JP2009505526A - Wireless communication device including integrated space-time optimal filter (JSTOF) using Cholesky decomposition and eigenvalue decomposition - Google Patents

Abstract

A wireless communication device may include a wireless transmitter and a wireless receiver. The wireless receiver may include a filter that reduces co-channel interference and may include a multi-channel space-time filter circuit. The multi-channel space-time filter circuit is configured to estimate n spatio-temporal filter weights and multi-channel impulse responses (CIR) based on Cholesky decomposition and eigenvalue decomposition, thereby dividing n pieces of communication signals. The signal part of is filtered. The filter may further include a multi-channel matched filter circuit that receives the signal from the multi-channel space-time filter circuit and is provided by a channel impulse response estimate from the space-time filter circuit. Has a filter response.

Description

(Field of Invention)
The present invention relates to wireless communication systems such as cellular communication systems, and more particularly to reducing unwanted interference by filtering received wireless signals.

(background)
To meet the requirements of Downlink Advanced Receiver Performance (DARP) standardized by 3rd Generation Mobile Communication System and 3rd Generation Mobile Communication Partnership Project (3GPP) Research has been conducted on interference canceling matched filters (ICMF) and joint demodulation (JDM). Some of these proposals are described in the following papers and literature non-patent documents 1 to 10.

In addition to addressing the DARP requirement, the current Global System for Mobile Communications (GSM) cellular system addresses co-channel interference (CCI) on the mobile station (MS) side. It is also necessary to deal with (interference). Some single channel structures and pre-filters have been used to facilitate interference cancellation and make some channel impulse response (CIR) estimation. Furthermore, some systems have used signal-to-interference ratio maximization to design an integrated single channel space-time filter and single channel CIR estimation. Other systems have used constrained minimization of mean square error to design single channel spatial filters. Other systems have used single channel spatial filters designed by rank 1 approximation of ML channel estimation. The target application of these systems is a base station that can use a physical antenna array including a plurality of antennas.
Liang et al., “A Two-Stage Hybrid Approach for CCI / ISI Reduction with Space-Time Processing”, IEEE Communication Letter, November 1997, Vol. 1, No. 6 Pipon et al., “Multichannel Receives Performance Comparison in the Presence of ISI and CCI”, 1997 13th Intl. Conf. on Digital Signal Processing, July 1997 Spagnolini, “Adaptive Rank-One Receiver for GSM / DCS Systems”, IEEE Trans. on Vehicular Technology, September 2002, Vol. 51, No. 5 "Feasibility Study on Single Antenna Interference Cancellation (SAIC) for GSM Networks," 3GPP TR 45.903 Version 6.0.1, Release 6, European Telecommunity4. “Radio Transmission and Reception (Release 6)”, 3GPP TS 45.005 Version 6.8.0; European Telecommunications Standards Institute, 2005 Stoica et al., “Maximum Likelihood Parameter and Rank Estimate in Reduced-Rank Multivariate Linear Regressions”, IEEE Trans. On Signal Processing, December 1996, Vol. 44, No. 12 Kristensson et al., “Blind Subspace Identification of a BPSK Communication Channel”, Proc. 30th Asilomar Conf. On Signals, Systems and Computers, 1996 Golub et al., "Matrix Computations", 3rd edition, 1996 Trefethen et al., “Numerical Linear Algebra”, 1997 Press et al., “Numeric Recipes in C”, 2nd edition, 1992

(Overview)
In general, the present disclosure relates to a wireless communication device, which may include a wireless transmitter and a wireless receiver. More specifically, the wireless receiver may include a filter that reduces channel interference and collectively estimates space-time filter weights and multi-channel impulse responses (CIR) based on Cholesky decomposition. Includes a multi-channel space-time filter circuit that filters the signal portion divided from the communication signal. The multi-channel matched filter circuit receives a multi-channel signal from the multi-channel space-time filter circuit and has a filter response provided by a channel impulse response estimate from the space-time filter circuit. A standard filter is operable when the interference level is below a predetermined threshold and can be formed as a matched filter, a cross-correlation circuit, and a switching mechanism that switches the signal portion to the matched filter and the cross-correlation circuit.

  In one aspect, the multi-channel space-time filter circuit includes a plurality of multipliers and delay circuits, each receiving n signal portions. Each of the multiplier and delay circuit is operable based on a spatio-temporal filter weight. Each of the multiplier and the delay circuit includes two multiplier circuits and a delay circuit. Each multiplier and delay circuit can operate with a one symbol delay. An integrated filter weight and channel estimator is operably connected to the multi-channel space-time filter, receives training sequence (TS) symbols and timing uncertainty data, and for multi-channel space-time filter circuits Generate spatiotemporal filter weights. The adder sums the data from the multiplier and delay circuit for each channel. The equalization circuit is operable with a multi-channel matched filter.

  Various objects, features and advantages will become apparent from the following detailed description when considered in conjunction with the accompanying drawings.

(Detailed description of preferred embodiments)
Several non-limiting embodiments are described more fully hereinafter with reference to the accompanying drawings, which illustrate preferred embodiments. However, these embodiments may be embodied in many other forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope to those skilled in the art. Like numbers refer to like elements throughout, and prime code notation is used to indicate similar elements in alternative embodiments.

  According to one embodiment, the third generation mobile communication system standardization project (3GPP) addresses the requirements for conformance to the Downlink Advanced Receiver Performance (DARP) standard, as well as the current global system for mobile communication Co-channel interference (CCI) on the mobile station (MS) side in the (GSM) communication system is also dealt with.

  In the embodiment shown in FIG. 1, a multi-channel pre-filter is provided that operates to adaptively and optimally cancel interference and provide a channel impulse response (CIR) estimate. In one non-limiting example, the pre-filter is an integrated space-time optimal filter (JSTOF: Joint Space-) that uses two important components: (1) multiple-input-multiple-output (MIMO). A multi-channel matched filter using a time optimal filter (2) and a multiple-input-single-output (MISO) can be used. In a typical mobile station using a single available antenna, a virtual antenna array is constructed internally by combining over-sampling with separation into real and imaginary parts that receive the samples. be able to.

  In one non-limiting embodiment, the signal from the virtual antenna array is fed into the JSTOF where the optimal weight for the interference cancellation filter using MIMO is estimated. At the same time, the multi-channel CIR for the desired signal is also estimated collectively. With the output of JSTOF, interference can be filtered and sent to a multi-channel matched filter using MISO. The filter response of the matched filter is provided by the CIR estimate from JSTOF.

  The output of the multi-channel matched filter is passed to a Viterbi equalizer, which removes inter-symbol interference (ISI) and provides soft decisions for further processing. The single channel response required for the equalizer can be formed by synthesis of convolutional CIR from JSTOF. This pre-filter automatically switches to a conventional or standard filter of a conventional receiver in any situation where AWGN (additive white Gaussian noise) is dominant, and in any situation where interference is dominant It is also possible to switch back to the receiver using JSTOF. This automatic switching function reduces losses in situations where AWGN is dominant.

An example of a DARP-compliant receiver pre-filter or interference cancellation filter using JSTOF is shown at 10 in FIG. 1, where the oversampling rate is 2, and from X ₁ (k) As indicated by X ₄ (k), the number of virtual antennas is 4 (M = 4). Throughout this description, the prefilter 10 may refer to an interference cancellation filter or JSTOF filter, and functions as a prefilter in a DARP-compliant receiver. A receiver incorporating this filter 10 can be described as a JSTOF receiver, as indicated by the dashed line 11 in FIG.

  FIG. 1 shows examples of various circuit blocks used in the filter 10. The input signal is received by the derotation circuit 12. The derotated output signal is split, a portion of which enters a conventional receiver filter 14 that includes a 2: 1 switch 16, and further includes a matched filter 18 and a shortened training sequence (TS). It is output to the cross-correlation circuit 20 that receives the symbols. The 2: 1 switch 16 is operable to enable switching between the filter 14 and the DARP-compatible prefilter 10 using JSTOF.

The other part of the output signal from the derotation circuit 12 is divided into even and odd samples as a part of the virtual antenna 24, and is further divided into a real signal and an imaginary signal again to obtain a multi-channel space-time filter circuit. The X ₁ (k) to X ₄ (k) input signals that enter the JSTOF circuit 30, which are also called, are formed. The output signal from the JSTOF circuit is fed to the multi-channel matched filter circuit 32, and the output signal therefrom is fed to the rescaling circuit 34 and then to the multiplexer circuit 36 as data (d ₁ ). Multiplexer circuit 36 also receives a channel (c ₁ ) response. When conventional filter 14 is connected, multiplexer 36 receives data (d ₂ ) and channel (c ₂ ) responses from matched filter circuit 18 and cross-correlation circuit 20. The signal is sent to the Viterbi equalizer 38 and becomes a soft decision output.

Further details of the JSTOF and multi-channel matched filter are shown in FIG. 2, where the number of time delay samples used in the JSTOF circuit is 2 (N = 2). It is shown in more detail where separate inputs X ₁ (k) to X ₄ (k) are received into JSTOF. The JSTOF circuit 30 includes channelized multipliers 40, 42, also referred to as a mixer, a delay unit 44, and an adder 46, which is a multi-channel matched filter 48 for each of the four channels shown. Further, the signal from the matched filter is sent to the adder 50. An integrated optimal filter weight and channel estimation circuit 52 receives the TS symbols and timing uncertainty signal and produces a weight (W _OPT ) for use in the mixers 40, 42.

  Therefore, as shown in FIG. 1, it is possible to incorporate the prefilter function into the conventional GSM receiver as described above by adding a prefilter department in parallel with the conventional matched filter. A conventional software / hardware Viterbi equalizer 38 can be used as it is. In one non-limiting example, an embedded DARP-capable receiver was simulated and examined against a DARP test case, where the receiver has a frame error rate (FER) for one of the AMR voice channels. : Frame error rate) exceeding the specified performance by a difference of 1.6 to 6.9 dB.

  FIG. 2a is a flowchart illustrating a high-level method associated with the described system, the various steps being shown as non-limiting examples. The various steps start with a reference number starting from 100. The incoming communication signal is derotated (block 100) and sent to the virtual antenna. The communication signal is divided into even and odd samples (block 102), and each even and odd sample is then divided into real and imaginary signal portions (block 104). The communication signal from the virtual antenna is fed into the JSTOF circuit, where the communication signal is multiplied and delayed (block 106) and then summed (block 108), all of which are the first multiple input multiple Part of an output (MIMO) integrated space-time optimal filter (JSTOF). After summing, the summed signal is fed into a multi-channel multiple-input single-output (MISO) matched filter circuit (block 110) and then summed (block 112) and then Viterbi equalizer as a single output signal (Block 114) in which soft decisions are made (block 116).

  In operation, derotation circuit 12 is operable for GMSK modulated signals and frequency offsets that are part of its signaling protocol. Before any derotation, the constellation is dynamic, but after derotation, the constellation becomes static, in other words, every symbol is usually rotated by symbols rotated to 0 and 180 degrees. Will be concentrated on the two points. Therefore, the GMSK signal can be treated as a normal binary phase shift keying (BPSK) signal. The forefront derotation is used for even and odd sampling, which is useful because of the oversampling rate. For example, in a conventional receiver, this rate is usually 1, ie 1 sample per symbol.

  The virtual antenna 24 can continuously increase the sampling rate to 2 samples per symbol in the order coming from the baseband filter to form two separate channels, even and odd. Prior to this process, the odd / even samples are interleaved in succession. These signals are then further divided into a real signal portion and an imaginary signal portion to form the four individual channels of the virtual antenna. It should be appreciated by those skilled in the art that other numbers (eg, one or more) of virtual antennas / channels may be used in some embodiments.

  As best shown in FIG. 2, the above signal is then fed into multipliers 40, 42 and a unit delay circuit 44 which delays, for example, one symbol, so that the signal is subjected to multiplication and delay processing. Subsequently, a multiplication operation is performed. This is apparent from the two multipliers 40 and 42 and one delay circuit 44. As shown, following this operation, the adder 46 adds up. This part of the system can operate as a multi-channel two-dimensional filter. One of the dimensions arises because of time delay, and the other is derived from the virtual antenna as described above, that is, the spatial dimension, and thus two dimensions are spatio-temporal. A filter will be formed.

  The multiplier receives weights from the integrated optimal filter weight and channel estimator 52 because each incoming signal is apparently in use with other channels. The integrated optimal filter weight and the weight obtained from the channel estimator 52 are fed to the multiplier.

  In a non-limiting example, the weight is also an 8 × 4 dimensional matrix, that is, 32 weights. With respect to training sequence symbols input to the integrated optimal filter weights and channel estimator 52, in some non-limiting examples, there are generally about 26 known symbols and which training sequence is in the packet. I know it is included. To determine timing, in one non-limiting example, a +/− 3 or 7 position search can be used. Impulse response of multichannel matched filters.

を使用して、システムがチャネル応答に整合し、整合フィルタを経ると信号が強化されるようにすることができる。

Can be used so that the system matches the channel response and the signal is enhanced after going through a matched filter.

  As shown in FIG. 1, although not essential, rescaling can occur as a convenience of hardware or software. This rescaling circuit 34 allows a larger operation on the input to a 4-bit or 5-bit Viterbi equalizer 38 as a non-limiting example. The dynamic range of the signal can be readjusted so that this signal can be fed into a 4-bit or 5-bit circuit.

As described above, multiplexer 36 receives data and channel response signals d ₂ and c ₂ from conventional filter receiver 14, or data and channel response signals d ₁ and c ₁ from JSTOF receiver 10. To enable switching between the two. The JSTOF receiver will cause some loss if there is no interference at all, in other words just pure white noise. In that case, the conventional receiver 14 can be used and functions properly. Thus, the circuit can be switched back to the conventional filter without the loss caused by the JSTOF receiver and the circuit. The switching is based on an estimate of SINR _OUT minus SINR _INP . If this value is below the threshold, the system determines that there is little interference and no interference cancellation by the JSTOF receiver is required. Thus, by switching the 2: 1 switch 16, the filter of the conventional receiver 14 is used.

  The circuit is operable in beamforming systems and other systems. This type of system also enables improved signal-to-interference ratios as well as improved bit error rate (BER). This is believed to affect the top-level protocol and telephone calls and other communication situations used with these circuits.

  In one embodiment, the multi-channel structure of the filter 10 using JSTOF is used, and the JSTOF circuit 30 using MIMO performs spatio-temporal filter weights and channel estimation that are different from prior art solutions. This circuit provides the ability to efficiently address interference suppression and improve performance for both synchronous and asynchronous interference. Some simulations have shown that some solutions do not provide the required performance in the light of DARP test cases, according to some prior art.

  The multi-channel matched filter circuit 32 using MISO is characterized by improving the overall error rate performance and reducing the complexity of the equalizer by omitting the multi-channel Viterbi equalizer. Built-in automatic switching between receivers using JSTOF and conventional receivers reduces losses under AWGN conditions.

  Appropriate receiver structures can be used to meet DARP requirements. An interference cancellation matched filter (ICMF) can address interference suppression using the example virtual antenna described and beamforming. The circuit is sensitive to channel impulse response (CIR) estimation errors of the desired signal. Integrated demodulation (JD) has shown desirable performance for various test cases. In addition to the difficulty of addressing the suppression of asynchronous interference, a high degree of computational complexity may be required to determine the CIR of interference.

  In one embodiment, the virtual antenna 24 can operate with adaptive space-time filtering, thus allowing the use of an integrated space-time optimal filter (JSTOF) circuit 30. One of the differences from ICMF is that the spatio-temporal filter weights used to suppress interference and the CIR estimate of the desired signal are estimated and optimized in JSTOF, On the other hand, in ICMF, these two are estimated separately. The JSTOF circuit 30 can be a multiple-input multiple-output (MIMO) circuit that utilizes the desired rank-down nature of the CIR matrix in a spatio-temporal arrangement. Simulations have shown satisfactory performance for various DARP test cases. If fixed-point Cholesky decomposition and EVD / SVD are feasible, the computational load is considered acceptable.

  This method has a certain degree of simplicity and low computational complexity. It is also a robust method because the system makes few hypotheses about the interference sources. In addition, since this solution is incorporated into the input data as a pre-processing step, the system can continue to use the existing equalizer structure. Thus, if available, the system can use the HW equalization accelerator.

  To support the evaluation of this technology, the system-level block error rate (BLER) simulator has been extended to support all interference models / scenarios used in the 3GPP DARP standard. became.

  Here, the simulation performance for the DARP test case using the JSTOF circuit will be described as follows. Of course, spatio-temporal processing for integrated interference reduction and channel estimation has been used in base stations where an array of M antennas is available. Assuming that the equivalent channel response for a single target user can be modeled as an L-tap Finite Impulse Response (FIR) filter, the snapshot sample of the received baseband signal is Represented,

ここで、ｘ（ｋ）は、アンテナからの出力を表現しているＭ×１ベクトルであり、Ｈは、アンテナ・アレイに関するチャネル応答を含んでいるＭ×Ｌ行列であり、ｓ（ｋ）は、対応する送信シンボルのＬ×１ベクトルであり、ｖ（ｋ）は、ＡＷＧＮおよび干渉を含んでいるＭ×１ベクトルである。式（１）の時空間的拡張は、ｘ（ｋ）の時間遅延バージョンＮ個を積み重ね、より大きいＭＮ×１ベクトル

Where x (k) is an M × 1 vector representing the output from the antenna, H is an M × L matrix containing the channel response for the antenna array, and s (k) is , L × 1 vector of the corresponding transmission symbol, and v (k) is an M × 1 vector containing AWGN and interference. The spatio-temporal extension of equation (1) stacks N time-delayed versions of x (k), resulting in a larger MN × 1 vector

にすることによって、以下のように得ることができ、

Can be obtained as follows,

ここで、ＭＮ×（Ｌ＋Ｎ−１）行列である

Where MN × (L + N−1) matrix

は、Ｈのブロック・テプリッツ（Ｔｏｅｐｌｉｔｚ）バージョンであり、

Is the Block Toeplitz version of H;

である。トレーニング・シーケンスに対応するサンプルは、以下のように収集することができ、

It is. Samples corresponding to the training sequence can be collected as follows:

ここでｐ＝Ｐ−Ｌ−Ｎ＋２であり、Ｐは、トレーニング・シーケンスのシンボル数であり、

Where p = P−L−N + 2, where P is the number of symbols in the training sequence,

はＭＮ×ｐ行列であって、

Is an MN × p matrix,

は、トレーニング・シンボルの（Ｌ＋Ｎ−１）×ｐ畳み込み行列である。統合型最適化（ｊｏｉｎｔ ｏｐｔｉｍｉｚａｔｉｏｎ）とは、時空間フィルタ用の自明でないＭＮ×１ウェイト・ベクトル

Is a training symbol (L + N−1) × p convolution matrix. Joint optimization is a non-trivial MN × 1 weight vector for space-time filters

と、フィルタを経た後の自明でない（Ｌ＋Ｎ−１）×１チャネル推定ベクトル

And (L + N−1) × 1 channel estimation vector that is not obvious after passing through the filter

を求めることによって、フィルタの出力残留干渉を最小化することであり、言い換えれば、以下の最適化問題を解くことである。

Is to minimize the residual output interference of the filter, in other words, to solve the following optimization problem.

最適ウェイトは、以下のように求めることができ、

The optimal weight can be obtained as follows:

最適チャネル推定値

Optimal channel estimate

は、行列

Is a matrix

の最小固有値に相当する固有ベクトルであり、ここで、

Is the eigenvector corresponding to the smallest eigenvalue of, where

である。

It is.

  Noise pulse interference component of spatio-temporal model of equation (3)

がもはや白色ではなく、未知の共分散行列

Is no longer white and the unknown covariance matrix

を用いて近似的にガウス分布されていると考えると、チャネルの最適な推定値

The approximate estimate of the channel

は、最尤（ＭＬ：ｍａｘｉｍｕｍ−ｌｉｋｅｌｉｈｏｏｄ）推定値となり、これは、以下の数量の最小化である。

Is the maximum-likelihood (ML) estimate, which is a minimization of the following quantity:

この限定されない時空間モデルにおいて、独立したチャネルの数は常にＭ以下であり、

In this non-limiting space-time model, the number of independent channels is always less than or equal to M,

は通常ランク落ちであって、すなわち、ランク

Is usually a rank drop, ie rank

である。ランク落ちＭＬ問題は、時空間フィルタのランク１近似に用いることができる。

It is. The rank drop ML problem can be used for rank-1 approximation of spatio-temporal filters.

  In one embodiment, the JSTOF circuit can use another approach to find an integrated optimal solution for filter weights and channel estimates.

のＭＬ推定値を求めることが可能となる。推定値は、以下のように分解することができ、

It is possible to obtain the ML estimated value of. The estimate can be decomposed as follows:

ここで、

here,

は

Is

の空間行列の推定値であり、

Is an estimate of the spatial matrix of

は

Is

の時間行列の推定値である。これらは、以下から得ることができ、

Is an estimate of the time matrix. These can be obtained from

ここで、

here,

は、コレスキー分解であり、

Is the Cholesky decomposition,

は、Ｍ個の固有ベクトルから成るが、この固有ベクトルは、行列

Consists of M eigenvectors, which are the matrix

の上位Ｍ個の固有値に相当する。

Correspond to the top M eigenvalues.

次のステップでは、時空間フィルタ用の最適ウェイトを、以下から得ることができ、

In the next step, the optimal weight for the spatio-temporal filter can be obtained from

さらに、最適チャネル推定値は、以下の通りである。

Further, the optimum channel estimation value is as follows.

次いで、式（１４）の最適時空間フィルタをアンテナ・アレイ２４からのサンプルへ適用することが可能になる。フィルタ３０の出力は依然としてＭ個のチャネルを有することが明らかであるので、これはＭＩＭＯシステムである。式（１５）の最適チャネル推定値を、多重チャネル整合フィルタ３２で使用することができる。その後、整合フィルタの出力が合成（合計）されて、リスケーリング回路３４にて、変更された所望のレベルへとリスケールされる。最終的な出力は、単一チャネルのサンプル・ストリームであり、これをビタビ等化器３８へ送り込むことができる。なお、ＪＳＴＯＦを経た後のチャネル・タップ数が、ＪＳＴＯＦを経る前のモデルのチャネル・タップ数Ｌと比較すると、Ｌ＋Ｎ−１へ変わっていることにも留意されたい。

The optimal spatio-temporal filter of equation (14) can then be applied to the samples from the antenna array 24. This is a MIMO system because it is clear that the output of the filter 30 still has M channels. The optimal channel estimate of equation (15) can be used by the multi-channel matched filter 32. Thereafter, the outputs of the matched filters are combined (summed) and rescaled to the changed desired level by the rescaling circuit 34. The final output is a single channel sample stream that can be fed into the Viterbi equalizer 38. It should be noted that the number of channel taps after passing through JSTOF is changed to L + N−1 as compared with the number of channel taps L in the model before passing through JSTOF.

  It has been observed by simulation that the JSTOF receiver causes an additional 1 dB loss under a pure AWGN situation compared to a conventional receiver using a conventional filter. In order to reduce losses, automatic switching means between JSTOF receivers and conventional receivers have been developed. The switching is based on the difference measurement of JSTOF input SINR and output SINR. If the difference is less than a predefined threshold, the JSTOF receiver is turned off and the conventional receiver is turned on. The input SINR is

の推定が式（１０）にて行われれば、容易に計算することができ、

Can be easily calculated if Eq. (10) is estimated,

出力ＳＩＮＲは、式（１４）および（１５）を元に計算することができる。

The output SINR can be calculated based on equations (14) and (15).

移動体側では、仮想アンテナ・アレイを、図１に示すように、オーバー・サンプリングと、実部および虚部への分離とを組み合わせることによって設置することができる。

On the mobile side, a virtual antenna array can be installed by combining over-sampling and separation into real and imaginary parts as shown in FIG.

  According to various embodiments, the integrated optimal MIMO space-time filter and channel estimation described in equations (14) and (15) enhances the performance of interference suppression. The MISO multi-channel matched filter 32 based on channel estimation in equation (15) improves the error rate performance and at the same time reduces the complexity of the Viterbi equalizer 38. The means of automatically switching between JSTOF receivers and conventional receivers reduces losses under pure AWGN conditions.

  JSTOF defined by equations (6)-(17) can be implemented in various ways from the viewpoint of numerical stability and computational complexity. The main difference is the autocorrelation matrix

の逆を計算する方法、およびチャネル

To calculate the inverse of, and channel

を低ランクで推定する方法である。

This is a method of estimating at a low rank.

  One of the implementations described above is

のコレスキー分解による行列の逆変換および式（１３）の行列

Matrix inverse transformation by Cholesky decomposition of and the matrix of equation (13)

の固有値分解である。具体的には、

Is the eigenvalue decomposition of. In particular,

は対称正定値であるので、コレスキー分解は、

Is a symmetric positive definite, so the Cholesky factorization is

になる。

become.

は、以下のように書き換えることができ、

Can be rewritten as

ここで、

here,

である。

It is.

  Of course, the inverse transform is actually

の平方根を用いて実行されており、後退代入によって逆変換の陽的計算を省いてもよい。さらに、

The explicit calculation of the inverse transformation may be omitted by backward substitution. further,

は、相互消去の構造のために、数値的に安定している。これは、

Is numerically stable due to the mutual erasure structure. this is,

の条件数がめったに３００を超えないことを示したシミュレーションによって検証された。当業者には当然のことながら、このことは、

This was verified by simulations showing that the condition number rarely exceeded 300. As will be appreciated by those skilled in the art,

の固有値分解が、通常の用途では過度に高度なアルゴリズムを必要としないことを意味すると言える。実際、この手法は、本明細書にて概説する手法の中でも最も低い計算複雑性を有する可能性がある。

It can be said that the eigenvalue decomposition of does not require an excessively sophisticated algorithm in normal applications. In fact, this approach may have the lowest computational complexity of the approaches outlined herein.

  One possible numerical problem is

のコレスキー分解であるが、それは、場合によっては条件数が比較的大きくなる可能性があり、正定値の特性が丸め誤差によって或る程度までオフセットされることがあり得るからである。ただし、シミュレーションでは、

This is because the condition number may be relatively large in some cases, and the positive definite characteristic may be offset to some extent by a rounding error. However, in simulation,

の条件数は、キャリア対干渉（Ｃ／Ｉ：ｃａｒｒｉｅｒ−ｔｏ−ｉｎｔｅｒｆｅｒｅｎｃｅ）比を非常に高くおよび非常に低くするといった一部の極端なシナリオにおいてさえも、１０^７未満であることが示された。

The condition number of was shown to be less than 10 ⁷ even in some extreme scenarios, such as very high and very low carrier-to-interference (C / I) ratios. .

  According to another embodiment, using QR decomposition in the sample region,

の逆変換の直接計算を省いてもよい。式（３）の

The direct calculation of the inverse transformation of may be omitted. Of formula (3)

は、フル列ランクを有するので、一意的なＱＲ分解を有し、

Has a unique QR decomposition because it has a full column rank,

ここで

here

は、直交列を伴うｐ×ＭＮ行列であり、

Is a p × MN matrix with orthogonal columns,

は、フルランクのＭＮ×ＭＮ上三角行列である。以下を示すことができ、

Is a full rank MN × MN upper triangular matrix. Can show:

さらに、式（１３）の

Furthermore, in equation (13)

は、以下にて再定義される

Is redefined as

を用いて式（１９）の形式で書くことができる。

Can be written in the form of equation (19).

低ランクのチャネル推定は、先の手法におけるように、

Low rank channel estimation, as in the previous method,

の固有値分解を用いて実行すればよく、（１４）の最適フィルタ・ウェイト行列は、以下のように変わる。

The optimal filter weight matrix in (14) changes as follows.

この手法は、

This technique is

であることを示すことができるので、基本的にはサンプル領域におけるコレスキー分解の同等バージョンである。これは、数値安定性を向上させてきたが、その代償として、ＱＲ分解の複雑性が高まり（一定サイズの行列に関しておよそ２倍の演算が必要になる）、サンプル行列が大きくなっている（Ｍ＝４、Ｎ＝２、Ｌ＝５である１つの事例では、約３倍の行を有する）。

This is basically an equivalent version of the Cholesky decomposition in the sample domain. This has improved numerical stability, but at the cost of increased QR decomposition complexity (approximately twice as much computation is required for a fixed size matrix) and a larger sample matrix (M One case where = 4, N = 2, L = 5 has about three times as many rows).

  The above two methods may be performed by backward substitution, but still require the inverse calculation of the triangular matrix. Here, if we look at yet another method, namely the singular value decomposition (SVD) method, in some applications, the inverse transformation of the matrix may be omitted, and the numerical stability is further improved. Can do. This technique starts with SVD of the sample matrix of equation (3),

ここで

here

は、直交列を伴うｐ×ＭＮ行列であり、

Is a p × MN matrix with orthogonal columns,

は、ＭＮ×ＭＮ直交行列であり、さらにΣ_ｘは、Σ_ｘ＝ｄｉａｇ（σ_１，・・・，σ_ＭＮ）であるＭＮ×ＭＮ直交行列であって、その対角線上に特異値を伴う。以下を示すことができる。

Is an MN × MN orthogonal matrix, and Σ _x is an MN × MN orthogonal matrix in which Σ _x = diag (σ ₁ ,..., Σ _MN ), with a singular value on its diagonal. The following can be shown.

式（１３）の

Of formula (13)

は、以下にて定義される

Is defined by

を用いて、やはり式（１９）の形を取る。

Then, the shape of the equation (19) is also used.

チャネル推定値は、

The channel estimate is

のＳＶＤによって求められ得、フィルタ・ウェイト行列は、以下のように記され得、

And the filter weight matrix can be written as:

ここで

here

は、

Is

の上位Ｍ個の右特異ベクトルを含む。この手法におけるＳＶＤは、先の２つの手法で使用されているコレスキー分解およびＱＲ分解よりも多くの計算を必要とすることがある。

Of the top M right singular vectors. SVD in this approach may require more computation than the Cholesky and QR decompositions used in the previous two approaches.

  Comparing the three approaches outlined above (ie Cholesky, QR, and SVD), the table in FIG. 9 shows step-by-step calculations for an example where M = 4, N = 2, and L = 5. ing. To determine the best timing for the burst, JSTOF searches a number of timing hypotheses and selects the one that corresponds to the smallest output residual as the best timing. The output residual is defined by

検索プロセスでは、基本的に、各仮説に関して表に記載された演算を繰り返すのであるが、連続したタイミング仮説に由来する入力サンプル行列は、列の追加および削除によって若干変わる。場合によってはアップデートおよびダウンデートのアルゴリズムが一部の演算に適用可能となり、全体的な計算負荷が軽減される可能性もあり得る。

The search process basically repeats the operations listed in the table for each hypothesis, but the input sample matrix derived from successive timing hypotheses varies slightly with the addition and deletion of columns. In some cases, update and downdating algorithms can be applied to some operations, and the overall computational burden may be reduced.

  what if

が時刻ｋにおけるサンプル行列を表すものとする。これは、式（３）から以下の式へ分割される。

Represents a sample matrix at time k. This is divided from equation (3) into the following equations:

ここで、

here,

時刻ｋ＋１におけるサンプル行列は、以下のように表される。

The sample matrix at time k + 1 is expressed as follows.

時刻ｋ＋１における自己相関行列は、以下の形を取る。

The autocorrelation matrix at time k + 1 takes the following form.

これは、ランク１ダウンデートとランク１アップデートとの組み合わせである。コレスキー分解のアップデート／ダウンデートのための双曲線回転によるアルゴリズムの１つについて、Ｇｏｌｕｂ他による、Ｍａｔｒｉｘ Ｃｏｍｐｕｔａｔｉｏｎｓ，第３版，１９９６年で説明されている。

This is a combination of rank 1 downdate and rank 1 update. One hyperbolic rotation algorithm for updating / downdating Cholesky decomposition is described in Golub et al., Matrix Computations, 3rd edition, 1996.

  Another applicable update / downdate algorithm, disclosed in Golub et al., Relates to QR decomposition based on Givens rotation. As will be appreciated by those skilled in the art, the presented techniques can be used for specific applications and will, of course, depend on factors such as available processing resources and computational complexity. Furthermore, as will be understood by those skilled in the art, other techniques can be used.

  The performance of receivers using JSTOF has been evaluated by Matlab simulation using an extended BLER simulation engine. The receiver parameters according to JSTOF can be set using various modes. Examples of values are as follows:

    1) An oversampling ratio (OSR) can be selected as 2, which maps to the number of virtual antennas (M), which is 4 in this non-limiting example. Simulations show that lowering OSR to 1 causes significant performance degradation.

    2) The number of time delay samples (N) can be selected as 2. However, increasing this number does not always improve performance.

    3) The low rank of the channel response matrix can be selected as M. Increasing or decreasing the rank does not necessarily improve performance.

    4) The threshold for automatic switching can be 4.75 dB.

    5) The soft decision output can be quantized with a 5-bit width. For DTS-5, increasing the width to 8 bits can improve performance slightly. Soft decision correction can be enabled.

  The AMR voice channel, TCH-AFS 12.2, can be used to evaluate JSTOF performance with respect to FER. Throughout the simulation, the propagation conditions can be assumed to be TU 50 km / h-1950 MHz. In the simulation, 1000 trials (blocks) were performed for each case.

  The FER of the receiver is shown in the graph of FIG. 3 as opposed to the carrier to interference (C / I) ratio. The differences compared to the prescribed reference performance are shown in the table below.

純粋なＡＷＧＮ状況下、およびＤＴＳ−５状況下の受信器性能について、それぞれ図４および図５のグラフにて、自動切り替え手段を伴う場合と伴わない場合に関して示されている。この手段によって、ＡＷＧＮでは、約１ｄＢ（ＦＥＲ＝１０％にて）ロスが低減し、ＤＴＳ−５では、わずかなロスしか生じなかった。

Receiver performance under pure AWGN and DTS-5 conditions is shown in the graphs of FIGS. 4 and 5, respectively, with and without automatic switching means. By this means, loss of about 1 dB (at FER = 10%) was reduced in AWGN, and only a slight loss was generated in DTS-5.

  A JSTOF receiver can also have a plurality of Viterbi equalizers, followed by a multi-channel matched filter to synthesize the soft decisions that have passed through the equalizer. The results are shown in the graph of FIG. 6 in comparison with the prototype.

  The performance can also be evaluated using a partially modified test case DTS-5R, where the asynchronous interference delay can be set. The performance at burst lengths of 0, 1/4, 1/2, and 3/4 is shown in the graph of FIG. The results show that JSTOF receiver performance degrades “slowly” with extreme interference delays.

  The receiver described above can be advantageously used for mobile radio devices (eg, cellular devices) as well as, for example, cellular base stations. An example of a mobile wireless communication device 1000 that may be used is further described in the example below with respect to FIG. The device 1000 includes a housing 1200, a keypad 1400, and an output device 1600 as specific examples. The output device shown is a display 1600, which is preferably a fully graphic LCD. Other types of output devices may be used instead. A processing device 1800 is included in the housing 1200 and is connected between the keypad 1400 and the display 1600. The processing device 1800 controls the overall operation of the mobile device 1000 in addition to the operation of the display 1600 in accordance with the operation of the keys on the keypad 1400 by the user.

  The housing 1200 may be elongated in the vertical direction, or may have other sizes and shapes (including clamshell housing structures). The keypad may include a mode selection key for switching between text input and telephone communication input, or other hardware or software.

  In addition to the processing device 1800, other parts of the mobile device 1000 are shown schematically in FIG. This includes communication subsystem 1001, short-range communication subsystem 1020, keypad 1400 and display 1600, and other input / output devices 1060, 1080, 1100, 1120, as well as memory devices 1160, 1180, and Various other device subsystems 1201 are included. Mobile device 1000 is preferably a two-way RF communication device having voice and data communication capabilities. In addition, the mobile device 1000 preferably has a function capable of communicating with other computer systems via the Internet.

  The operating system software executed by the processing device 1800 is preferably stored in a fixed storage device such as a flash memory 1160, but may be other such as a read only memory (ROM) or similar storage element. May be stored in different types of memory devices. In addition, system software, specific device applications, or portions thereof, may be temporarily loaded into volatile storage, such as random access memory (RAM) 1180. Communication signals received by the mobile device may be stored in the RAM 1180.

  Processing device 1800 enables execution of software applications 1300A-1300N on device 1000 in addition to operating system functions. A predetermined set of applications that control the basic operation of the device, such as data and voice communications 1300A and 1300B, may be installed on the device 1000 during manufacture. In addition, a personal information manager (PIM) application may be installed during manufacturing. The PIM is preferably capable of organizing and managing data items and task items such as e-mail, calendar events, voice mail, reservations and the like. Further, the PIM application is preferably capable of sending and receiving data items via the wireless network 1401. PIM data items are continuously integrated, synchronized, and updated over the wireless network 1401 for corresponding device user data items that are stored or associated with the host computer system. It is preferable.

  Communication functions including data and voice communication are performed through the communication subsystem 1001, and possibly through the short-range communication subsystem. Communication subsystem 1001 includes a receiver 1500, a transmitter 1520, and one or more antennas 1540 and 1560. In addition, the communications subsystem 1001 also includes a processing module such as a digital signal processor (DSP) 1580 and a local oscillator (LO) 1601. The specific design and implementation of the communication subsystem 1001 depends on the communication network in which the mobile device 1000 intends to operate. For example, the mobile device 1000 includes a communication subsystem 1001 that operates on a mobile data communication network of Mobitex ™, Data TAC ™, or General Packet Radio Service (GPRS). Or designed to work with any of a variety of voice communication networks such as AMPS, TDMA, CDMA, WCDMA, PCS, GSM, EDGE, etc. . Other types of data and voice networks, whether separate or integrated, can be utilized with mobile device 1000. Mobile device 1000 may also conform to other communication standards such as 3GSM, 3GPP, UMTS, and the like.

  Network access requirements vary depending on the type of communication system. For example, in the Mobitex and DataTAC networks, mobile devices are registered on the network using a unique personal identification number or PIN associated with each device. On the other hand, in a GRPS network, network access is associated with a subscriber or device user. Thus, in order for a GPRS device to operate on a GPRS network, it requires a subscriber identity module, commonly referred to as a SIM card.

  If the necessary network registration or validation procedure is complete, the mobile device 1000 can transmit and receive communication signals over the communication network 1401. A signal from the communication network 1401 received by the antenna 1540 is sent to the receiver 1500, which performs signal amplification, frequency down-conversion, filtering, channel selection, and analog / digital conversion. obtain. Analog-to-digital conversion of the received signal allows the DSP 1580 to perform more complex communication functions such as demodulation and decoding. Similarly, the signal to be transmitted to network 1401 is processed (eg, modulated and encoded) by DSP 1580 and then provided to transmitter 1520 for digital-to-analog conversion, frequency up-conversion, filtering, amplification, and , Transmission to the communication network 1401 (s) via the antenna 1560 takes place.

  In addition to processing the communication signal, the DSP 1580 controls the receiver 1500 and the transmitter 1520. For example, the gain applied to the communication signal at receiver 1500 and transmitter 1520 may be adaptively controlled through an automatic gain control algorithm implemented in DSP 1580.

  In the data communication mode, received signals such as text messages or web page downloads are processed by the communication subsystem 1001 and input to the processing device 1800. The received signal is then further processed by processing device 1800 and output to display 1600 or some other auxiliary I / O device 1060. A device user may use keypad 1400 and / or any other auxiliary I / O device 1060 such as a touchpad, rocker switch, thumbwheel, or other type of input device, such as an email message, etc. Data items can also be created. The created data item may then be transmitted over the communication network 1401 via the communication subsystem 1001.

  In the voice communication mode, the overall operation of the device is substantially similar to the data communication mode, except that the received signal is output to the speaker 1100 and the signal for transmission is generated by the microphone 1120. Another voice or audio I / O subsystem, such as a voice message recording subsystem, may be implemented on the device 1000. In addition, the display 1600 may be used in a voice communication mode, eg, to display caller identification information, duration of a voice call, or other voice call related information. .

  Although the short-range communication subsystem allows communication between the mobile device 1000 and other nearby systems or devices, they need not be similar devices. For example, a short-range communication subsystem may include infrared devices and related circuitry and components, or Bluetooth ™ communication modules, and provide communication with similarly usable systems and devices.

  Those skilled in the art who have the benefit of the teachings presented in the foregoing description and the associated drawings will envision numerous variations and other embodiments of the invention. Therefore, it should be understood that the present invention is not limited to the specific embodiments disclosed, and such modifications and embodiments as described above should be included in the scope of the present invention.

FIG. 1 is a block diagram of a downlink advanced receiver performance (DARP) capable receiver utilizing an integrated space-time optimal filter, according to an illustrative embodiment. FIG. 2 is a more detailed block diagram of the integrated space-time optimal filter and multi-channel matched filter shown in FIG. 1 according to an exemplary embodiment. FIG. 2A is a block diagram of a method according to an exemplary embodiment. FIG. 3 is a graph illustrating the performance of a DARP-enabled receiver utilizing an integrated space-time optimal filter for various DARP test cases. FIG. 4 is a graph illustrating the performance of an integrated space-time optimal filter receiver according to an exemplary embodiment using additive white Gaussian noise (AWGN), with and without automatic switching means. It is. FIG. 5 is a graph showing the combined spatio-temporal optimal filter receiver performance according to an exemplary embodiment using DTS-5, with and without automatic switching means. FIG. 6 is a graph comparing the performance of single and multiple Viterbi equalizers using an 8-bit SD limiter for simulation, according to an exemplary embodiment. FIG. 7 is a graph illustrating the performance of an integrated space-time optimal filter receiver and a partially modified test case according to an exemplary embodiment. FIG. 8 is a schematic block diagram of an exemplary model wireless communication device that may be used in accordance with an exemplary embodiment. FIG. 9 is a table comparing three techniques for performing Cholesky decomposition, QR decomposition, and singular value decomposition (SVD) calculations according to this disclosure.

Claims (23)

A wireless communication device,
It has a wireless transmitter and a wireless receiver,
The wireless receiver includes a filter that reduces co-channel interference in the communication receiver, the filter comprising:
Multiple channels that filter n signal parts divided from a communication signal by jointly estimating spatio-temporal filter weights and multi-channel impulse responses (CIR) based on Cholesky decomposition and eigenvalue decomposition A spatio-temporal filter circuit;
A multi-channel matched filter circuit that receives a multi-channel signal from the multi-channel space-time filter circuit and has a filter response provided by a channel impulse response estimate from the space-time filter circuit. device.   The said filter is further connected to the said multi-channel space-time filter circuit, The virtual antenna which further divides | segments the said communication signal into the n-number real signal part and imaginary signal part which were sampled odd-numbered and even-numbered further. Wireless communication device.   The wireless communication device according to claim 1, wherein the multi-channel space-time filter circuit includes at least one multiplier that multiplies each signal unit by a separate space-time filter weight.   Each of the at least one multiplier includes a pair of the multipliers connected in parallel, and the multi-channel space-time filter circuit is connected to one input of the pair of multipliers. The wireless communication device according to claim 3, further comprising a separate delay circuit for the unit.   The wireless communication device of claim 4, wherein the communication signal includes a plurality of symbols, and each of the multiplier and delay circuit has a delay of about one symbol associated with the plurality of symbols.   The wireless communication device of claim 3, wherein the filter further comprises a separate summing circuit that sums the outputs of the multipliers for each channel.   The filter receives training sequence symbols and timing uncertainty data, and includes a space-time filter weight for the multi-channel space-time filter circuit and a multi-channel impulse response for the multi-channel matched filter circuit. The wireless communication device of claim 1, further comprising an integrated optimal filter weight and channel estimator.   The wireless communication device of claim 1, wherein the filter further includes an equalization circuit downstream of the multi-channel matched filter circuit. A wireless communication device,
It has a wireless transmitter and a wireless receiver,
The wireless receiver includes a filter system that reduces co-channel interference, the filter system comprising:
An integrated space-time filter,
Multiple channels that filter n signal parts divided from a communication signal by jointly estimating spatio-temporal filter weights and multi-channel impulse responses (CIR) based on Cholesky decomposition and eigenvalue decomposition A spatio-temporal filter circuit;
A multichannel matched filter circuit that receives a multichannel signal from the multichannel space-time filter circuit and has a filter response provided by a channel impulse response estimate from the space-time filter circuit. A spatio-temporal filter,
An alternative filter that operates when the interference level is less than a predetermined threshold, including a matched filter, a cross-correlation circuit, and a switching mechanism that switches the n signal units to the matched filter and the cross-correlation circuit; A wireless communication device including an alternative filter.   The filter system further includes a virtual antenna circuit connected to the multi-channel space-time filter circuit and dividing the communication signal into n real and imaginary signal parts sampled odd and even. A wireless communication device according to 1.   10. The wireless communication device of claim 9, wherein the multi-channel space-time filter circuit includes at least one multiplier that multiplies each signal unit by a separate space-time filter weight.   The at least one multiplier includes a pair of the multipliers connected in parallel, and the multi-channel space-time filter circuit is connected to an input of one of the pair of multipliers. The wireless communication device according to claim 11, further comprising a separate delay circuit for the unit.   13. The wireless communication device of claim 12, wherein the communication signal includes a plurality of symbols, and each of the multiplier and delay circuit has a delay of about one symbol associated with the plurality of symbols.   The wireless communication device of claim 11, wherein the filter system further comprises a separate summing circuit that sums the outputs of the multipliers for each channel.   The filter system receives training sequence symbols and timing uncertainty data, and a space-time filter weight for the multi-channel space-time filter circuit and a multi-channel impulse response for the multi-channel matched filter circuit 10. The wireless communication device of claim 9, further comprising an integrated optimal filter weight and channel estimator that generates   The wireless communication device of claim 9, wherein the filter system further includes an equalization circuit downstream of the multi-channel matched filter circuit. A method for reducing co-channel interference in a wireless communication device, comprising:
Providing the wireless communication device with a wireless receiver including a multi-channel space-time filter circuit and a multi-channel matched filter circuit;
Dividing the communication signal into n signal parts;
Filtering n signals in the multi-channel spatio-temporal filter circuit to collectively estimate spatio-temporal filter weights and multi-channel impulse response (CIR) based on Cholesky decomposition and eigenvalue decomposition When,
Receiving a multiple signal from the spatiotemporal filter circuit within a multichannel matched filter circuit having a filter response provided by a channel impulse response estimate from the spatiotemporal filter circuit.   18. The method of claim 17, wherein dividing comprises sampling the communication signal into even and odd samples and dividing the even and odd samples into a real signal portion and an imaginary signal portion.   The method of claim 17, further comprising summing and rescaling the output of the matched filter to a desired level.   The method of claim 19, further comprising equalizing the single channel signal after rescaling to a desired level.   The method of claim 17, further comprising filtering the n signal portions in an alternative filter if the interference level is less than a threshold.   The method of claim 17, further comprising multiplying each signal portion based on a spatio-temporal filter weight.   23. The method of claim 22, further comprising summing the signal portion for each channel after multiplication.

Images