Adaptive compression of feedback channel based on second order channel statistics. RU patent 2478258.

FIELD: radio engineering, communication.

SUBSTANCE: method involves determining individual statistics for a plurality of channel tap coefficients of a communication channel between a transmitting station and a receiving terminal; and individually quantising said plurality of channel tap coefficients at corresponding quantisation bit rates which are determined based on said statistics to generate quantised channel tap coefficients, wherein the total number of bits allocated to said plurality of channel tap coefficients is fixed; and transmitting said quantised channel tap coefficients from the receiving terminal to the transmitting station.

EFFECT: implementing methods of compressing channels status feedback, which are adapted for different channel tap distributions.

52 cl, 11 dwg

Prior art

The present invention relates generally to transfer feedback channel status in the mobile network, and, more specifically, to a method and device for compression feedback channel status adaptive way.

The use of multiple antennas at the transmitter and/or receiver systems wireless attracted considerable attention over the past decade due to possible improvements as the radio coverage and speed of data transmission. Unlike systems with one antenna, where information channel status not significantly improves throughput, a substantial increase in throughput can be achieved in systems with multiple antennas where precise information channel status is available transmitter. In a system based on a multiplexing frequency division multiplexing (FDD), the receiver normally displays the channel information feedback state of the channel on the transmitter. Despite the fact that the assumption about the perfection of state information channel on the transmitter unrealistic due to bandwidth limitations imposed on the feedback channel, and associated delays caused by finding the signal back and forth, it was shown that even partial knowledge about the channels on the transmitter can achieve a significant increase in throughput compared with systems that do not take into account the information of the state of the channel. However, the feedback on the detailed information link state spends useful bandwidth of the back line. Consequently, there is considerable interest in designing effective ways to reduce feedback information channel status without significant expenditure of bandwidth back line.

One approach to feedback channel status uses unstructured block or vector (VQ) to reduce feedback on the information of the state of the channel. Although, in theory, unstructured VQ can achieve optimum achievable compression, the complexity of unstructured VQ grows exponentially with the piece size to speed. For example, in a MIMO system with 4 transferring and 2 reception antenna size unstructured VQ, proposed in the literature, can obtain the value 4*2*2 (real and imaginary parts of each coefficient of the report of the channel)=16. Resource requirements for storage and computing resources applicable to large unstructured VQ, may be prohibitively high in practical application for permission quantization (or initial speeds of coding), which is achieved acceptable accuracy.

Separately from computational complexity, another problem unstructured VQ is their inability to adapt to a different channel statistics. The majority of the proposed technologies quantization compression feedback channel status suggest that the reports of the MIMO channel are independent and identically distributed (IID) the spatial dimension. In practice, however, the statistical distribution channels MIMO often highly spatial and frequency. VQ, designed on the basis of IID-assumptions may not provide the desired performance to a wide range of channel statistics that is usually obtained in wireless environments.

On the other hand, the design of the flat VQ to consider all of the possible distribution of the report of the channel, at the same time maintaining reasonable accuracy quantization is not practical.

Accordingly, there is a need for methods of compression feedback channel status, which can be adapted to different distributions of the report of the channel, at the same time maintaining acceptable accuracy and complexity.

The essence of the invention

The present invention relates to a method and apparatus for providing in the form of feedback information about the channel, using adaptive vector . Method and device for use channel statistics of the second order (for example, dispersion) for compression feedback instantaneous characteristic of spatially-correlated MIMO channel. Many low-dimensional vector (VQ) with various resolutions (or speeds) various coefficients channel reports. Resolution of each VQ adaptively selected based on dispersions of the corresponding report of the channel. When you use different permissions quantization for the report of the channel with varying significance, the distribution of points of quantization can be done in a similar distribution corresponding to the optimal VQ designed for a specific channel statistics, which leads to almost optimal performance from a much lower complexity, computationally and storage.

In one typical variant of the incarnation as a compressed feedback instant channel characterization, and channel statistics serves as feedback to the transmitter. Compressed feedback instant channel characterization is provided as a back quick feedback channel. Channel statistics served as feedback to a transmitter on a slow feedback channel through which information is sent from the receiver back significantly less often than the quick feedback channel. In an alternative embodiment, it's useful when the noise spectrum is relatively flat in the frequency spectrum, all or part of the required channel statistics can be calculated directly on the transmitter, based on the assumption that statistics forward and reverse channels are .

In some versions of the incarnation, feedback channel can be converted to another area before the quantization of a channel assessments. For example, in the variant of the incarnation, which is suitable for systems MIMO-OFDM, channel characteristics measured in the frequency domain can be converted in the channel reports the time domain. Channel estimation temporary area, which fall under a predefined delay variation, selected, and then further transformed by the spatial dimension of the «own» region. The resulting transformed rates individually using with different speeds (or permissions), adaptive calculated in accordance with the variance of the transformed coefficients.

Feedback channel status decoded transmitter, using the code books quantization for the relevant speeds (or permission) to obtain estimates of the transformed coefficients, i.e. quantized transformed coefficients. Rate or resolution of each quantizer is calculated the same way as in the receiver, based on the relative variance of the corresponding converted factor. Subsequently inverse transformations are applied to the quantized converted odds to get version of the channel characteristics frequency domain. Based on this information about the channel can be calculated in the transmitter optimal corrector preliminary coding attributable to each thread encoding speed and/or quality indicator channel (CQI) for each frequency.

Brief description of drawings

Figure 1 - illustration of a typical communication system.

Figure 2 illustrates a typical communication system that uses a scheme of adaptive feedback.

Figure 3 illustrates a typical communication system that uses a scheme of adaptive feedback.

Figure 4 illustrates a typical coding method feedback on the quality of the channel in accordance with one variant of the incarnation.

Figure 5 illustrates a typical way to decode feedback on the quality of the channel in accordance with one typical variant of the incarnation.

Figure 6 illustrates a typical encoder feedback for OFDM system.

Fig.7 - illustration of a typical decoder feedback for OFDM system.

Fig.8 illustrates a typical processor conversion for the encoder feedback OFDM shown in Fig.6.

Figure 9 illustrates a typical processor conversion decoder feedback OFDM shown in Fig.7.

Figure 10 illustrates performance for the MIMO system according to the present invention.

Figure 11 is an illustration of the performance of the scheme of adaptive feedback, shown in Figure 6. and 7.

Detailed description

Referring now to the drawings, the typical variants of embodiment of the present invention described in the context of the communication system 10 with multiple antennas shown in figure 1. Communication system 10 with a variety of antennas may, for example, contain a system with multiple inputs and a single output (MISO), or a system with multiple inputs and multiple outputs (MIMO). Specialists in this field of technology however, we should understand that the principles illustrated pivot options incarnation, could be applied to other types of communications systems.

Communication system 10 with multiple antennas contains the first station 12, transmits the signal via the standard communication channel 14 at the second station 16. The first station 12 is mentioned here as a transmitting station, while the second station 16 is mentioned here as a receiving station. Specialists in a given field of technology will take into account that each of the first station of 12 and the second station on 16 may include the transmitter and receiver for bidirectional communication. Communication line from the transmitting station 12 to the receiving station 16 is called the descending line. Communication line from the receiving station 16 to the transmitting station 12 is called the ascending line of communication. In one typical variant of the incarnation, the transmitting station 12 is the base station in your wireless network, and reception places. 16 is the mobile station. The present invention can be used, for example, to send data from the base station 12 to the mobile station 16 through the channel high-Speed Packet Data downlink (HSPDA) in WCDMA systems.

The transmitting station 12 transmits signals from a variety of antennas to the receiving station of 16, which may include one or more receiving antennas. Unlike systems with one antenna, which involve a single antenna, both the transmitting and receiving stations 12 and 16, increasing system throughput can be implemented, if the broadcaster has 12 for detailed information about the channel characterization for channel 14 from the transmitting station 12 to the receiving station 16. Receiving station 16 calculates channel estimation 14 from the transmitting station 12 to the receiving station 16 and transmits feedback channel status to the transmitting station 12 through the feedback channel 18. However, the provision of feedback information about the channel from the receiving station 16 to the transmitting station 12 consumes useful bandwidth of the back line, which could otherwise be used to migrate user data. In systems with multiple antennas amount of feedback channel status increases sharply with the number of pairs of transmitting and receiving antennas.

Figure 2 explains the typical transmitter 100 transmitting station 12 and receiver 200 receiving station 16. Receiver 200 uses technology vector quantization, to reduce feedback channel status. For clarity, it is assumed that the communications system 10 uses a variety of antennas for transmitting station 12 and single antenna to the receiving station 16. The principles described here apply in full and on lots of aerials of the receiving station 16.

The transmitting station 12 (for example, a base station) sends signals

generated processor 102 the transmitted signal to the receiving station of 16 (for example, a mobile station). There are M downlink channels (one from each of the transmitting antenna). Channels downlink from the transmitting station 12 to the receiving station 16 are assumed to be linearly invariant time-TV with canal characteristic g m (t) in the time domain and G m (f) in the frequency domain. signal r(t), adopted at the receiving station has the form:

Equation 1

where * denotes the convolution, and v(t) - the noise of the main frequency bands. m-th channel downlink can be modeled as:

Equation 2

where a m,k - channel coefficients channel from m-th antenna and t to the delay. Tool 204 channel estimation in the receiving station 16 generates a score channel downlink in accordance with:

Equation 3

where m=1,...,M, and T is the sampling period used for quantization delays τ to . Note that for Q in equation 3 is not necessary to be equal To equation 2. Channel estimation feature processor 202 accepted signals for demodulation adopted signal r(t). Additionally, channel estimation is entered in the coder 206 feedback. Encoder feedback 206 takes channel assessment of the 204 channel estimation, channel coefficients in and takes the quantized channel factors as feedback to the transmitting station 12.

Touted channel characteristics for channel downlink from one of the transmitting antenna can be conceptually linked to the discrete-time filter, finite impulse response Q with nonzero coefcients of the report, for example:

Equation 4

where m=1,...,M. Hence, the problem of transfer to the transmitting station 12 equivalent to the problem of transfer .

An implementation option shown in figure 2, uses the technology of adaptive quantization, which assigns a higher bit more meaningful reports channel and fewer and less significant channel reports. Distribution of bits adaptive calculated on the basis of long-term statistics report of the channel, such as the relative power or variance reporting channel to a predefined measure distortion reaction resulting quantized channel characteristics has been minimized to the total number of bits available. Use two logical feedback channel: channel 18a feedback low-speed (slow channel of feedback), as the feedback distribution of bits and more high-speed channel 18b feedback (quick feedback channel) transfer as feedback quantized coefficients of the report of the channel. In this variant the implementation of channel statistics (e.g., variance reports channel) gather in front quantization. Information regarding the number of bits allocated for the quantization of each report of a signal periodically sent back to the transmitting station 12 through the slow channel 18a feedback. Information concerning the quantized versions (according to the current distribution of bits) evaluation of each specific implementation channel periodically sent back through the fast channel 18b feedback.

Coder 206 feedback includes many multi-speed or variable-speed vector 212 tool 214 calculate the metrics and management tool (controller) 216 speed. Variable-speed vector 212 individually quantizing channel coefficients for each channel 14. Rate or resolution of each quantizer 212 chosen individually on the basis of relevant statistics the report of the channel. Tool 214 calculate the metrics calculates statistics such as the variance, each report channel of each channel 14 and delivers statistics on reports of channels on the controller 216 speed. In this variant of the incarnation of the band statistics calculated to quantization. Controller 216 speed specifies the number of bits allocated to each 212. The number of bits allocated 212, corresponds to the speed or resolving the quantizer 212. Quantized channel coefficients are sent to the transmitting station 12 quick channel 18b feedback. Allocation of bits specified by the controller 216 speed, served as a feedback to the transmitting station 12 to slow 18a feedback. In the alternative, the controller 216 speed may provide for a feedback channel statistics from the funds 214 calculate the metrics, and the allocation of bits can be calculated from the statistics on the transmitting station 12.

Decoder 104 feedback in the transmitting station 12 contains a variety of decoders 110 quantization and controller 112 speed. Decoders 110 quantization form the evaluation of quantum channel factors on the basis of received bits taken by the rapid channel 18b feedback. Decoding rate or resolution is determined by the controller 112 speed based on feedback on the distribution of bits from the controller 216 speed in the receiving station 16. In the alternative, the controller 216 speed in the receiving station 16 could provide feedback statistical metric of funds 214 calculate the metrics, and the controller 112 speed in transmitting station 12 could calculate the appropriate allocation of bits.

Figure 3 explains an implementation option that eliminates the slow channel 18a feedback. The same reference numbers are used in figure 2 to denote the same components. In embodiment, shown in figure 3, the transmitting station calculates the channel statistics channel ascending line of communication, which is the same as statistics for channel downlink, and determines the distribution of bits of channel statistics. In this case, the channel statistics collected after quantization so that one and the same statistics could be formed in a transmitting station, 12, and in the receiving station 16. The specialists in this field of technology should be clear that the metrics used to calculate the distributions of bits in the current period, speed control, will be used to determine the distribution of bits in the following period, speed control. Tool 214 calculate the metrics calculates statistics, such as variance) for each channel report on the basis of quantum channel coefficients. Variance or other statistical indicators are served on the controller 216 velocity, which determines the distribution of bits for variable-speed vector quantizer 212. Decoder 104 feedback in the transmitting station 12 takes a quantum channel coefficients. Tool 214 calculate the metrics uses quantum channel coefficients adopted in the current period to control the speed to calculate the distributions of bits for the next period of speed control. Distribution of bits that is computed in the previous period, speed control, used decoders quantization, to determine evaluation of quantum channel coefficients.

The options for implementing shown in figure 2 and 3, the distribution of bits for the Q channel reports can be calculated so that the root mean square difference between the estimated channel characterization, and its quantized version has been minimized, as described below. Let and denote the real and imaginary parts of the evaluated report of the channel, respectively, and let denote the k-th vector channel report. Let Q k (·) denotes the vector quantizer 212 dimension 2M N k points quantization used for quantization . The original encoding speed Q k (.) defined as that means the number of bits allocated for the quantization of each () element. The aim is to find the optimum of a vector distribution of bits R=(R 1 ,R 2 ,...,R (Q ) to minimize the amount of RMS distortion for all reports channel, defined as:

Equation 5

Distortion D(R k )report of the channel is:

Equation 6

The above-mentioned problem of optimization difficult for the exact solution, since the distortion D(R k ) is a highly nonlinear function R k . However, a good approximate solution can be displayed using the asymptotic formula Bennett-Enthusiasm- (Bennet-Zador-Gersho) to D(R k ), defined as:

Equation 7

To quantize the coefficients of the report of the channel at different speeds according to their dispersions, receiving station 16 and transmitting station 12 should keep respectively encoders 206 and decoders 104 many with different initial velocities encoding. Because of the speed, calculated using equation 9, may not exactly reflect the available speeds, can be performed certain operations rounding when calculating velocities {R, k }. To ensure that the final speed after rounding will not exceed the bandwidth 18 feedback, we can calculate the speed for reporting channel series, as

Equation 10

where k=1,2,...,Q denotes the approximation R j due to rounding. It can be noted that where =R j for all j=1,2,...,k-1, R k calculated by the equations of 9 and 10 will be the same. For a guarantee of good performance, preferably calculate the speed in descending order of the respective variances of the report of the channel and use operations rounding to for the dominant report of the channel is ensured sufficient number of bits.

Calculation of distributions of bits in accordance with equations 8 and 9 is one typical variant of carrying out the invention, which is based on the respect of the standard deviation of each report of the channel to the geometric mean of the standard deviation of all the reports of the channel. Other variants of the invention includes the calculation of distributions of bits on the basis of the arithmetic mean of a function dispersions channel reports in accordance with:

Equation 11

where k=1,2,...,Q denotes the set of monotone increasing functions. For example, when f k (x)=log(y k ,x)/2, equation 11 identical to equation 8. Alternatively, when f k ( x )= for all k, bit allocation is calculated on the basis of the relative values of the standard deviation of each report channel relative to the mean standard deviation.

More generally, if s denotes some long-term statistics with regards to the channel characteristics (for example, in a preferred embodiment calculation of distributions of bits for the various reports of the channel can be expressed as:

Equation 12

where k=1,2,...,Q and

k (·)to represent a certain function of the distribution of bits designed for k-th channel report. Equation 12 can be calculated consistently with:

Equation 13

where k=1,2,...,Q denotes the approximation R j due to rounding.

As mentioned above, for realization of the invention in which the transmitting station 12 and receiving station 16 must be implemented multiple encoders and decoders with different speeds and levels of distortion so that different levels of quantization can be provided according to the measured statistics. Alternatively, you can use a single vector quantizer with tree structure (TSVQ)to provide different levels of quantization. Encoder for TSVQ stores balanced tree encoding hyperplanes depth d-1, i.e. each node trees, indexed by a sequence of bits corresponds to normal () vector pb multidimensional hyperplanes and a threshold value . For example, the depth of the tree may be selected as d=2MQR. Having (estimated) report vector channel coding process begins with the root tree node with the appropriate hyperplane and calculates:

Equation 14

where q 1 (x) means one bit scalar quantizer, the output of which is equal to unity if x0, or zero if x0. At the next level coder 206 calculates:

Equation 15

using a hyperplane that corresponds to a value Encoder 206 repeats this process at the subsequent levels and calculates:

Equation 16

where b=(b[1],s[2],...,a[n-1]), unless and until it reaches the number of bits R k allocated for quantization . At this time coder 206 produces a sequence of bits R k (b[1],s[2],...,b[k ]) for the vector of the report of the channel.

Admission coded bit stream (b[1],s[2],..., b[k ]), decoder 104 TSVQ forms report channel for wood-based decoding depth b, whose nodes at each level contains quantized reports channel with an appropriate level of quantization. Hyperplane used at each level depends on the output bits, calculated at the previous levels. In addition, hyperplanes used in TSVQ (along with the corresponding tree decoding quantized vectors)are created to match the statistical distribution.

In practical communication channel reports are slow to change from one moment of time feedback to another. Thus, can be used differential quantization channel reports. In this case, the concepts described here can work in combination with any scheme of differential quantization for the quantization of changes in the channel reports from one point in time to another.

The principles of the present invention can be applied to the systems based on a multiplexing with orthogonal frequency division multiplexing (OFDM). In OFDM system adopted signal frequency domain can be modeled as:

Equation 17

where k=1,2,..., N, H, f [k] - matrix n R n T , indicating the channel characterization of MIMO, r[k] - received signal, s[k] - a transmitted signal and w[k] - component noise and interference on the frequency of the k-th subcarrier in the system wireless OFDM n T transmitting antennas and n R a reception antenna, respectively. Noise component W[k] suggests statistically independent frequency, but its covariance matrix, denoted by R

E { w [k] w [k] H }may vary with frequency, where E {•} denotes the expected value of the quantity inside the brackets.

Receiving station 16 evaluates the channel and the variance of the noise. Channel characteristics, corresponding transformation to white noise is defined as:

Equation 18

where k=1,2,..., N. We assume that certain statistics of the second order is available on the transmitting station 12. For example, channel statistics of the second order can be collected at the receiving station 16 by averaging on many implementations, observed over a period of time, and then sent to the transmitting station 12 to slow feedback 18a, as previously described. Alternatively, when the noise spectrum is relatively flat, at least the part of channel statistics can also be calculated directly at the transmitting station 12, using the property channel statistics on the forward and return channels 14.

Figure 4 illustrates a typical way of 50 executed by the encoder 206 feedback for channel coding assessments in accordance with one variant of the incarnation. Coder 206 feedback channel takes assessment of the 204 channel estimation and computes statistics, such as variance) for each channel coefficients (step-52). Controller 216 speed determines the speed for a related set of multi-speed 212 on the basis of channel statistics (step-54). Multi-speed 212 forth on an individual basis quantizing the corresponding channel coefficients at speeds of certain speed controller on the basis of statistics channel coefficients (step-56). In some cases, implementation, statistics, calculated before quantization for the current period, speed control, is used to define the initial speeds of coding. In other variants of implementation, statistics, calculated after quantization for the current period, speed control, is used to define the initial speeds of coding for the next period of speed control.

Figure 5 illustrates a typical way of 60 performed by the decoder 104 feedback to decode channel assessments, according to one typical variant of implementation. Controller 112 speed decoder 104 feedback determines the source encoding speed for a variety of decoders 110 quantization . Decoders 110 quantization next decode channel assessment, using the velocities determined on the basis of feedback on the distribution of bits from the speed controller step-64). In some cases, the implementation of the speed can be determined based on the feedback by the distributions of bits or channel statistics from the encoder 206 feedback (step-62). In other variants of implementation, the statistics are calculated for the current period, speed control based on feedback quantized estimates of the channel, can be used in a subsequent period, speed control, to determine the speed of decoders for 110 quantization.

Figure 6 illustrates a typical encoder 300 feedback for the receiving station 16 per OFDM system. Coder 300 feedback includes filter-Converter 302 to white noise, CPU 304 conversion module 306 zoom tool 308 calculate the metrics, 310 controller speed and variable-speed vector 312. Channel characteristics frequency region of funds 204 channel estimation and the covariance matrix of noise are introduced into the filter-Converter 302 to white noise. Filter-Converter 302 to white noise first performs a conversion operation to white noise by channel decorrelation of characteristics for each frequency on the corresponding square root of the covariance of the noise, according to equation 18 to form given to white noise channel description . The white noise channel feature then converted processor 304 conversion, as described in more detail below, in the vector coefficients

), where n

c denotes the number of converted channel coefficients. Module 306 zoom scales converted channel factors in X by their respective standard deviations. Scaled and converted channel coefficients next on an individual basis by the relevant variable-speed vector 312 (or with variable resolution). Vector 312 performed independently for different speeds (or permissions) on the basis of, for example, samples Gaussian IID with zero mean with unit variance. Vector , for example, may contain two-dimensional vector . In addition, vector 312 higher dimensions can also be used for quantization of two or more of the transformed coefficients together.

Speed (or resolution)used for quantization of each of the converted factor, adaptively selected on the basis of a set of dispersions channel coefficients frequency domain. Tool 308 calculate the metrics computes the variance of the converted channel coefficients. Controller 310 speed determines the allocation of bits for each vector quantizer 312-based dispersions channel coefficients. For example, having the aggregate reserves of B bits

total number of bits B

k , used for quantization factor X

k can be chosen in accordance with:

Equation 19

As shown in equation 19, the number of bits allocated to a specific price depends on how large its dispersion relative to the geometric mean of all dispersions. After quantization encoded bits are sent to the transmitting station 12 through a fast line 18b feedback.

Fig.7 illustrates decoder 400 feedback on the transmitting station 12 for OFDM system. Decoder 400 feedback draws operations applied by the encoder 300 feedback on the receiving station 16, for the formation of a quantized assessment given to white noise channel characteristics . Decoder 400 feedback includes many multi-speed decoders 402, 404 module scaling processor 406 inversion in the controller 408 speed. On the basis of the adopted bits decoders 402 quantization form of assessment converted channel coefficients. Controller 408 speed indicates the distribution of bits for each decoder 402, what determines the rate or resolution to this decoder 402. Distribution of bits used decoders 402 quantization, calculated controller 402 speed in the same way as in the receiving station 16, on the basis of the relative dispersion of the transformed coefficients, which in turn can be obtained from the statistical information provided by the reception station 16 to slow 18a feedback. Module 404 zoom scales of assessment converted channel coefficients their respective standard deviations. In conclusion, the processor 406 reverse conversion applies the reverse conversion to the scaled restored converted odds to get version of the white noise channel characteristics .

Many important values to maximize bandwidth and system can be obtained from the quantized to white noise channel characteristics . For example, the best concealer prior coding denoted by P[k], which maximizes throughput of the communication line on the k-th frequency can be calculated in accordance with:

Equation 20

where U H [k] denote the matrix columns which are native vectors of a matrix and D (p

[k],..., p

[k]) denotes a diagonal matrix with diagonal elements , which are:

Equation 21

where j=1,2,...,n T , - set of eigenvalues and 0 is selected that

. In addition, {p j [k]} can also be used as indicators of the quality of the channel (CQI) on different frequencies and various own mods, which is often a need for resource planning and adaptation of telephone lines.

The matrix U TR can either be provided as feedback to a transmitting station 12, using a slow network 18a feedback, or, alternatively, it can be estimated, using the measurements of the ascending line. This transformation corresponds to the complete Transformation of - (Karhunen-Loeve) (KLT) in respect of spatial channel reports. Except U TR, transmitting station 12 also requires a variation of each component of X[n], which can also be made available to the transmitting station 12 through the slow channel 18a feedback to calculate the appropriate distribution of the raw bits for a given stock of bits.

According to another model variant of the invention, a spatial transformation is performed according to all n W , where U T mean matrix c private vectors of the correlation matrix channels of transfer of size n T n R used instead of eigenvectors of f full . Matrix f TX correlation transmission channels is:

Equation 23

Note that the matrix f TX correlation transmission channels can be obtained from the full matrix f full correlation channels. In particular, item f th in the i-th row and j-th column followed by the corresponding size n T n R in the f full , for example

where [A] m:n,p, q denotes size (n-m+1) (q-p+1), taken from lines with m-th through n and from columns p-q-th matrix A.

The matrix U T can be either provided as feedback to a transmitting station 12, using a slow network 18a feedback, or alternatively, it can be estimated using the measurements of the reverse communication lines. In addition to U of T , the transmitting station 12 also requires the dispersion of each component X[n], which can also be made available to the transmitting station 12 through the slow channel 18a feedback to calculate the appropriate distribution of the raw bits for a given stock of bits.

In accordance with another embodiment, the spatial conversion is performed according to all n

W , where U R denotes the matrix with private vectors of the correlation matrix accessible channels of size n R n R , defined by:

Equation 24

Matrix f RX can be obtained from the f full summing, its diagonal size n R n R , for example:

Like U T , matrix U R can be either provided as feedback to a transmitting station 12, using a slow network 18a feedback, or alternatively, it can be estimated using the measurements of the reverse communication lines. In addition to U of T and U R , the transmitting station 12 also requires the dispersion of each component X[n], which can also be made available to the transmitting station 12 through the slow channel 18a feedback to determine the appropriate allocation of raw bits for a given stock of bits.

According to yet another variant of the invention, the spatial transformation is performed in accordance X[n]= , where W N denotes a transformation matrix IFFT, element of which in the i-th row and j-th column is set exp{-j2ij/N}. In the case of a transmitting station 12 also requires the dispersion of each component X[n], which can also be made available to the transmitting station 12 through the slow channel 18a feedback to determine the appropriate allocation of raw bits for a given stock of bits.

At the transmitting station 12, reverse transformation is applied to the restored converted odds to get restored to the white noise channel characteristic frequency domain, as shown in Figure 9. converted vector X is first divided into a set of quantized converted vector channel reports

. Reverse spatial transformation is applied then to each report X[n]to obtain the relevant given to white noise channel characterization of the time domain. The white noise channel feature then padded with zeros to form , which then is converted back into the frequency domain, through the operation of FFT to form given to white noise channel characteristic frequency domain.

Note that trimming operation and additions zeros shown in Figure 6 and 7 may be omitted for reporting channel with small dispersion, because for such a scenario, no raw bits, as a rule, will not stand out.

In this subsection we show the advantage in performance, provided by the invention through a system of MIMO-OFDM. The total bandwidth of the system, as expected, is 5 MHz FFT size equal to 512. The number of employed carriers is 300, which are evenly divided into sections 25 (12 carriers in each). The interval between sub carriers amounts to 15 kHz. Performance characteristics are modeled using a spatial model of channels with pedestrian In-channel profile in the micro environment.

Figure 10 shows the corresponding invention performance for four transmitting antennas and two receiving antennas. In particular, the graph is drawn, depending on the stock bits, the difference in the level of SNR required for the achievement of the ergodic capacity (for example, using 5 bits per channel) between the ideal case where the transmitting station 12 has perfect knowledge of the instantaneous state of the channel and the case where an instantaneous state of the channel is compressed using the invention, before providing feedback from the transmitting station 12. Stock bits normalized according to the number of partitions available in the system. As shown in Figure 10, the use of spatial correlation among the various elements of the channel matrix through various spatial transformations is very beneficial in terms of reducing the volume of quick feedback. For example, to achieve within 1 dB from ideal bandwidth closed-loop should be about 3.5 bits section (a total of 3.5 x 25=63 bits in the range), if applied uneven distribution of bits only according to various reports channel temporary area without spatial transform. However, if a fixed FFT-transformation is applied to each channel price as described in the previous section, less than 2 bits section (a total of 50 bits in the range)to achieve within 1 dB from ideal performance in a closed circuit. In addition, if, instead, one of the spatial transformations KLT described in the previous section must be less than 1 bit per partition (a total of 25 bits in the range)to achieve within 1 dB from ideal performance. If they are 2 bits section (a total of 50 bits) in the return line of communication can be achieved within 0.5 dB from ideal performance in a closed circuit.

11 additionally shows the operating characteristics of the circuit adaptive feedback, as shown in figure 4 and 5, assuming four transmitting antennas and one receiving antenna. In this case, if you use the uneven distribution of bits only according to various reports channel temporary area without spatial transformation, need about 2 bits per section (a total of 2 x 25=50 bits in the range)to achieve within 1 dB from ideal bandwidth in a closed circuit. However, if a fixed FFT-transformation is applied to each channel factor, as described in the previous section, about 1 bit per partition (a total of 25 bits in the range)to achieve within 1 dB from ideal performance in a closed circuit. In addition, if one of the spatial transformations KLT described in the previous section, about 0.4 bits section (a total of 10 bits in the range) to achieve within 1 dB from ideal performance. If you let the 1 bit per partition (a total of 25 bits) in the return line to achieve within 0.5 dB from ideal performance in a closed circuit.

Furthermore, the encoder 206 feedback can, in some cases, implementation, in addition, contain CPU 304 conversion to convert the mentioned channel coefficients with the aim of creating translated channel coefficients. Processor 304 transformation may, in some cases, implementation, convert channel coefficients in the frequency domain channel coefficients time area, select the mentioned coefficients temporary area within a predefined scatter delays and may, in some cases, implementation, further referred to convert selected channel coefficients temporary area in the channel coefficients own.

Metric calculator can, in some cases, implementation, define individual statistics for each channel coefficient, and individual statistics can, in some variants of implementation, include the relative channel power ratio. Speed controller may, in some cases, implementation, referred to determine a speed for the mentioned channel factors in descending order referred relative power. Individual statistics can, in some cases, implementation, contain the variance associated with the link factor. Controller 216, 310 speed may, in some cases, implementation, define referred to speed on the basis of statistics collected for the current period, speed control, and referred to the statistics are calculated before quantization of a channel of the coefficients in the current period, speed control.

Coder 206 feedback can, in some cases, implementation, transmit speed on slow feedback and transmits quantum channel coefficients on the rapid feedback channel.

Encoder feedback can, in some cases, implementation, in addition, contain module 306 scaling to scale the mentioned channel coefficients on the basis of the statistics referred to channel coefficients to quantization. Controller 216, 310 speed may, in some cases, implementation, determine the allocation of bits for the mentioned on the basis of the abovementioned statistics. Multi-speed 212, 312 may, in some cases, implementation, contain encoder vector with tree structure, designed on the basis of the abovementioned statistics.

Additionally, the decoder feedback to decode channel coefficients, quantized variable-speed quantizer, includes controller 408 speed to determine the appropriate speed for many channel coefficients and decoder 402 quantization for individual decoding mentioned many channel coefficients at speeds specified in the abovementioned controller 408 speed. Mentioned controller 408 speed may, in some cases, implementation, referred to determine the appropriate speed for a variety of channel factors by reception mentioned speeds from receiving station over slow feedback. Mentioned controller 408 speed may, in some cases, implementation, referred to determine the appropriate speed for the canal coefficients by receiving the statistics referred to channel coefficients from the receiving station for slow feedback and calculations referred to speeds on the basis of the abovementioned adopted statistics. Referred adopted statistics can, in some cases, implementation, contain dispersion of the mentioned channel coefficients. Mentioned controller 408 speed may, in some cases, implementation, referred to determine the appropriate speed for a variety of channel coefficients by measuring the statistics referred to channel ratios and calculation mentioned speeds on the basis of the abovementioned measured statistics. Mentioned controller 408 speed may, in some versions of the implementation of the measure referred to statistics in the first period of the speed control and calculates referred to speed on the basis of the abovementioned measured statistics in the second period, speed control. The mentioned statistics can, in some cases, implementation, contain dispersion of the mentioned channel coefficients.

Decoder 104 feedback can, in some cases, implementation, in addition, contain CPU 406 conversion to convert the mentioned channel coefficients with the aim of creating translated channel factors before decoding. Referred processor 406 transformation may, in some cases, implementation, convert mentioned channel coefficients coefficients in the frequency domain. Decoder 104 feedback can, in some cases, implementation, in addition, contain module 404 scaling to scale the mentioned channel coefficients on the basis of the statistics referred to channel coefficients and decoding mentioned scaled channel coefficients. Decoder 110 quantization of composition decoder 104 feedback can, in some cases, implementation, be implemented using the decoder vector quantizer with tree structure.

The present invention may, of course, be done in other ways than those specifically set forth herein, without departing from the essential characteristics of the invention. These options implementation should be considered in all respects illustrative and not restrictive, and all the dispersion within the semantic meaning and equivalence accompanying the claims are covered by it.

1. Method implemented by the receiving terminal and intended for quantization feedback channel status, comprising stages: determine the individual statistics for many of the coefficients of reference channel for the communication channel between the broadcaster and the receiving terminal; quantizing on an individual basis mentioned set of coefficients of reference channel on the corresponding bit speeds of quantization, which are determined on the basis of the abovementioned statistics, to generate the quantized coefficients of reference of the channel, and the total number of bits allocated to the said many of the coefficients of reference channel is fixed; and transmit the quantized coefficients of reference channel of the receiving the terminal referred to the transmitting station in the communication system.

2. The method according to claim 1, additionally contains a stage at which convert the mentioned coefficients of reference channel for obtaining the transformed coefficients of reference channel.

3. The method of claim 2, where the transformation of the mentioned coefficients of the report of the channel to obtain the transformed coefficients of reference channel: convert coefficients of reference channel frequency domain into the coefficients of reference channel temporary field; select the coefficients of reference channel temporary area within a predefined scatter delays; and convert mentioned selected coefficients of reference channel temporary area in the coefficients of reference of the channel's own.

4. The method according to claim 1, wherein when determining the statistics for many of the coefficients of reference channel determine the individual statistics for each coefficient of the reference channel.

5. The method according to claim 4, in which the individual statistics contains the relative power of the coefficient of reference channel.

6. The method according to claim 5, which referred to speed determined in descending order referred relative power.

7. The method according to claim 4, in which the individual statistics contains the variance associated with the coefficient of reference channel.

8. The method according to claim 7, which referred to speed determined in descending order referred dispersions.

9. The method according to claim 1, which referred to speed are determined on the basis of the statistical data collected for the current period, speed control, referred to calculate statistics before quantization of the coefficients of reference channel in the current period, speed control.

10. The method of claim 9, which referred to speed transmitted over slow feedback, in fact the quantized coefficients of reference channel is transmitted by rapid feedback channel.

11. The method according to claim 1, which referred to speed are determined on the basis of the statistical data collected for the previous period, speed control, referred to calculate statistics after quantization of the coefficients of reference channel in the previous period the speed control.

12. The method according to claim 11, in which the quantized coefficients of reference channel is transmitted by rapid feedback channel.

13. The method according to claim 1, wherein when determining the mentioned coefficients of reference channel mentioned coefficients of reference channel convert to white noise.

14. The method according to claim 1, wherein during the quantization on an individual basis referred to many factors of reference channel on the relevant speeds, defined on the basis of the abovementioned statistics, the scale mentioned coefficients of reference channel based on the statistics mentioned coefficients of reference channel and quantizing mentioned scaled coefficients of reference channel.

15. The method according to claim 1, wherein during the quantization on an individual basis referred to many factors of reference channel on the relevant speeds, defined on the basis of the abovementioned statistics to determine the distribution of bits for the ensemble of the coefficients of reference channel on the basis of the abovementioned statistics; and quantizing on an individual basis mentioned set of coefficients of reference channel at speeds defined on the basis of such distributions bits.

16. Encoder feedback in the receiving terminal, designed for the quantization of a feedback channel status, contains: a tool to calculate the metrics to calculate the individual statistics for many of the coefficients of reference channel for the communication channel between the broadcaster and the receiving terminal; and a variety of multi-speed for the quantization on an individual basis referred to many factors of reference channel on bit speeds of quantization, which are determined on the basis of the abovementioned statistics, the total number of bits used by the encoder feedback for the quantization of the said many of the coefficients of reference channel is fixed; and tool speed control for the above velocities for the mentioned .

17. Encoder feedback item 16, additionally contains the processor conversion to convert the mentioned coefficients of reference channel for obtaining the transformed coefficients of reference channel.

18. Encoder feedback according to clause 17, allows the processor to transform converts the coefficients of reference channel frequency domain into the coefficients of reference channel time area, select the mentioned coefficients temporary area within a predefined scatter delays and converts mentioned selected coefficients of reference channel temporary area in the coefficients of reference of the channel's own.

19. Encoder feedback item 16, in which the calculation of the metric determines individual statistics for each coefficient of the reference channel.

20. Encoder feedback .19 in which the individual statistics contains the relative power of the coefficient of reference channel.

21. Encoder feedback 20, in which the speed control determines referred to speed for the mentioned coefficients of reference channel in descending order referred relative power.

22. Encoder feedback .19 in which the individual statistics contains the variance associated with the coefficient of reference channel.

25. Encoder feedback paragraph 24, the encoder feedback transmits the speed to slow feedback and passes the quantized coefficients of reference channel for rapid feedback channel.

26. Encoder feedback item 16, in which the speed control determines the speed based on the statistics compiled for the previous period, speed control, referred to the statistics are calculated after quantization of the coefficients of reference channel in the previous period the speed control.

27. Encoder feedback .26, encoder feedback passes the quantized coefficients of reference channel for rapid feedback channel.

28. Encoder feedback item 16, additionally contains whitening filter to convert the mentioned coefficients of reference channel to white noise.

29. Encoder feedback according to clause 17, additionally contains a module for scaling scaling mentioned coefficients of reference channel based on the statistics mentioned coefficients of reference channel before quantization.

30. Encoder feedback item 16, in which the speed control determines the distribution of bits for the mentioned on the basis of the abovementioned statistics.

31. Encoder feedback item 16, in which multi-speed contain encoder from vector with tree structure, designed on the basis of the abovementioned statistics.

32. The method implemented in the transmitting station and designed to decode feedback channel status, comprising stages: take a receiving terminal feedback channel status, including many quantized coefficients of reference channel for the communication between a sending station and a receiving terminal, said a lot of quantized coefficients of reference channel quantized on an individual basis and the total number of bits allocated for the said many of the coefficients of reference channel is fixed; identify the individual bit rates quantization for the many factors of reference channel; and decode the mentioned set of coefficients of reference channel on the basis of such individual bit rates quantization.

33. The method according to .32, which in determining the appropriate speeds of quantization for many coefficients of reference channel accept referred to speed quantization from the receiving station for slow feedback.

34. The method according to .32, which in determining speeds for many coefficients of reference channels receive statistics mentioned coefficients of reference channel from the receiving station for slow feedback and calculate referred to speed on the basis of the abovementioned adopted statistics.

35. The method according to clause 34, in which the said adopted statistics contains dispersion mentioned coefficients of reference channel.

36. The method according to .32, which in determining the appropriate speeds of quantization for many coefficients of reference channel measure the statistics mentioned coefficients of reference channel and referred to calculate the speed of quantization for the mentioned many coefficients of reference channel on the basis of the abovementioned measured statistics.

37. The method according to clause 34, which referred to the statistics, measured in the first period, speed control, used to calculate the velocities in the second period, speed control.

38. The method according to clause 37, in which the said statistics contains dispersion mentioned coefficients of reference channel.

39. The method according to .32, additionally contains a stage at which convert the mentioned coefficients of reference channel for obtaining the transformed coefficients of reference channel.

40. The method according to clause 37, where the transformation of the mentioned coefficients of reference channel for obtaining the transformed coefficients of reference channel mentioned coefficients of reference channels are transformed coefficients in the frequency domain.

41. The method according to .32, additionally contains a stage at which scale the mentioned coefficients of reference channel based on the statistics mentioned coefficients of reference channel and decode mentioned scaled coefficients of reference channel.

42. Decoder feedback in a transmitting station, designed to decode the coefficients of reference channel provided as feedback by the receiving terminal, containing: the speed control for determination of individual bit rates of quantization for many coefficients of reference channel for the communication between a sending station and a receiving terminal, said a lot of quantized coefficients of reference channel quantized on an individual basis and the total number of bits allocated for the said many of the coefficients of reference channel is fixed; and decoder quantization for decoding mentioned many coefficients of reference channel the basis of such individual bit rates quantization.

43. Decoder feedback item 42, where the said means of control speed determines referred to the appropriate speed of quantization for many coefficients of reference channel through the reception mentioned speeds quantization from the receiving station for slow feedback.

44. Decoder feedback item 42, where the said means of control speed determines referred to the appropriate speed of quantization for many coefficients of reference of a channel by means of reception of statistics mentioned coefficients of reference channel from the receiving station for slow feedback and calculations referred to speeds of quantization on the basis of the abovementioned adopted statistics.

45. Decoder feedback item 44, in which the said adopted statistics contains the variance of the mentioned coefficients of reference channel.

46. Decoder feedback item 42, where the said means of control speed determines referred to the appropriate speed of quantization for many coefficients of reference channel by measuring the statistics mentioned coefficients of reference channel and calculations referred to speeds of quantization on the basis of the abovementioned measured statistics.

47. Decoder feedback p.46, where the said means of control the speed of the measures referred to statistics in the first period of the speed control and calculates referred to speed quantization on the basis of the abovementioned measured statistics in the second period, speed control.

48. Decoder feedback .47 in which the said statistics contains dispersion mentioned coefficients of reference channel.

49. Decoder feedback item 42, advanced translation of the mentioned coefficients of reference channel for obtaining the transformed coefficients of reference channel to decode.

50. Decoder feedback § 49, in which the mentioned transformation coefficients of reference channel for obtaining the transformed coefficients of reference channel mentioned coefficients of reference channels are transformed coefficients in the frequency domain.

51. Decoder feedback item 42, additionally performs scaling of the mentioned coefficients of reference channel based on the statistics mentioned coefficients of reference channel and decoding mentioned scaled coefficients of reference channel.

52. Decoder feedback item 42, the decoder quantization implemented using the decoder vector quantizer tree structure.