WO2006070834A1 - Mimo receiving apparatus, mimo communication system, and mimo receiving method - Google Patents

Abstract

To prevent occurrence of any transmission rate reduction, which would otherwise be caused by an overhead using a known symbol, even if the number of transmission/reception antennas is increased so as to increase the number of users to be accommodated. A signal separating part (211) is provided which uses an independent component analyzing method to perform a signal separation. In this way, signals mixed with each other on a transmission path can be separated from each other even in the absence of any known symbol for signal separation. Thus, the occurrence of any transmission rate reduction can be prevented which would otherwise be caused by an overhead using a known symbol for signal separation.

Description

 MIMO receiving apparatus, MIMO communication system, and MIMO receiving method

 [0001] The present invention relates to a MIMO receiver, a MIMO communication system, and a MIMO reception method used in a wireless communication system such as a wireless LAN (Local Area Network) or a cellular system.

 Background art

 [0002] In a wireless communication system, channel multiplexing techniques for realizing higher speed transmission or accommodating more users are roughly classified as follows from the viewpoint of physical resources such as frequency, time, and space. Each channel multiplexing technology has the following tendencies when the number of channels is increased with the same transmission rate per channel.

 [0003] -FDM (Frequency Division Multiplex): The required bandwidth increases according to the number of channels.

 • TDM (Time Division Multiplex): The transmission clock increases with the number of channels, and the bandwidth increases.

 • CDM (Code Division Multiplex): The spreading ratio needs to be increased according to the number of channels, which increases the bandwidth.

 • SDM (Space Division Multiplex): The number of antennas increases with the number of channels (MIM

O transmission), but the band does not increase.

[0004] Thus, SDM is the only method that does not increase the bandwidth even if the number of channels increases.

MO (Multiple-Input Multiple-Output) transmission must be put into practical use in the future! Currently, it is actively researched as a technology. Regarding MIMO transmission technology, for example, patent literature

There is one disclosed in 1.

[0005] Figure 1 shows a schematic diagram of the currently studied MIMO transmission. Fig. 1A shows the schematic configuration of the transmitter, Fig. 1B shows the frame configuration of the transmission signal, and Fig. 1C shows the schematic configuration of the receiver.

[0006] Transmitter 10 converts m-sequence signals formed by MIMO transmitter 11 into radio signals by transmission radio circuits 12-1 to 12-m, respectively, and then transmits m antennas TAN-1 to T AN—Send from m

 [0007] The receiver 20 converts the signals received by 11 (11≥111) antennas 1 ^^^ 1 to 1 ^^ 11 into baseband signals by the receiving radio circuits 21-1 to 21-n, respectively. After conversion, the data is input to the MIM receiver 22. The MIMO receiver 22 inputs the outputs of the reception radio circuits 21-1 to 21-n to the switch circuit 24. The signal output from the reception radio circuit 21-1 is input to the frame synchronization unit 23.

 [0008] The frame synchronization unit 23 controls the switch circuit 24 based on a synchronization symbol included in the received signal. As a result, the signal in the known symbol interval is input to the transfer function estimation units 25-1 to 25-n, and the signal in the data interval is input to the signal separation unit 26. Incidentally, the MIMO receiver 22 in FIG. 1C is configured on the assumption that the frames of the received signals are identical.

 [0009] The transfer function estimators 25-1 to 25-n estimate the transfer functions between the transmitting and receiving antennas based on the known symbols transmitted from the transmitting antennas TAN-1 to TAN-m. The transfer function is sent to the signal separator 26. The signal separation unit 26 separates signals mixed on the transmission path using a transfer function between the transmission and reception antennas.

 [0010] A specific description will be given. As shown in FIG. 1B, one transmission frame is divided into a known symbol transmission period and a data transmission period. In the known symbol transmission period, known symbols from the antennas TAN 1 to TAN-m are transmitted at different times (in a time division manner). During the data transmission period, multiple data symbols are transmitted from each antenna TAN-l to TAN-m at the same time. In the example shown in FIG. 1B, each antenna Dinghachi-1 to Dinghachi-111 sends a TDM (Time Division Multiple) signal for k channels.

 [0011] Signals transmitted from antennas TAN-1 to TAN-m are mixed on the transmission path and received by n antennas RAN-1 to RAN-n of receiver 20. The MIMO receiver 22 estimates the transfer function by detecting the known symbol pattern distortion for each receiving antenna by the transfer function estimation units 25-l to 25-n, and separates the signal based on the estimation by the signal separation unit 26. To do.

[0012] Here, when the service area is narrow and frequency selective phasing due to a multipath transmission path delay difference can be ignored, the received known symbol value is divided by the transmitted known symbol value. Therefore, the transfer function is obtained as a single complex number. Therefore, the following equation is established between the transmission symbol and the reception symbol, and signal separation can be performed by solving this simultaneous linear equation for the transmission symbol.

 [Number 1]

- -m 【ま j番目の送信アンテナからの送信シンボル Xi, i = l, 2,…! ! Is the received symbol at the i-th receiving antenna

--m [m Transmit symbol from jth transmit antenna

 Wij is the transfer coefficient between the jth transmit antenna and the ith receive antenna

 [0013] In other words, the SDM scheme performs user signal separation based on the difference in radio wave propagation path between a plurality of transmission / reception antennas. Specifically, a known symbol signal that is different for each transmission antenna is added. The signal is sent out, and the inverse characteristics of the total transfer function between the transmitting and receiving antennas are obtained by detecting how the known symbol signal is distorted and received at each receiving antenna, and the signal is obtained by multiplying this by the received signal. Isolate. Therefore, in this method, by increasing the number of transmitting and receiving antennas, the number of accommodated users can be increased without increasing the transmission band.

 Patent Document 1: Japanese Patent Laid-Open No. 2001-237751

 Disclosure of the invention

 Problems to be solved by the invention

[0014] However, in the conventional SDM system, since a known symbol of each transmission antenna is assumed to be received without interference from other signals, the moment when a certain transmission antenna transmits a known symbol. The other antennas need to stop transmitting. For this reason, there is a problem that if the number of antennas is increased and the number of antennas is increased in order to realize a full-scale service, the known symbol transmission time increases, and high-speed communication becomes impossible due to the overhead. It was. That is, as is clear from FIG. 1B, when the number of transmission antennas is increased, the known symbol transmission period in one frame is increased and the data transmission period is shortened. As a result, the transmission speed decreases. [0015] For example, assuming outdoor cafes, when one user including passages such as the area for each (1. 8m) ² in the area of radius 20m is assumed to be distributed, approximately 384 users (π Χ 20 ^2/1. 8 ² ) needs to be accommodated, but if it is divided into 6 sectors and communicated, it must be able to accommodate 64 users per sector (= 384 4Z6). In this case, if 8 multiplex TDM is performed, MIMO transmission with 8 antennas (= 64Z8) is required. Therefore, if the ratio of the known symbol transmission time per antenna to the data transmission time is 1Z8 or higher, data cannot be sent during MIMO transmission.

 [0016] Also, in an uplink in which each transmitting antenna is of a different terminal (that is, in a system in which a signal transmitted from m terminals is received by one receiver and a signal from each terminal is separated) However, since it is indispensable for each terminal to synchronize the frame, a feedback mechanism between transmission and reception such as time alignment is required. As a result, the system becomes complex and the synchronization accuracy must be taken into account, which can cause a further decrease in transmission speed.

 [0017] An object of the present invention is to prevent a decrease in transmission speed due to overhead due to known symbols even when the number of transmission / reception antennas is increased in order to increase the number of accommodated users compared to the conventional method. It is to provide a MIMO receiving apparatus, a MIMO communication system, and an Ml MO receiving method.

 Means for solving the problem

 [0018] The MIMO receiving apparatus of the present invention inputs a plurality of received signals received by a plurality of antennas, and performs blind signal separation processing using an independent component analysis method on the plurality of received signals. A configuration is adopted that includes signal separation means for obtaining a plurality of independent separated signals and demodulation means for demodulating the plurality of separated signals obtained by the signal separation means.

 [0019] According to this configuration, since signal separation is performed using the independent component analysis method, signal separation is performed even when there is no known symbol for signal separation, compared to conventional signal separation processing. Will be able to. As a result, it is not necessary to arrange known symbols for signal separation in a time division manner, so that the data transmission period can be increased.

[0020] Further, the MIMO receiving apparatus of the present invention employs a configuration further comprising order correcting means for rearranging the order of the plurality of separated signals obtained by the signal separating means. [0021] According to this configuration, it is possible to correct the indefiniteness of the signal sequence caused by the independent component analysis method, and thus it is possible to obtain the correct sequence data.

 The invention's effect

 [0022] According to the present invention, even if there is no known symbol for signal separation, it becomes possible to separate signals mixed on the transmission path, so that transmission speed reduction due to overhead due to known symbols does not occur. It is possible to realize a MIMO receiver and a MIMO reception method.

 Brief Description of Drawings

 1 is a diagram for explaining the configuration and operation of a conventional MIMO transmission system. FIG. 1A shows a configuration of a transmitter, FIG. 1B shows a frame configuration of a transmission signal, and FIG. 1C shows a receiver. Diagram showing the configuration of

 [Figure 2] Diagram for explaining the principle of MIMO transmission

 FIG. 3 is a diagram for explaining the difference between independent and uncorrelated, FIG. 3A shows signals that are independent from each other, and FIG. 3B is a diagram that shows signals that are uncorrelated but independent from each other.

 [Figure 4] Block diagram showing a configuration example for performing signal separation processing by the independent component analysis method. [Figure 5] This is a two-dimensional diagram showing changes in the received signal distribution by Sphering. 5B shows the signal after DC removal, and FIG. 5C shows the signal after whitening.

 [Figure 6] List of independent component analysis algorithms

 7 is a diagram for explaining the configuration and operation of Embodiment 1, FIG. 7A shows the configuration of the transmitter, FIG. 7B shows the frame configuration of the transmission signal, and FIG. 7C shows the configuration of the receiver. FIG. 8 is a diagram for explaining the configuration and operation of Embodiment 2. FIG. 8A shows the configuration of the transmitter, FIG. 8B shows the frame configuration of the transmission signal, and FIG. 8C shows the configuration of the receiver. FIG. 9 is a diagram for explaining the configuration and operation of Embodiment 3. FIG. 9A shows the configuration of the transmitter, FIG. 9B shows the frame configuration of the transmission signal, and FIG. 9C shows the configuration of the receiver. The best mode for carrying out the invention

[0024] If the inventor of the present invention applies the independent component analysis method already proposed in the field of speech or the like to the Ml MO receiver, the signal separation process can be performed even if there is no known symbol for signal separation. Therefore, the present invention has been achieved. That is, the gist of the present invention is that signal separation is performed using an independent component analysis method in the Ml MO receiver. The present invention also proposes various devices for performing processing using the independent component analysis method in the MIMO receiver.

 [0025] When MIMO reception processing by the independent component analysis method is performed as in the present invention, it is not necessary to add a known symbol for signal separation to the transmission signal. In the following description, a scheme without adding known symbols for signal separation is called blind MIMO transmission.

 [0026] (1) Principle

 First, before describing the configuration of the embodiment, the principle of the present invention will be described.

[0027] (1-1) Classification of separation processing in MIMO receiver

 Figure 2 shows a schematic diagram of MIMO transmission. In Fig. 2, transmitter 30 transmits M channel signal s (t) s (t) generated by M channel signal generator 31 using M antennas.

 1 M

 To do. These transmission signals s (t) to s (t) are mixed in the transmission path. And 40 receivers

 1 M

 N signals X (t) to x (t) are received by mixing N transmission signals in the transmission path by N antennas. This process is expressed as x (t) = As (t) using matrix A.

1 N

 It is. The signal separator 41 provided in the receiver 40 multiplies the received signal x (t) by the matrix W to obtain N separated signals y (t) to y (t).

 1 N

[0028] Here, blind MIMO transmission receiver 40 does not know the number M of transmission channels, and therefore M and N are generally assumed to have different forces M≤N. If M = N and there is an inverse matrix of matrix A, then when W = A ^_1 , y (t) = s (t) and ideal signal separation is realized. In addition, when there is no inverse matrix or when M <N, it is considered that the generalized inverse matrix is optimal instead of the inverse matrix of matrix A. However, for simplicity, the following explanation assumes that there is an inverse matrix of matrix A with M = N.

 [0029] In the receiver 40, a known symbol inserted in each transmission signal s (t) to s (t) is used.

 1 M

The power of obtaining WA ^_1 and using this W to obtain the separation signal y (t) to y (t)

 1 M

 This is separation processing in a general MIMO receiver.

[0030] In contrast, in order to realize blind MIMO transmission, the inventor of the present invention determines a principal factor analysis method that determines the separated signals y (t) to y (t) to be uncorrelated with each other. (Hereafter this Simply using PCA (sometimes referred to as Principal Component Analysis) and independent component analysis methods that determine that the separated signals are independent of each other (hereinafter this is sometimes simply called ICA (Independent Component Analysis)). ) Was considered as a candidate.

 [0031] Based on the following considerations, it was considered that blind MIMO could be realized satisfactorily by using the independent component analysis method.

 [0032] (1 2) Difference between independence and correlation

 Roughly speaking, the difference between PCA and ICA is whether they are separated by force independence, which is uncorrelated, and is largely different from uncorrelated and independent in the following points. According to the results, “the power that is uncorrelated if it is independent is not necessarily independent because it is uncorrelated”.

 [0033] When uncorrelated and independent are expressed by equations, the signals s (t) and s (t) with an average value of 0 are expressed as

 1 2

 Become. However, Ε [·] indicates a set average.

[Equation ₂ ] Uncorrelated: E [ _Sl .s ₂ ] = 0 (2)

[Equation 3]

 (3) Independent: p (si, S2) = p (si) -pvs2

[0034] Here, as shown in the following equation, if equation (3) holds and is independent, it is certainly uncorrelated.

 [Equation 4]

E [s ^ 82]-/ s ₁ -s ₂ p (s ₁₎ s ₂ ) ds ₁ ^, ds ₂ = / s ^ ps ^ dsx ■ / s ₂ p (s ₂ no ds ₂ = E [s · E [s ₂ ] = 0

(Four)

 However, even if equation (2) holds and is uncorrelated, the independent condition of equation (3) does not always hold.

 This will be briefly described with reference to FIG. If the signals s (t) and s (t), which are uniformly distributed from -1.0 to +1.0 with an average of 0, are independent of each other, the distribution is as shown in Fig. 3A. this

 1 2

The signals s '1 (t) and s' 2 (t) obtained by converting the signal pair counterclockwise by 45 ° are Next formula

 [Equation 5]

E [s, ₁ .s'J = E Ίξ (st)-S2 (t) (si (t) + s ₂ (t)) = (E [si ² ] E [S2 ² ]) = 0

(5) More uncorrelated.

[0037] However, the distribution of s' (t) and s' (t) is as shown in Fig. 3B. For example, s' (t) has a large value.

 1 2 1

 S '(t) must take a small value and vice versa,

 2

 Matching. So s '(t) and s' (t) are uncorrelated but not independent.

 1 2

 [0038] In this way, “the power is uncorrelated if it is independent, it is not always independent because it is not correlated”, so it can be said that ICA is a stricter signal separation algorithm than PCA.

 [0039] As described above, in the present invention, since signal separation using the independent component analysis method (ICA) is performed, compared with the case where blind separation is performed using PCA, better signal separation results are obtained. You can get fruits.

 [0040] (1-3) Independent Component Analysis (ICA)

 (1 3— 1) Overview

 Figure 4 shows a configuration example for realizing signal separation processing by ICA. First, N systems of received signals x (t) received by N antennas are input to a Sphering processing unit 50 as a preprocessing unit. Incidentally, the reception signal x (t) is a baseband signal after reception radio processing.

[0041] In the Sphering processing unit 50, first, the DC component of each received signal is removed by the centering unit 51, and the eigenvalues and eigenvectors of the covariance matrix Rxx of the received signal vector X are obtained by the SΛ matrix calculating unit 53. Is calculated. Then, by calculating x ′ = Sx by the Whitening unit 52, the received signal x (t) after DC removal is converted into an unrelated received signal vector x ′ with the same power. The signal subjected to such preprocessing is input to the ICA processing unit 60.

Here, FIG. 5 shows, in a two-dimensional image, a state in which the distribution power of the received signal of N = 2 is changed by the Sphering processing unit 50 as a preprocessing unit. The signal shown in Fig. 5A is centered as shown in Fig. 5B by the centering unit 51, and the state shown in Fig. 5C is further executed by the whitening unit 52. It is said. Incidentally, whitening here is not to flatten the frequency spectrum but to make the eigenvalues uniform as shown in Fig. 5C.

 The ICA processing unit 60 includes an independent separating unit 61 and an R (orthogonal) matrix calculating unit 62. The R matrix calculator 62 determines the orthogonal matrix R so that the elements of the received signal vector x ′ are independent of each other. The independent separation unit 61 separates signals by performing an operation of y = Rx ′. The separated signal y (t) obtained by the ICA processing unit 60 is sent to a rearrangement 'level adjusting unit 70 as a post-processing unit. By the way, each of the above parameters can be expressed as x (t) = A 's (t) and y (t) = Wx (t) = RS A' x (t) (However, this is a description when the Centering process is omitted).

 Rearrangement The level adjustment unit 70 rearranges and adjusts the level of the separated signal y (t). As a result, a signal corresponding to the transmitted M-sequence signal can be obtained from the ICA-processed signal with an unstable separation order and level. In other words, since the signal separation rule other than independence is not used in the ICA processing, the separation order and level become indefinite, and the rearrangement level adjustment unit 70 performs post-processing. This rearrangement 'level adjustment unit 70 can be realized by a conventional channel estimation process using V for wireless transmission.

 [0045] By the way, ICA has proposed a plurality of algorithms with various viewpoints. For example, it is known that Sphering using PCA cannot be well-correlated when receiver noise cannot be ignored, and an algorithm that does not require Sphering has been proposed! In the following, how to use a suitable ICA algorithm will be described specifically for reference when implementing the present invention. In the following, we will focus on ICA using the gradient method, which is relatively easy to understand.

 [0046] (1-3-2) Basic concept (conditions and reasons for separation based on independence only)

 First, the following conditions are necessary to perform ICA by the gradient method (however, ICA including Sphering is used here).

 [0047] Condition 1: The average value of each element of the source signal vector s (t) is 0.

 E [s] = 0, k = l, 2, ..., M

 k

 Condition 2: Each element of the source signal vector s (t) is independent of each other.

p (s = p (s) ρ (s) p) Where p is the probability density distribution function of each signal and is unknown but k for Gaussian distribution

 Make it not exist.

 Condition 3: Each element of source signal vector s (t) is mixed by time-invariant linear combination to become M received signals X (t)

 k

 x (t) = A-s (t)

Under the above conditions 1 to 3, the matrix W (= ^A_1 ) is estimated so that each element of the y (= Wx) vector becomes independent.

[0048] Condition 1 does not lose generality if direct current removal processing is performed in advance. Condition 2 is equivalent to, for example, E [s] = E [s, s,..., S] = E [s] -E [s] ---- E [s]. Condition 3 is

 1 2 M 1 2 M

 It shows that neither the multipath nor the receiver noise is initially considered. Therefore, E [y] = E [x] = E [k k s] = 0. Note that Ε [·] means the set average.

 k

 [0049] Next, the basic idea of the gradient method ICA is based on four forms ((i) the most intuitive form, (ii) a form arranged from the viewpoint of the maximum likelihood estimation method, and (iii) Sphering) (Iv) a format that is more generalized from the viewpoint of information geometry) and explain it in relation to each other. Since the gradient method is easy to obtain a relatively intuitive image, this association is possible.

 [0050] (i) The most intuitive format

 The estimated value W is preferably corrected sequentially as in the following equation.

 [Equation 6]

W (t + 1) -W (t) = Δ W (t), Wij (t) = -ν-[yi (t)]-φ [ _yj (t)]] where η is a small positive number, Φ, is an arbitrary measurable function (6)

 [0051] Equation (6) means that the (i, j) element of the estimated value W is corrected by mapping with respect to y (t) and y (t). Since y = Wx, W is essentially a transfer coefficient to the X force y, and should be used to correct it. However, here, instead of X, y containing these information is used.

[0052] First, when equation (6) is added for t, the following equation is obtained. [Equation 7]

Wij (t) -W (0) = Δ Wij (t) =-η .∑ Φ [yi (t)]-φ [ _yj (t) l

 t t (7)

[0053] At this time, if y (t) and y (t) become independent from each other, the following equation can be obtained by substituting the time-average ^^-average with the assumption of the ergodic nature of the signal: it can.

 [Equation 8]

Wij (t) = Wij (0)-v "Ε [[ _yi (t)]] -E [φ [yj (t)]]… constant

[0054] The collective average is exactly a force that is a function of time.

Equation (8) means that W (t) converges to a constant value.

[0055] Thus, if AW is expressed as a function of y and y in equation (6), it can be easily shown that W (t) converges if the elements of vector y are independent. Since each element of the source signal vector s is assumed to be independent from each other, the vector y should match the source signal vector s after convergence (as shown in the example of Fig. 3). The elemental power of y obtained by multiplying an independent s by a matrix is not necessarily independent of each other, but on the contrary, it can be separated if it is independent.

[0056] However, in the above discussion, it is V ヽ that touches specifically what the functions φ and φ should be. In other words, Erika (8) suggests that the convergence value differs depending on how these functions are selected. In fact, because AW is a function of y and y (including the entire received vector X, not X), the components of the ICA separation signal, vector y, are unordered. In addition, independence (in other words, waveform) is used for convergence of Eq. (8), but the amplitude is irrelevant, so the scale of the separated signal is also undefined. In other words, the ICA operates only for the purpose of making the separated signals independent on the assumption that the elements of the source signal vector s are independent of each other, so the order and scale of the separated signals become indefinite (that is, the waveform Are maintained, but the order and size are indeterminate). Since ICA has these uncertainties, the convergence value naturally changes depending on how the functions φ and φ are selected. Of course, the convergence speed has also changed. Tetsushimatsu 6

 [0057] Incidentally, when ICA is performed, the order and scale of the separated signals become indefinite.

= Rather than converge to A ^_1, can be expressed as converges to W = PDA ^{_ 1.} Here, matrix D is an appropriate diagonal matrix that transforms the scale, and matrix P is a sort that has one "1" in each row and column element.

 [0058] (ii) Format arranged from the viewpoint of maximum likelihood estimation

 The ICA algorithm has been studied from the viewpoint of maximum likelihood estimation and information theory. At present, the following formula is often used.

 [Equation 9]

 W (t + 1) -W (t) = Δ W (t), Δ W (t) =-v -Fix I W (t))-W (t)

 Where F (x I W (t) is called the estimation function,

 Any measurable function Φ (y) is expressed as

F (x IW (t)) = I one φ (y (t)) y (t) ^T , E [φ (y) y ^T ] = I (9)

 [0059] Equation (9) is derived by returning to the original meaning of the gradient operation, and is particularly called a natural gradient method. Assuming that the probability density function p (s) of the source signal vector is already known, the estimation function F (x I W) can be solved as follows by applying the maximum likelihood estimation algorithm.

[Equation 10]

Maximum likelihood estimation solution: F (x IW (t)) = I-φ (γ) γ ^τ , φ (γ) =

 dy (10)

 In other words, equation (10) is the solution that is originally desired. However, in ICA, which is a blind estimation, ρ (s) is unknown, so Eq. (10) replaces this part with a convenient estimation function that does not include p (s). In fact, in Eq. (10), only the variable name is changed to p (y), and an arbitrary measurable function φ (y) is introduced instead.

[0061] The estimation function sets AW (t) in equation (9) to 0 only when the estimated value W (t) matches the true value ^A_1. So that the update stops and _c

 [Equation 11]

 Conditions that the estimation function F (x I W (t)) must satisfy:

When W (t) = A- 1

W (t) A—when ¹

(1 1)

 [0062] Furthermore, E [] means the estimated value W (t) at each time and the probability density function of the source signal vector.

 W, p (S)

 Since the set average for p (s) is !!, equation (11) is also a blind estimation condition that does not depend on p (s).

 [0063] Here, regarding the estimation function of equation (9), when considering the ij component of the matrix φ (y) yT, if i ≠ j, it becomes 0 by independence, and if i = j, it is a nonzero number Therefore, each element function of φ (y) can be scaled so that

 [Equation 12]

E Φ (γ) γ ^Ύ (12) Therefore, the estimation function of Eq. (9) satisfies the condition of Eq. (11)! /.

 [0064] On the other hand, when the estimation function F (x I W) is also given to the force of equation (11), the convergence value is E [F (

 W, p (S) x I w)] = o It can be obtained as the solution of the following equation by the law of force algebra that can be obtained as a solution. This is called an estimation equation.

 [Equation 13]

 T J F 0¾ I W (t) = 0 T is sufficiently large t = l

 (13)

 [0065] (iii) Format based on Sphering

The estimation function of ICA with Sphering as shown in Fig. 4 is as follows. W (t + 1) -W (t) = Δ W (t), Δ W (t) =-η ■ F (x IW (t)). W (t) where F (x | W (t )) Is called the estimation function,

 Any measurable function φ (y) can be expressed as

 F (x I W (t))

=-Φ (y (t)) y (t) T + y (t) φ (y (t)) T, E [φ (y) y ^T ] —E [y φ (y) T] = 0

(14)

[0066] In ICA in this case, W (t) must be an orthogonal matrix, so WW ^T = I, and the estimation function is ε Δ and W is updated to (Ι + ε Δ) W. Even if it is done, this must be satisfied. Then, the following equation is obtained. If the term of ε ² is ignored, it must be ΔΤ + Δ = 0! /

 [Equation 15]

(1+ ε A) W- [(1+ ε A) W] ^T = 1+ ε Δτ + s Δ +2 ² Δ Δ ^τ = 1 (15)

 [0067] In other words, the estimation function Δ (ε is only for matching the scale and can be omitted because it is not important for ICA), the diagonal component is 0, and the ij component and the ji component have a different sign Must be in format. In order to satisfy this condition, the estimated function as shown in Eq. (11) and the difference between its transpositions should be used.

 [Equation 16]

Δ = (l- (y) yT)-(ΐ- ( _γ ) _γ τ) τ =-φ (y) yT + y φ (y) T

 (16)

 [0068] (iv) A form in which the viewpoint of information geometry is more general

 There are an infinite number of candidates for the estimation function as described above. From the theory of blueprint geometry, it is shown that if you select an estimated function force of the form Other types of estimation functions generally have large estimation errors.

[Equation 17] W (t + 1) -W (t) = Δ W (t), Δ W (t) =-ηF (x IW (t)) W (t) where F (x IW (t)) Is called the estimation function,

 Any measurable function Φ (y) is expressed as

F (x IW (t)) = Ι- α φ (y (t)) y (t) T + _y ( _t ) φ ( _y (t)) T

 (17)

 [0069] Note that the estimation function F (x IW (t)) multiplied by the regular matrix R is also an estimation function (R is more accurately a reversible mapping from matrix to matrix and depends on W. It is better to write R (W).

 [0070] (1 3— 3) Algorithm types

 As described above, the present invention is characterized in that blind signal separation that does not require a known signal for signal separation is realized by using an independent component analysis algorithm. Figure 6 shows a typical independent component analysis algorithm proposed so far. In the present invention, the conventional independent component analysis algorithm in FIG. 6 can be selected from among these independent component algorithms according to the multi-antenna radio system to be used. It may be modified and used to adapt to the above.

 (2) Embodiment 1

 FIG. 7 is a diagram for explaining the configuration and operation of the first embodiment of the present invention. FIG. 7A shows the configuration of transmitter 100 of this embodiment, FIG. 7B shows the frame configuration of a signal transmitted from each antenna TAN-1 to TAN—m of transmitter 100, and FIG. 7C shows this configuration. A configuration of a receiver 200 of the form is shown.

[0072] As shown in FIG. 7A, the transmitter 100 includes a MIMO transmission apparatus 110, a transmission radio circuit 111-1 ~: L 11 m, and a plurality of antennas Dingpachi? And have. MIMO transmission apparatus 110 performs baseband processing on signals transmitted from antennas TAN-l to TAN-m. For example, the signals transmitted from each antenna TAN-l to TAN-m are subjected to error correction coding and modulation (mapping), and the signal is distributed to each antenna TAN-l to TAN-m. Each of the transmission radio circuits 111 1 to 111 m converts the baseband signal input from the MIMO transmission apparatus 110 into a radio signal. As shown in FIG. 7C, receiver 200 includes a plurality of antennas RAN-1 to RAN-n, reception radio circuits 201-1 to 201-n, and a MIMO receiver 210. Each reception radio circuit 201-1 to 201-n converts the radio signal received by each antenna RAN-1 to RAN-n into a baseband signal, and separates the converted baseband signal from the MIMO receiver 210. Send to part 211.

 [0074] The signal separation unit 211 obtains separated signals y to y by performing signal separation processing using the independent component analysis method as described in the above section (1). In the present embodiment, a case will be described in which the signal processing unit 211 performs signal separation using, for example, the Jutten and Herault algorithm among the independent component analysis algorithms shown in FIG. In this case, the signal separation unit 211 obtains the separation signals y to v by executing the following equation. The following equation is an algorithm when m = n.

 [Equation 18]

W (t + 1) = W (t) + AW (t), A Wij (t) = -η ■ yi (t) 3. Yj (t) where, W (t) is transmitted at time t coefficient matrix Estimated value of

 Δ W (t) is the W (t) correction matrix at time t

 Δ Wij (t) is the ϋ element of W (t) at time t

 yi (t) is the ίth separation signal

 η is an appropriate real number, i = 0, l, ...! II, j = 0, l, ~, m (18)

 When equation (18) is executed, the estimated value W (t) approaches the actual transfer coefficient matrix, and the received signal can be correctly separated using the inverse matrix. Incidentally, in mobile communication, the conditions required for the algorithm of equation (18) are:

 1. Transmit signals are independent of each other

2. While W (t) is concentrating on convergence, the transfer coefficient matrix is only constant, and neither known symbols nor synchronization are required. As a result, in this embodiment, neither a known symbol for signal separation nor a synchronization process for signal separation is required. The separated signals V to y are sent to the demapping units 212-1 to 212-11 and the frame synchronization units 214-1 to 214-11. [0076] The frame synchronization units 214-1 to 214-n detect the frame synchronization timing using the separated signals y to y, and the synchronization timings are determined as demapping units 212-l to 212-n and error correction decoding units. 213—1 to 213—n are notified. As described above, in the present embodiment, frame synchronization is detected for each of the separated signals y to y which does not detect frame synchronization using a signal before signal separation. As a result, it is possible to detect frame synchronization by using signals whose mutual interference has been reduced by signal separation.

 [0077] Thus, in this embodiment, if independent component analysis is used, the feature that separation processing can be performed without frame synchronization before signal separation is effectively used, and signal separation is performed. Frame synchronization is made.

 The separated signals y to y are demapped by the demapping units 212-1 to 212-n and then error-corrected and decoded by the error correction decoding units 213-1 to 213-n. The data after error correction decoding is sent to the order correction unit 215.

 [0079] The order correction unit 215 observes the data input from each of the error correction decoding units 213-l to 213-n, and rearranges the data according to the observation result. For example, the data is rearranged by observing information such as IP addresses contained in the data. This makes it possible to return the order of signals that have become indefinite due to the independent signal analysis algorithm to the correct order.

[0080] Next, the operation of the MIMO communication system of the present embodiment will be described.

Transmitter 100 transmits a signal having a frame configuration as shown in FIG. 7B. In Fig. 7B, the horizontal axis is the time direction. FIG. 7B shows an example in which signals Ul to Uk for k channels are time-division multiplexed and transmitted in one frame. That is, in the example of FIG. 7B, channel 1 signal U1 is transmitted from m antennas TAN-1 to TAN-m at the same time, and channel 2 signal U2 is transmitted from m antennas TAN-1 at the same time. ~ TAN—sent from m

,,, Channel 111 signal 111 ^ is transmitted from 111 antennas 1-8 to 1-8 at the same time. Thus, for example, if different signals are transmitted from antennas TAN-1 to TAN-m as signals Ul to Uk of channels l to k, signals for m X k channels can be transmitted within one frame period in the same time. It becomes.

Here, in transmitter 100 of the present embodiment, known symbols for signal separation are used. Do not send. Thus, as is apparent from comparison with FIG. IB showing a conventional transmission frame, there is no known symbol transmission period in one frame, so that one frame can be used as a data transmission period. As a result, the amount of transmission data can be increased by the amount of not transmitting a known symbol for signal separation as compared with the conventional case.

[0083] When there is no known system for signal separation, the receiver 200 that receives a signal executes an independent component analysis algorithm by the signal separation unit 211 of the MIMO receiver 210, thereby The signals mixed in are separated into independent signals y to y.

 Here, generally, when independent component analysis is performed, the order of the separated signals becomes indefinite. In consideration of this, the MIMO receiving apparatus 210 of the present embodiment is provided with an order correcting unit 215, and the order correcting unit 215 corrects the order of the separated signals V to y.

 [0085] Forcibly, according to the present embodiment, by providing the signal separation unit 211 that performs signal separation using the independent component analysis method, even if there is no known symbol for signal separation, the transmission line can be used. Since the mixed signals can be separated, it is possible to prevent the transmission rate from being lowered due to the overhead due to the known symbols for signal separation.

 As a result, even if the number of antennas is increased so that a large number of users can be accommodated, it is possible to prevent a decrease in transmission speed. Even when applied to the uplink, it is possible to eliminate the need for highly accurate synchronization between signals. In other words, in the uplink where each transmitting antenna has a different terminal (ie, m terminals), each transmitted antenna receives a signal transmitted by one receiver. In a system that separates signals from terminals), it is essential for each terminal to synchronize the frame, so a feedback mechanism between transmission and reception such as time alignment is required. Since it is not necessary to do this, the system can be simplified and the transmission speed can be increased.

 Further, according to the present embodiment, by providing the order correcting unit 215, the order of signals that have become indefinite by performing the independent signal analysis algorithm can be returned to the correct order. Become.

[0088] (3) Embodiment 2

FIG. 8, which shows parts corresponding to those in FIG. 7 with the same reference numerals, shows the configuration and operation of the second embodiment. It is a figure where it uses for description. Fig. 8A shows the configuration of transmitter 300 according to the present embodiment, Fig. 8B shows the frame configuration of each antenna TAN-1 to TAN-m force of transmitter 300, and Fig. 8C shows the configuration of this embodiment. 1 shows a configuration of a receiver 400 of the embodiment.

 Transmitter 300 of the present embodiment shown in FIG. 8A arranges known symbols for identifying transmission antennas TAN-l to TAN-m as shown in FIG. 8B. Specifically, different known symbols are transmitted from each of the transmission antennas TAN-1 to TAN-m. For example, transmit known symbols with different symbol patterns between antennas. This known symbol is formed by the MIMO transmitter 310.

 [0090] The receiver 400 of the present embodiment shown in FIG. 8C is different from the receiver of Embodiment 1 in that it replaces the frame synchronization units 214-1 to 214-n with a frame synchronization 'signal identification unit. 411—1 to 41 1—n. The frame synchronization 'signal identification units 411-1 to 411-n detect the frame synchronization timing using the separated signals y to y, and use them to detect the demapping units 212-1 to 212-n and the error correction decoding unit 213— 1 to 213—Notify n. Power! The frame synchronization 'signal identification units 411 1 to 411 n transmit the transmission signals TAN—l to TAN—m from which the separation signals V to y are transmitted based on the known symbols included in the separation signals y to y. The identification information is sent to the order correction unit 412.

 [0091] Based on the identification information from the frame synchronization 'signal identification units 411-l to 411-n! /, The order correction unit 412 receives the data input from the error correction decoding units 213-1 to 213-n. Rearrange in order. As a result, even if the order of the separated signals becomes indefinite by the independent component analysis method, data in the correct order can be finally obtained.

 As described above, according to the present embodiment, a known symbol for identifying a transmission antenna is arranged in a signal to be transmitted from each of transmission antennas TAN-1 to TAN—m, and independent components are arranged. Compared with the first embodiment, the order of the signals separated using the analysis method is rearranged in the correct order based on the known symbols, so that any transmission antenna can transmit the separated signal. Therefore, it is possible to quickly rearrange the signals after separation by the independent component analysis method.

That is, as in the first embodiment, in the method of observing data such as the IP address included in the received data, it is generally not possible to perform order control so that the process does not proceed to the upper layer. However, in the present embodiment, the order of data can be identified by physical layer processing. As a result, for example, when applied to packet transmission involving retransmission, there is an effect of preventing a decrease in throughput.

 [0094] In addition, since the known symbol for transmitting antenna identification in this embodiment is not used for signal separation, the transmission time should not be overlapped as in the conventional known symbol for signal separation! / Because there is no need to pay attention, that is, there is no need to transmit in time division, so even if the number of transmitting antennas increases, the transmission speed does not decrease.

 [0095] Some independent component analysis algorithms have nonholonomic properties, and when the number of transmitting antennas m is smaller than the number of receiving antennas n, n separated signals y to y (n -M) can be reduced to 0 (in the case of principal component analysis, n-m signals that should be 0 also appear). Therefore, it is more preferable to identify no signal by the frame synchronization / signal identification unit 412.

 [0096] (4) Embodiment 3

 FIG. 9 which shows the parts corresponding to those in FIG. 8 with the same reference numerals is a diagram for explaining the configuration and operation of the third embodiment. Fig. 9A shows the configuration of transmitter 300 of this embodiment, Fig. 9B shows the frame configuration of signals transmitted from each antenna TAN-1 to TAN-m of transmitter 300, and Fig. 9C shows the configuration of this embodiment. The structure of the receiver 500 of a form is shown.

 This embodiment is different from Embodiment 2 in that level / phase correction sections 511-1 to 511-n are provided in MIMO receiving apparatus 510 of receiver 500! Level / Positive Compensation Positive parts 511-1 to 511-11 correct the level and phase of the separated signals y to y input from the signal separating part 211, and the corrected signals are de-mapped parts 212-l to Send to 212-n. At this time, the level 'phase correction units 511-1 to 511-n are based on the reception amplitude and reception phase of the known symbols included in the separated signals y to y and determine the distortion added in the transmission path. Estimate and correct the distortion.

This distortion estimation / correction processing is the same as what is generally called channel estimation * correction. In practice, such distortion estimation / correction processing cannot be applied when other interference waves are superimposed on a known symbol, but in this embodiment, the signal separation unit 211 performs SN (signal pairing). Distortion estimation and correction processing is performed on the separated signals y to y after the improvement of the (noise ratio). Therefore, it is possible to perform correction processing with high accuracy.

 [0099] In summary, according to the present embodiment, in addition to the configuration of the second embodiment, a level for correcting distortion of separated signals y to y based on known symbols in separated signals y to y ' By providing the phase correction units 511-l to 511-n, in addition to the effects of the second embodiment, high-precision distortion correction is possible, and it can be applied to forging transmission lines and multilevel modulation systems. Become.

 [0100] This specification is based on Japanese Patent Application No. 2004-379671 filed on Dec. 28, 2004. All the contents are included here.

 Industrial applicability

[0101] The MIMO receiver, MIMO communication system, and MIMO reception method of the present invention can perform signal separation even when a known signal for signal separation is not arranged, and can be applied to a wireless system such as a wireless LAN or a cellular system. It is suitable for wide application.

Claims

The scope of the claims  [1] By inputting a plurality of received signals received by a plurality of antennas and subjecting the received signals to blind signal separation processing using an independent component analysis method, a plurality of independent separated signals are obtained. Signal separation means;  A MIMO receiver comprising: demodulating means for demodulating the plurality of separated signals obtained by the signal separating means. [2] The apparatus further comprises order correction means for rearranging the order of the plurality of separation signals obtained by the signal separation means.  The MIMO receiver according to claim 1. [3] The order correcting means is included in the separated signal! /, Specifies a transmitting antenna, and rearranges the order of the plurality of separated signals based on a known symbol.  The MIMO receiver according to claim 2. [4] The image processing apparatus further includes distortion correction means for correcting distortion of the plurality of separated signals based on known symbols transmitted from the transmitting antennas included in the separated signals. The MIMO receiver according to claim 1, wherein: [5] A transmitter that transmits transmission signals including different known symbols from multiple antennas, and multiple reception signals received by multiple antennas, and blinds that use independent component analysis for the multiple reception signals Signal separation means for obtaining a plurality of independent separated signals by performing signal separation processing, and the plurality of separated signals based on the known symbols included in the separated signals! Order correction means for rearranging the order of  A MIMO communication system comprising: [6] A signal separation step for separating received signals received by multiple antennas into independent signals using an independent component analysis method;  An order correction step for rearranging the order of the separated signals;  MIMO reception method including: