CN100544328C

CN100544328C - An Improved Optimal Zero-Forcing Serial Interference Removal Detection Method

Info

Publication number: CN100544328C
Application number: CNB2005100804988A
Authority: CN
Inventors: 罗振东; 赵明; 刘思杨; 刘元安
Original assignee: Beijing University of Posts and Telecommunications
Current assignee: Shenzhen Tinno Wireless Technology Co Ltd
Priority date: 2005-07-07
Filing date: 2005-07-07
Publication date: 2009-09-23
Anticipated expiration: 2025-07-07
Also published as: CN1697431A

Abstract

The invention provides an improved optimal zero-forcing serial interference cancellation detection method, which is suitable for a multiple-input multiple-output (abbreviated as: MIMO) system or a communication system that can be modeled as a MIMO system. The basic principle is: use the zero-forcing weight matrix obtained in the previous stage of detection to accurately deduce the zero-forcing weight matrix and zero-forcing weight vector of the next stage of detection. Compared with the traditional optimal zero-forcing serial interference deletion detection method, the detection method provided by the invention effectively reduces the computational complexity without loss of performance.

Description

An Improved Optimal Zero-Forcing Serial Interference Removal Detection Method

技术领域 technical field

本发明涉及无线通信系统的信号检测技术，尤其涉及多输入多输出(简写为：MIMO)系统的信号检测技术。The present invention relates to a signal detection technology of a wireless communication system, in particular to a signal detection technology of a multiple-input multiple-output (abbreviated as: MIMO) system.

背景技术 Background technique

最新的研究显示：在无线衰落环境下采用多个发射天线和接收天线可以成倍提高无线通信系统的信道容量。这种采用多个收发天线的系统通常被称为多输入多输出(MIMO)系统。由于MIMO系统能够突破无线频率资源限制，有效提高系统频谱效率，因此被认为是未来高速无线通信系统的主要物理层技术之一。国内外学者对MIMO系统的相关技术已经做了大量的深入研究工作，其中MIMO系统的检测方法是一个重要研究热点。The latest research shows that using multiple transmitting antennas and receiving antennas in a wireless fading environment can double the channel capacity of a wireless communication system. Such systems employing multiple transmit and receive antennas are generally referred to as multiple-input multiple-output (MIMO) systems. Since the MIMO system can break through the limitation of wireless frequency resources and effectively improve the system spectrum efficiency, it is considered to be one of the main physical layer technologies for future high-speed wireless communication systems. Scholars at home and abroad have done a lot of in-depth research on the related technologies of MIMO systems, among which the detection method of MIMO systems is an important research hotspot.

1998年贝尔实验室的Golden和Foschini等人提出了一种最优迫零串行干扰删除的MIMO检测方法(或称为V-BLAST检测方法)。该方法按最大信噪比的排列顺序对发射的M路并行信号进行连续M级检测，在每级检测前删除已检测信号对未检测信号的干扰，每级检测时需要计算迫零加权向量，然后利用迫零加权向量恢复其对应的发射符号。贝尔实验室利用其搭建的实验平台证明了：在室内富反射环境下，当平均信噪比(SNR)为24-34dB时采用上述方法的MIMO系统的频谱效率可达到20-40bit/s/Hz；如此高的频谱效率利用传统的技术是无法达到的。In 1998, Golden and Foschini of Bell Laboratories proposed a MIMO detection method (or V-BLAST detection method) for optimal zero-forcing serial interference deletion. This method performs continuous M-level detection on the transmitted M-channel parallel signals according to the order of the maximum signal-to-noise ratio, and deletes the interference of the detected signal on the undetected signal before each level of detection. The zero-forcing weighted vector needs to be calculated for each level of detection. The zero-forcing weight vector is then used to recover its corresponding transmitted symbol. Bell Labs used its experimental platform to prove that: in an indoor rich reflection environment, when the average signal-to-noise ratio (SNR) is 24-34dB, the spectral efficiency of the MIMO system using the above method can reach 20-40bit/s/Hz ; Such a high spectral efficiency cannot be achieved using traditional techniques.

然而，最优迫零串行干扰删除检测方法需要进行M次求伪逆运算，在天线数比较多的情况下，该方法的复杂度很高，难以实现。能否降低该方法的运算复杂度是其能否实际应用的关键，也是摆在国内外研究人员面前的一个难题。However, the optimal zero-forcing serial interference cancellation detection method needs to perform M times of pseudo-inverse calculations. In the case of a large number of antennas, the complexity of this method is very high and it is difficult to implement. Whether the computational complexity of this method can be reduced is the key to its practical application, and it is also a difficult problem for researchers at home and abroad.

发明内容 Contents of the invention

针对最优迫零串行干扰删除检测方法具有很高的运算复杂度所带来的问题，本发明提供了一种大幅度降低运算复杂度的检测方法。Aiming at the problem caused by the high computational complexity of the optimal zero-forcing serial interference deletion detection method, the present invention provides a detection method that greatly reduces the computational complexity.

本发明提供一种改进的最优迫零串行干扰删除检测方法，将传统方法中计算迫零加权矩阵的方法改造成为逐级递推的方法，即利用第m-1级检测中得到的迫零加权矩阵G_m-1推导下一级(即：第m级)检测的迫零加权矩阵G_m和迫零加权向量g_m。The present invention provides an improved optimal zero-forcing serial interference deletion detection method, transforming the method of calculating the zero-forcing weighted matrix in the traditional method into a step-by-step recursive method, that is, using the force obtained in the m-1th level detection The zero-forcing weighting matrix G _m-1 deduces the zero-forcing weighting matrix G _m and the zero-forcing weighting vector g _m for detection at the next level (namely: the m-th level).

令c_m为第m级检测所对应的发射天线在H_m中对应的列序号，H₁＝H，H表示MIMO信道矩阵，H_m(2≤m≤M)为删除H_m-1的第c_m-1列所得到的矩阵；G′_m表示删除第m级检测的迫零加权矩阵G_m的第c_m行得到矩阵，

表示H_m的第c_m列，‖·‖表示矩阵的Frobenius范数或向量的模。则所述的递推方法的基本步骤包括：Let c _m be the column number corresponding to the transmitting antenna corresponding to the m-th level detection in H _m , H ₁ =H, H represents the MIMO channel matrix, H _m ( _2≤m≤M ) is the first The matrix obtained by c _m-1 column; G _′ _m means that the matrix obtained by deleting the c m row of the zero-forcing weighted matrix G _m detected by the mth level,

represents the c _m column of H _m , and ‖·‖ represents the Frobenius norm of the matrix or the modulus of the vector. The basic steps of the described recursive method then include:

计算第1级检测的迫零加权矩阵

，同时判别H是否列满秩，再将G₁的第c₁行进行共轭转置后得到第1级检测的迫零加权向量g₁。这里，表示H的伪逆。Calculate the zero-forcing weighting matrix for level 1 detection

, and at the same time determine whether H is full rank, and then perform conjugate transposition on row c ₁ of G ₁ to obtain the zero-forcing weighted vector g ₁ of the first-level detection. here, represents the pseudo-inverse of H.

如果H列满秩，则利用第一种递推方法计算后面的M-1级检测的迫零加权矩阵和迫零加权向量，即：If column H is full rank, use the first recursive method to calculate the zero-forcing weight matrix and zero-forcing weight vector for subsequent M-1 level detection, namely:

先根据如下递推公式计算G_m：First calculate G _m according to the following recursive formula:

${G G}_{m m} = = {G G}_{m m - - 11}^{' '} - - {| | | | {g g}_{m m - - 11} | | | |}^{22} {G G}_{m m - - 11}^{' '} {g g}_{m m - - 11} {g g}_{m m - - 11}^{H h};;$

再将G_m的第c_m行进行共轭转置后得到第m级检测的迫零加权向量g_m。Then perform conjugate transposition on the c _mth row of G _m to obtain the zero-forcing weighted vector g _m of the m-th level detection.

如果H列缺秩，则利用第二种递推方法计算后面的M-1级检测的迫零加权矩阵和迫零加权向量，即：If column H lacks rank, use the second recursive method to calculate the zero-forcing weighted matrix and zero-forcing weighted vector for subsequent M-1 level detection, namely:

根据如下递推公式计算G_m：Calculate G _m according to the following recursive formula:

先计算 $α_{m - 1} = 1 - g_{m - 1}^{H} h_{c_{m - 1}};$ calculate first $α_{m - 1} = 1 - g_{m - 1}^{h} h_{c_{m - 1}};$

如果α_m-1＝0，则通过下式计算G_m：If α _m-1 =0, G _m is calculated by the following formula:

${G G}_{m m} = = {G G}_{m m - - 11}^{' '} - - {| | | | {g g}_{m m - - 11} | | | |}^{- - 22} {G G}_{m m - - 11}^{' '} {g g}_{m m - - 11} {g g}_{m m - - 11}^{H h};;$

如果α_m-1≠0，则通过下式计算G_m：If α _m-1 ≠0, G _m is calculated by the following formula:

${G G}_{m m} = = {G G}_{m m - - 11}^{' '} + + \frac{11}{{α α}_{m m - - 11}} {G G}_{m m - - 11}^{' '} {h h}_{{c c}_{m m - - 11}} {g g}_{m m - - 11}^{H h};;$

本发明的有益效果在于，所提出的改进的最优迫零串行干扰删除检测方法采用低复杂度的递推方法，利用前一级检测的运算结果来递推计算后一级检测的迫零加权矩阵和迫零加权向量，避免了复杂度很高的直接求矩阵伪逆运算。与传统方法相比，该方法在保证检测性能不变的前提下，大幅度降低了最优迫零串行干扰删除检测方法的运算复杂度。The beneficial effect of the present invention is that the proposed improved optimal zero-forcing serial interference deletion detection method adopts a low-complexity recursive method, and uses the calculation results of the previous stage of detection to recursively calculate the zero-forcing of the subsequent stage of detection. The weighted matrix and the zero-forcing weighted vector avoid the high-complexity direct matrix pseudo-inverse operation. Compared with the traditional method, this method greatly reduces the computational complexity of the optimal zero-forcing serial interference removal detection method under the premise of ensuring the same detection performance.

附图说明 Description of drawings

图1示出了MIMO系统的原理框图；Fig. 1 shows a functional block diagram of a MIMO system;

图2示出了最优迫零干扰删除检测方法的流程图；Fig. 2 shows the flowchart of optimal zero-forcing interference deletion detection method;

图3示出了采用Greville方法计算第1级迫零加权矩阵以及判别MIMO信道矩阵是否列满秩的流程图；FIG. 3 shows a flow chart of calculating the first-level zero-forcing weighted matrix and judging whether the MIMO channel matrix is full rank by using the Greville method;

图4示出了检测第p_m路信号的流程图；Fig. 4 shows the flow chart of detecting the p _m road signal;

图5示出了计算第m级检测的迫零加权矩阵的流程图；Fig. 5 shows the flow chart of calculating the zero-forcing weighted matrix of the m-level detection;

图6示出了本发明提供的改进的最优迫零干扰删除检测方法与传统方法的性能比较结果。Fig. 6 shows the performance comparison results between the improved optimal zero-forcing interference deletion detection method provided by the present invention and the traditional method.

具体实施方式 Detailed ways

下面通过附图和实施例对本发明进行详细阐述。The present invention will be described in detail below through the accompanying drawings and examples.

本发明的检测方法适用于MIMO系统，或是能够建模为MIMO系统的其它通信系统。例如，本发明可直接用在多入多出正交频分复用(MIMO-OFDM)系统的任意一个子载波上，也可用于码分多址系统的多用户检测。The detection method of the present invention is applicable to MIMO systems, or other communication systems that can be modeled as MIMO systems. For example, the present invention can be directly used on any sub-carrier of a multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) system, and can also be used for multi-user detection in a code division multiple access system.

图1示出了MIMO系统的原理框图。在发射端，数据比特首先被映射成为信号星座中的信号，经过串并变换后形成多路并行的基带信号，然后经过调制后分别从多根不同的天线同时发射出去；经过无线信道衰落后，来自不同发射天线的信号与噪声叠加后被多根天线同时接收，经过解调后生成多路并行基带信号，MIMO检测器利用信道估计器产生的信道状态信息从基带信号中恢复出原始数据。实际系统中，数据比特在映射之前可以先经过编码和交织，相应的在接收机输出数据之前要经过解交织和译码。该系统的基带信号输入输出关系可以被表示为：Fig. 1 shows a functional block diagram of a MIMO system. At the transmitting end, the data bits are first mapped into signals in the signal constellation, and after serial-to-parallel conversion, multiple parallel baseband signals are formed, and then modulated and transmitted from multiple different antennas at the same time; after wireless channel fading, Signals and noise from different transmitting antennas are superimposed and received by multiple antennas at the same time. After demodulation, multiple parallel baseband signals are generated. The MIMO detector uses the channel state information generated by the channel estimator to recover the original data from the baseband signals. In an actual system, the data bits can be coded and interleaved before being mapped, and correspondingly deinterleaved and decoded before the receiver outputs the data. The baseband signal input-output relationship of the system can be expressed as:

y＝Hx+εy=Hx+ε

上式中，x＝[x₁ x₂…x_M]^T表示发射信号向量，M表示发射天线数目，[·]^T表示矩阵或向量的转置，x_m表示从第m根发射天线发射的信号；ε＝[ε₁ ε₂…ε_N]^T表示噪声向量，N表示接收天线数目，ε_n表示第n根接收天线接收到的噪声；y＝[y₁ y₂…y_N]^T表示接收信号向量，y_n表示第n根接收天线接收到的信号；H是N×M维的矩阵，表示MIMO信道矩阵，其第n行第m列的元素h_n，m表示从第m根发射天线到第n根接收天线的基带信道衰落因子，在进行MIMO检测处理之前，首先要通过信道估计器获得信道矩阵的估计值(为了方便描述，文中把MIMO信道矩阵的估计值仍记为H)。本发明涉及图1所示系统的MIMO检测器部分。In the above formula, x=[x ₁ x ₂ ... x _M ] ^T represents the transmitted signal vector, M represents the number of transmitting antennas, [ ] ^T represents the transposition of the matrix or vector, and x _m represents the signal transmitted from the mth transmitting antenna signal; ε=[ε ₁ ε ₂ ...ε _N ] ^T represents the noise vector, N represents the number of receiving antennas, ε _n represents the noise received by the nth receiving antenna; y=[y ₁ y ₂ ... y _N ] ^T represents Received signal vector, y _n represents the signal received by the nth receiving antenna; H is an N×M dimensional matrix, which represents the MIMO channel matrix, and the element h _n in the nth row and the mth column of it represents the signal transmitted from the mth root The baseband channel fading factor from the antenna to the nth receiving antenna, before performing MIMO detection processing, the estimated value of the channel matrix must first be obtained through the channel estimator (for the convenience of description, the estimated value of the MIMO channel matrix is still recorded as H in this paper) . The present invention relates to the MIMO detector portion of the system shown in FIG. 1 .

图2示出了最优迫零串行干扰删除检测方法的流程图，用于图1中的MIMO检测器部分。这里将第i个发射天线发射的信号简称为第i路信号，i＝1，2，…，M。所示的流程在检测开始201后进入步骤203进行第1级检测初始化，包括初始化第1级检测的接收向量y₁、等效信道矩阵H₁和序号向量f₁，其具体步骤如下：FIG. 2 shows a flowchart of an optimal zero-forcing serial interference cancellation detection method for the MIMO detector part in FIG. 1 . Here, the signal transmitted by the i-th transmitting antenna is referred to as the i-th signal for short, i=1, 2, . . . , M. The flow shown enters step 203 after the detection starts 201 to initialize the first-level detection, including initializing the receiving vector y ₁ , the equivalent channel matrix H ₁ and the serial number vector f ₁ of the first-level detection. The specific steps are as follows:

(1)第1级检测的接收向量y₁等于接收向量y，即y₁＝y；(1) The receiving vector y ₁ of the first-level detection is equal to the receiving vector y, that is, y ₁ =y;

(2)第1级检测的等效信道矩阵H₁就等于MIMO信道矩阵H，即H₁＝H；(2) The equivalent channel matrix H ₁ of the first-level detection is equal to the MIMO channel matrix H, that is, H ₁ =H;

(3)第1级检测的序号向量f₁的元素为自然数1至M，且为升序排列，即f₁＝[1 2…M]^T。(3) The elements of the serial number vector f ₁ detected at the first level are natural numbers 1 to M, and are arranged in ascending order, that is, f ₁ =[1 2 . . . M] ^T .

此后，在步骤205，检测第p₁路信号，得到

其具体步骤如下：Thereafter, in step 205, the _p1th signal is detected to obtain

The specific steps are as follows:

(1)计算迫零加权矩阵G₁，即

其中

表示矩阵或向量的伪逆；并且确定列满秩标志位R的值，即当H₁为列满秩时R＝1，否则R＝0；(1) Calculate the zero-forcing weighted matrix G ₁ , namely

in

Representing the pseudo-inverse of matrix or vector; and determining the value of the full-rank flag bit R, that is, R=1 when H ₁ is full-rank, otherwise R=0;

(2)取出矩阵G₁中模值最小的行所对应的行序号c₁；(2) take out the row sequence number _c1 corresponding to the row with the smallest modulus value in the matrix _G1 ;

(3)取出f₁的第c₁个元素得到第1级检测对应的发射天线序号p₁。(3) Take out the c ₁ th element of f ₁ to obtain the serial number p ₁ of the transmitting antenna corresponding to the first-level detection.

(4)对G₁的第c₁行进行共轭转置得到第1级迫零加权向量g₁，并利用g₁对接收向量y₁进行迫零加权，得到第p₁路信号的判决统计量，即 $z_{p_{1}} = g_{1}^{H} y_{1},$ 其中，[·]H表示矩阵或向量的共轭转置；(4) Perform conjugate transposition on row c ₁ of G ₁ to obtain the first-level zero-forcing weighted vector g ₁ , and use g ₁ to perform zero-forcing weighting on the received vector y ₁ to obtain the decision statistics of the p _1- th signal quantity ,Right now $z_{p_{1}} = g_{1}^{h} {they}_{1},$ Among them, [ ] H represents the conjugate transpose of matrix or vector;

(5)对判决统计量进行硬判决，得到第p₁路信号的判决值

。(5) For decision statistics Make a hard decision to get the decision value of the _p1th signal

.

在步骤207，对计数变量m赋初始值，即m＝2。In step 207, an initial value is assigned to the counting variable m, ie m=2.

在步骤209，对计数变量m进行判断，如果m≤M成立，则进入步骤211，反之则进入步骤217。In step 209, the counting variable m is judged, if m≤M is established, then enter step 211, otherwise, enter step 217.

在步骤211，第m级检测初始化，包括初始化第m级检测的等效接收向量y_m、等效信道矩阵H_m和序号向量f_m，其具体步骤如下：In step 211, the detection initialization of the mth stage includes initializing the equivalent reception vector y _m , the equivalent channel matrix H _m and the serial number vector f _m of the mth stage detection, and the specific steps are as follows:

(1)从y_m-1中删除

对其它发射信号的干扰得到第m级检测的等效接收向量y_m，即

y_{m} = y_{m - 1} - h_{c_{m - 1}} {\hat{x}}_{p_{m - 1}} .

这里，

表示H_m-1的第c_m-1列。(1) delete from y _m-1

The interference to other transmitted signals obtains the equivalent receiving vector y _m of the m-th stage detection, namely

{the y}_{m} = {the y}_{m - 1} - h_{c_{m - 1}} {\hat{x}}_{p_{m - 1}} .

here,

Indicates the c _m-1 column of H _m-1 .

(2)删除第m-1级检测的等效信道矩阵H_m-1的第c_m-1列得到第m级检测的等效信道矩阵H_m。(2) Delete the c m-1 column of the equivalent channel matrix H _m-1 of the _m-1 detection level to obtain the m-th level detection equivalent channel matrix H _m .

(3)删除第m-1级检测的序号向量f_m-1的第c_m-1个元素得到第m级检测的序号向量f_m。(3) Delete the c m _-1th element of the sequence number vector f _m-1 of the m-1th level detection to obtain the m-th level detection sequence number vector f _m .

在步骤213，检测第p_m路信号，得到判决值

，图4给出了这一步骤的具体流程。In step 213, detect the p _m- th channel signal, and obtain the decision value

, Figure 4 shows the specific flow of this step.

在步骤215，计数变量m加1，即m＝m+1。In step 215, the counting variable m is incremented by 1, ie m=m+1.

然后，该流程在步骤217退出。Then, the process exits at step 217 .

图3示出了采用Greville方法计算第1级迫零加权矩阵以及判别MIMO信道矩阵H是否列满秩的流程图，用于实现图2中步骤205的(1)步骤。记A_k为H的前k列构成的子矩阵，a_k为H的第k列，k＝1，2，…，M。图3中的流程是在计算矩阵伪逆的开始进入步骤301。接着，步骤303，计算A₁的伪逆，即

这里，‖·‖表示矩阵的Frobenius范数或向量的模。在步骤305，对计数变量k赋初始值，即k＝2。在步骤307，对计数变量k进行判断，如果k≤M成立，则进入步骤309，反之则进入步骤317。在步骤309，计算d_k和q_k，即和q_k＝a_k-A_k-1d_k。然后，在步骤311，根据q_k、d_k和A_k-1计算即：FIG. 3 shows a flow chart of calculating the first-level zero-forcing weighted matrix and judging whether the MIMO channel matrix H has a full rank using the Greville method, which is used to implement step (1) of step 205 in FIG. 2 . Note that A _k is the sub-matrix formed by the first k columns of H, a _k is the kth column of H, and k=1, 2, . . . , M. The flow in FIG. 3 enters step 301 at the beginning of calculating the pseudoinverse of the matrix. Next, step 303, calculate the pseudo-inverse of A ₁ , namely

Here, ‖·‖ denotes the Frobenius norm of a matrix or the modulus of a vector. In step 305, an initial value is assigned to the counting variable k, ie k=2. In step 307, the counting variable k is judged, if k≤M is established, then go to step 309, otherwise go to step 317. In step 309, d _k and q _k are calculated, namely and q _k = a _k - A _{k - 1} d _k . Then, in step 311, calculate according to q _k , d _k and A _k-1 Right now:

实际中，由于计算精度的限制，当‖q_k‖的值小于某个极小的正数(即：‖q_k‖<γ，这里γ为一个极小的正数，其大小由实际的计算精度确定)时就可以认为‖q_k‖＝0，反之则认为‖q_k‖≠0。In practice, due to the limitation of calculation accuracy, when the value of ‖q _k ‖ is less than a certain very small positive number (ie: ‖q _k ‖<γ, where γ is a very small positive number, its size is determined by the actual calculation When the accuracy is determined), it can be considered that ‖q _k ‖=0, otherwise it can be considered that ‖q _k ‖≠0.

在步骤313，按下式递推计算A_k的伪逆：In step 313, the pseudo-inverse of A _k is recursively calculated as follows:

在步骤315，计数变量k加1，即k＝k+1。In step 315, the counting variable k is incremented by 1, ie k=k+1.

在步骤317，根据‖q_M‖确定满秩标志位R的值，若‖q_M‖≠0说明矩阵H为列满秩，即：R＝1；若‖q_M‖＝0说明矩阵H为列缺秩，即：R＝0。In step 317, determine the value of the full-rank flag R according to ‖q _M ‖, if ‖q _M ‖≠0 shows that the matrix H is full rank, that is: R=1; if ‖q _M ‖=0 shows that the matrix H is Column missing rank, ie: R=0.

然后，该流程在步骤319退出。Then, the process exits at step 319 .

图4示出了检测第p_m路信号以得到

的具体流程，用于实现图2中的步骤213。图4中的流程是在完成图2中的步骤211后进入步骤401。接着，在步骤403，计算迫零加权矩阵G_m。此后，在步骤405，取出矩阵G_m中模值最小的行所对应的行序号c_m。在步骤407，取出f_m的第c_m个元素得到第m级检测对应的发射天线序号p_m。在步骤409，对G_m的第c_m行取共轭转置得到第m级迫零加权向量g_m，并利用迫零向量g_m对接收向量y_m进行线性加权，得到第p_m路信号的判决统计量

即

z_{p_{m}} = g_{m}^{H} y_{m} .

在步骤411，对判决统计量

进行硬判决，得到第p_m路信号的判决值

然后，该流程在步骤413退出。Figure 4 shows the detection of the p _m signal to obtain

The specific process is used to implement step 213 in FIG. 2 . The process in FIG. 4 enters into step 401 after completing step 211 in FIG. 2 . Next, in step 403, the zero-forcing weighting matrix G _m is calculated. Thereafter, in step 405, the row number c _m corresponding to the row with the smallest modulus value in the matrix G _m is taken out. In step 407, the c _m th element of f _m is taken out to obtain the transmit antenna serial number p _m corresponding to the mth stage detection. In step 409, take the conjugate transpose of the c _m row of G _m to obtain the mth level zero-forcing weighted vector g _m , and use the zero-forcing vector g _m to linearly weight the receiving vector y _m to obtain the p _m- th signal decision statistic

Right now

z_{p_{m}} = g_{m}^{h} {the y}_{m} .

In step 411, the decision statistics

Make a hard decision to obtain the decision value of the p _mth channel signal

Then, the process exits at step 413 .

图5示出了计算第m级检测的迫零加权矩阵G_m的具体流程(2≤m≤M)，用于实现图4中的步骤403。该流程开始后进入步骤503，对计数变量m进行判断，如果m≤M-1成立，则进入步骤505，反之(即：当m＝M时)则进入步骤517。FIG. 5 shows a specific flow (2≤m≤M) of calculating the zero-forcing weighted matrix G _m of the m-th level detection, which is used to realize step 403 in FIG. 4 . After the process starts, enter step 503 to judge the counting variable m, if m≤M-1 is established, then enter step 505, otherwise (ie: when m=M) then enter step 517.

在步骤505，删除第m-1级检测的迫零加权矩阵G_m-1的第c_m-1行得到

In step 505, delete the c m-1 row of the zero-forcing weighting matrix G _m-1 _detected at the m-1th level to obtain

在步骤507，对H₁的满秩标志位R进行判断，如果R＝1成立，则进入步骤513，反之则进入步骤509；理论上来说，只要H的元素间具有一定的独立性(即：不完全相关)，H列满秩的概率就为1，然而在实际中由于信道的不理想、计算精度有限等因素可能会导致H₁列缺秩的情况。In step 507, the full rank flag R of _H1 is judged, if R=1 is set up, then enter step 513, otherwise then enter step 509; Theoretically speaking, as long as there is certain independence between the elements of H (that is: Incomplete correlation), the probability that the H column is full rank is 1, but in practice, due to factors such as unsatisfactory channels and limited calculation accuracy, the H ₁ column may lack a rank.

在步骤509，判断α_m-1的值是否为0，如果α_m-1＝0成立，则进入步骤511，反之则进入步骤513。这里，α_m-1的计算公式为： $α_{m - 1} = 1 - g_{m - 1}^{H} h_{c_{m - 1}} .$ 实际中，由于计算精度的限制，当α_m-1的绝对值小于某个非常小的正数(即：|α_m-1|<γ，其中|·|表示绝对值，γ为一个非常小的正数，其大小由实际的计算精度确定)时就可以认为α_m-1＝0；反之，如果|α_m-1|≥γ则认为α_m-1≠0；In step 509 , it is judged whether the value of α _m-1 is 0, if α _m-1 =0, go to step 511 , otherwise go to step 513 . Here, the calculation formula of α _m-1 is: $α_{m - 1} = 1 - g_{m - 1}^{h} h_{c_{m - 1}} .$ In practice, due to the limitation of calculation accuracy, when the absolute value of α _m-1 is less than a very small positive number (ie: |α _m-1 |<γ, where |·| represents the absolute value, γ is a very small α _m-1 = 0; on the contrary, if |α _m-1 |≥γ, then α _m-1 ≠0;

在步骤511，利用公式(1)递推计算G_m，公式(1)如下：In step 511, G _m is recursively calculated using formula (1), and formula (1) is as follows:

${G G}_{m m} = = {G G}_{m m - - 11}^{' '} - - {| | | | {g g}_{m m - - 11} | | | |}^{22} {G G}_{m m - - 11}^{' '} {g g}_{m m - - 11} {g g}_{m m - - 11}^{H h} - - - - - - ((11))$

在步骤513，利用公式(2)递推计算G_m，公式(2)如下：In step 513, G _m is recursively calculated using formula (2), and formula (2) is as follows:

${G G}_{m m} = = {G G}_{m m - - 11}^{' '} + + \frac{11}{{α α}_{m m - - 11}} {G G}_{m m - - 11}^{' '} {h h}_{{c c}_{m m - - 11}} {g g}_{m m - - 11}^{H h} - - - - - - ((22))$

在步骤515，直接计算第M级检测的迫零加权矩阵

注：任意的迫零加权矩阵G_m(2≤m≤M)均可由G_m-1递推得到；然而当m＝M，

H_{M} = h_{c_{M}}

为一列向量，此时所述的递推方法可简化为直接计算第M级迫零加权向量，即：

g_{M} = G_{M}^{H} = {| | h_{c_{M}} | |}^{- 2} h_{c_{M}} .

In step 515, directly calculate the zero-forcing weighting matrix of the Mth level detection

Note: Any zero-forcing weight matrix G _m (2≤m≤M) can be recursively obtained from G _m-1 ; however when m=M,

h_{m} = h_{c_{m}}

is a column of vectors, the recursive method described at this time can be simplified to directly calculate the Mth level zero-forcing weighted vector, namely:

g_{m} = G_{m}^{h} = {| | h_{c_{m}} | |}^{- 2} h_{c_{m}} .

然后，该流程在步骤517退出。Then, the process exits at step 517 .

图6示出了本发明提供的方法与传统方法的性能比较结果。横坐标表示信噪比(SNR)，纵坐标表示的是误比特率(BER)，该系统的发射天线和接收天线数目均为4，信道是独立同分布的MIMO平坦瑞利衰落信道，所采用的调制方式是16QAM。仿真结果显示本发明提供的方法与传统方法性能完全相同。Fig. 6 shows the performance comparison results of the method provided by the present invention and the traditional method. The abscissa represents the signal-to-noise ratio (SNR), and the ordinate represents the bit error rate (BER). The number of transmitting and receiving antennas in this system is 4, and the channel is an independent and identically distributed MIMO flat Rayleigh fading channel. The modulation method is 16QAM. Simulation results show that the performance of the method provided by the invention is exactly the same as that of the traditional method.

下面以发射天线数目和接收天线数目相等(M＝N)为例，简要分析本发明提供的方法的运算复杂度。这里以一次复数乘法的计算量为方法复杂度的单位，忽略加减法、比较、选择等相对简单的处理，只计算乘除法的复杂度。传统最优迫零串行干扰删除检测方法复杂度达到M⁴的水平；由于采用了简单的递推方法计算迫零加权矩阵，本发明提供的方法复杂度仅为M³的水平。显然，本发明提供的方法可以大幅度降低最优迫零串行干扰删除检测方法的运算复杂度。Taking the number of transmitting antennas equal to the number of receiving antennas (M=N) as an example, the computational complexity of the method provided by the present invention is briefly analyzed below. Here, the calculation amount of complex number multiplication is used as the unit of method complexity, and relatively simple processing such as addition and subtraction, comparison, and selection is ignored, and only the complexity of multiplication and division is calculated. The complexity of the traditional optimal zero-forcing serial interference deletion detection method reaches the level of ^M4 ; since a simple recursive method is used to calculate the zero-forcing weighted matrix, the complexity of the method provided by the present invention is only at the level of ^M3 . Obviously, the method provided by the present invention can greatly reduce the computational complexity of the optimal zero-forcing serial interference cancellation detection method.

Claims

1. An improved optimal zero-forcing serial interference deletion detection method, which performs continuous M-level detection on the transmitted M-channel parallel signals in a certain order, and deletes the detected signal and the undetected signal before each level of detection interference, the zero-forcing weighted vector needs to be calculated at each level of detection, and then the corresponding transmitted symbol is recovered by using the zero-forcing weighted vector; the method is characterized in that the zero-forcing weighted matrix and the zero-forcing weighted vector required for the m-th level detection are Obtained by recursion from the zero-forcing weighted matrix generated during the m-1th level detection, where m=2, 3,..., M, the basic steps of recursion are:

First, calculate the zero-forcing weighting matrix for level 1 detection

, and at the same time determine whether H is full rank, and then perform conjugate transposition on row c ₁ of G ₁ to obtain the zero-forcing weighted vector g ₁ of the first-level detection, where H represents the MIMO channel matrix,

Indicates its pseudo-inverse, c ₁ is the column number corresponding to the transmit antenna corresponding to the first-level detection in H;

Then, judge whether H is full rank; if H is full rank, use the first recursive method to calculate the subsequent zero-forcing weight matrix and zero-forcing weight vector for M-1 level detection; if H lacks rank, use The second recursive method calculates the zero-forcing weight matrix and zero-forcing weight vector for subsequent M-1 level detection.

2. The detection method according to claim 1, wherein the steps of using the first recursive method to calculate the zero-forcing weighted matrix _G and the zero-forcing weighted vector _g for the m-level detection are as follows:

Delete the c m-1th row of the zero-forcing weighted matrix G _m-1 _of the m-1th level detection to get the matrix

Wherein, c _l is the column sequence number corresponding to the transmitting antenna corresponding to the l-level detection in H _l , l=1, 2, ..., M, when l=1, H _l =H, when l=2, 3,..., _H1 is the matrix obtained by deleting the column _C1-1 of _H1-1 during M;

Calculate the zero-forcing weighted matrix G _m of the m-th level detection according to the following recursive formula:

{G G}_{m m} = = {G G}_{m m - - 11}^{' '} - - {| | | | {g g}_{m m - - 11} | | | |}^{- - 22} {G G}_{m m - - 11}^{' '} {g g}_{m m - - 11} {g g}_{m m - - 11}^{H h};;

The zero-forcing weighted vector g _m of the m-th level detection is obtained after the conjugate transposition of the cm-th row _{of G m} _.

3. The detection method according to claim 1, wherein the steps of using the second recursive method to calculate the zero-forcing weighted matrix _G and the zero-forcing weighted vector _g for the m-level detection are as follows:

calculate

α_{m - 1} = 1 - g_{m - 1}^{h} h_{c_{m - 1}},

in, Indicates the c _l column of H _l , l=1, 2,..., M;

If α _m-1 =0, G _m is calculated by the following formula:

{G G}_{m m} = = {G G}_{m m - - 11}^{' '} - - {| | | | {g g}_{m m - - 11} | | | |}^{- - 22} {G G}_{m m - - 11}^{' '} {g g}_{m m - - 11} {g g}_{m m - - 11}^{H h};;

If α _m-1 ≠0, G _m is calculated by the following formula:

{G G}_{m m} = = {G G}_{m m - - 11}^{' '} + + \frac{11}{{α α}_{m m - - 11}} {G G}_{m m - - 11}^{' '} {h h}_{{c c}_{m m - - 11}} {g g}_{m m - - 11}^{H h};;

4. The detection method according to claim 1, 2 or 3, characterized in that the method for calculating the zero-forcing weighted vector for the Mth level detection is:

g_{m} = {| | h_{c_{m}} | |}^{- 2} h_{c_{m}} .

5. The detection method according to claim 3, characterized in that, when |α _m-1 |<γ, it is judged that α _m-1 = 0; when |α _m-1 |≥γ, it is judged that α _m-1 ≠0; here, γ is a very small positive number set according to the actual calculation accuracy.

6. The detection method according to claim 1, characterized in that the detection method is applicable to MIMO systems and other communication systems that can be modeled as MIMO systems.