In short, any metric that has a lower processing load than the square metric, such as a metric that performs an operation that does not include multiplication, may be used.

  In the embodiment, the Manhattan metric is used for selecting the modulation candidate point, and the square metric is used for generating the likelihood information. However, for selecting the modulation candidate point and generating the likelihood information, Only the Manhattan metric may be used, or only the square metric may be used.

1,200 Receiving Device 2 Rearrangement Block QR Decomposition Unit 3, 24 Received Signal Conversion Unit 4, 25 Likelihood Information / Demodulation Result Generation Unit 5, 26 Modulation Point Recording Memory 6, 10 MIMO Demodulation Unit 7, 11, 28, 32 Candidate point selection unit 8, 29 Interference component calculation unit 9, 30 Subtraction unit 13, 14 Square metric calculation unit 21 FFT processing unit 22 Channel estimation unit 23 Rearrangement QR decomposition unit 27, 31 Metric calculation unit 41 Norm calculation unit 42 Correlation Detection unit 43 Propagation path matrix conversion unit 44 Block house holder conversion unit 100 Transmitter 101 S / P conversion unit 102 MIMO-OFDM modulation unit 201 MIMO-OFDM demodulation unit 202 P / S conversion unit 300 MIMO-OFDM transmission system

Claims (3)

Signals transmitted from a plurality of transmission antennas are received via a plurality of reception antennas, propagation path responses between the plurality of transmission antennas and the plurality of reception antennas are estimated, and a matrix of the propagation path responses is QR In a MIMO receiving apparatus that performs decomposition and MIMO demodulation,
Reordering column vectors constituting the matrix of propagation path responses, performing block QR decomposition on the rearranged propagation path responses, and obtaining a unitary matrix and block triangular matrix; a reordering block QR decomposition unit;
A reception signal conversion unit that converts a matrix of reception signals received by the plurality of reception antennas based on a unitary matrix generated by the rearrangement block QR decomposition unit;
A matrix of reception signals converted by the reception signal conversion unit is obtained by multiplying a block triangular matrix generated by the rearrangement block QR decomposition unit by a matrix of transmission signals transmitted from the plurality of transmission antennas. When the block triangular matrix is composed of a plurality of block matrices obtained by decomposing the block triangular matrix,
In the bottom row from the bottom row to the top row corresponding to the block matrix constituting the transformed reception signal matrix and the block triangular matrix, the bottom row corresponding to the block matrix of the transformed reception signal matrix. Obtain metrics based on all modulation candidate points for the bottom row element, the bottom row element corresponding to the block matrix of the block triangular matrix, and the bottom row element corresponding to the block matrix of the transmit signal matrix , Select a predetermined number of modulation candidate points with a small value of the metric,
In the rows from the next row to the top row of the bottom row, the row elements corresponding to the block matrix of the transformed received signal matrix, the row elements corresponding to the block matrix of the block triangular matrix, and Obtaining the metric based on all modulation candidate points for elements corresponding to the block matrix of the transmit signal matrix, selecting a predetermined number of modulation candidate points with a small metric value;
A demodulator that demodulates the transmission signal based on a metric at a predetermined number of modulation candidate points selected in each row from the bottom row to the top row;
The rearrangement block QR decomposition unit
Calculating a norm for each of the column vectors constituting the matrix of the channel response, and a norm calculating unit for selecting a column vector having the smallest norm;
The column vector selected by the norm calculation unit is set to the leftmost column in the channel response matrix, and each column vector other than the selected column vector is between the selected column vector. A correlation detection unit for detecting a correlation value;
By rearranging column vectors other than the selected column vector in ascending order of correlation values detected by the correlation detection unit, and setting the remaining column except the left end in the channel response matrix, the channel A propagation path matrix conversion unit for converting a response to the rearranged propagation path response;
A block house holder conversion unit that performs block QR decomposition on the rearranged propagation path responses converted by the propagation path matrix conversion unit by block house holder conversion and obtains a unitary matrix and a block triangular matrix ; MIMO receiving device. The MIMO receiver according to claim 1 , wherein
A MIMO receiving apparatus, wherein the metric is a square metric or a Manhattan metric. The MIMO receiver according to claim 1 , wherein
The demodulator
In the bottom row from the bottom row to the top row corresponding to the block matrix constituting the transformed reception signal matrix and the block triangular matrix, the bottom row corresponding to the block matrix of the transformed reception signal matrix. Manhattan metrics based on all modulation candidate points for the bottom row element, the bottom row element corresponding to the block matrix of the block triangular matrix, and the bottom row element corresponding to the block matrix of the transmit signal matrix. Determining, selecting a predetermined number of modulation candidate points having a small value of the Manhattan metric, determining a square metric at the selected predetermined number of modulation candidate points,
In the rows from the next row to the top row of the bottom row, the row elements corresponding to the block matrix of the transformed received signal matrix, the row elements corresponding to the block matrix of the block triangular matrix, and The Manhattan metric is obtained based on all modulation candidate points for the elements corresponding to the block matrix of the transmission signal matrix, a predetermined number of modulation candidate points having a small value of the Manhattan metric are selected, and the selected predetermined Find the square metric at a number of modulation candidate points,
A MIMO receiving apparatus, characterized by generating likelihood information by soft decision based on a square metric at a predetermined number of modulation candidate points selected in each line from the bottom line to the top line.

Images