WO2011140833A1 - 一种盲均衡器及盲均衡处理方法 - Google Patents
一种盲均衡器及盲均衡处理方法 Download PDFInfo
- Publication number
- WO2011140833A1 WO2011140833A1 PCT/CN2011/070029 CN2011070029W WO2011140833A1 WO 2011140833 A1 WO2011140833 A1 WO 2011140833A1 CN 2011070029 W CN2011070029 W CN 2011070029W WO 2011140833 A1 WO2011140833 A1 WO 2011140833A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- weight
- weight vector
- update
- vector
- module
- Prior art date
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
- H04L25/03006—Arrangements for removing intersymbol interference
- H04L25/03012—Arrangements for removing intersymbol interference operating in the time domain
- H04L25/03019—Arrangements for removing intersymbol interference operating in the time domain adaptive, i.e. capable of adjustment during data reception
- H04L25/03057—Arrangements for removing intersymbol interference operating in the time domain adaptive, i.e. capable of adjustment during data reception with a recursive structure
- H04L25/0307—Arrangements for removing intersymbol interference operating in the time domain adaptive, i.e. capable of adjustment during data reception with a recursive structure using blind adaptation
Definitions
- the invention belongs to the field of communications. In particular, it relates to a blind equalizer and a blind equalization processing method.
- Equalization is a technique used to reduce distortion and compensate for signal loss. Since blind equalization does not require the use of training sequences, system bandwidth efficiency can be improved. In addition, for some communication systems, it is difficult to obtain a correct training sequence at the receiving end when the receiving end loses synchronization or the like. Blind equalization provides a practical means of eliminating the harmful effects of the channel.
- CMA Constant Modulus Algorithm
- the weight update equation for the CMA equalizer uses the steepest descent gradient algorithm:
- the equalizer has N tap coefficients, also called weights; Representing a weight vector; Is the (positive) iteration step, which controls the convergence speed; the equalizer input signal vector is , the output signal is ; Cost function middle Obtaining the partial derivative, that is, ; upper corner Indicates that the conjugate operation is taken.
- the weights are adjusted according to the CMA weight update equation to reduce the error term. That is, the deviation between the output of the equalizer and the constant mode is reduced until the equalizer converges.
- the (positive) iteration step size of the DSE-CMA algorithm which controls the convergence speed of the DSE-CMA algorithm; Representing a conjugate operation; Indicates the error term of DSE-CMA, . among them, Express output signal ; Error term for CMA , That is, the random signal added before quantization, Indicates the amplitude of the jitter, which is a normal number; , with Uniformly distributed And subject to independent and identically distributed jitter random signals. , sgn(.) represents a symbolic operation, thus simplifying the calculation.
- the DSE-CMA algorithm reduces the error term by adjusting the weights until the equalizer converges.
- the DSE-CMA algorithm simplifies calculations by turning a large number of update multiplications into symbolic operations.
- DSE-CMA The purpose of the algorithm is to maintain the robustness of the CMA while reducing complexity.
- the convergence rate of the DSE-CMA algorithm is slower and the steady-state performance is not good enough.
- the object of the present invention is to solve the problem of slow convergence speed and poor steady state performance existing in the existing blind equalizer technology.
- a blind equalizer includes a weight update unit and a filter
- the weight update unit includes:
- a first weight update module configured to update the first weight vector by using a jitter symbol error-constant modulus algorithm
- a second weight update module for using a maximum a posteriori probability theory as a decision basis, using a steepest descent gradient algorithm, Maximizing the logarithm of the local posterior probability density function, updating the second weight vector;
- a weight combination module configured to merge the updated first weight vector and second weight vector
- the filter is configured to perform equalization processing on the received signal vector according to the weight vector obtained by combining the weight combining modules. .
- Another object of the present invention is to provide a problem aimed at solving the prior art, and to provide a
- the blind equalization processing method includes the following steps:
- the steepest descending gradient algorithm is used to make the local posterior probability density function.
- the logarithmic value is the largest, and the second weight vector is updated;
- the received signal vector is equalized and output.
- the density function has the largest logarithmic value, and the weight vector is updated. Then, the received signal vector is equalized and output according to the updated weight vector, and a blind equalizer is realized, which can improve the convergence speed and steady state performance.
- FIG. 1 is a structural block diagram of a blind equalization system provided by the prior art
- FIG. 2 is a structural block diagram of a blind equalizer according to an embodiment of the present invention.
- FIG. 3 is a block diagram of a blind equalization system in which a blind equalizer is provided according to an embodiment of the present invention
- FIG. 5 is a diagram of inter-symbol interference (ISI) simulation provided by an embodiment of the present invention
- ISI inter-symbol interference
- FIG. 7 is a flowchart of an implementation process of a blind equalization processing method according to an embodiment of the present invention.
- the error term of the first weight update module is minimized and the local a posteriori probability used by the second weight update module is adopted.
- the density function has the largest logarithmic value, and the weight vector is updated. Then, the received signal vector is equalized and output according to the updated weight vector, and a blind equalizer is realized.
- FIG 2 The structure of the blind equalizer provided by the embodiment of the present invention is shown, and only parts related to the embodiment of the present invention are shown for convenience of description.
- the blind equalizer can be used in a receiver of a wireless mobile communication system, and can be a software unit, a hardware unit or a combination of hardware and software running in the receivers, wherein:
- the weight update unit 201 according to the steepest descent gradient algorithm, in order to make the first weight update module 201
- the error term is minimized and the second weight update module 202 employs a local posterior probability density function (p.d.f.
- the logarithmic value is the largest, updating the weight vector;
- the filter 202 performs equalization processing on the received signal vector according to the weight vector updated by the weight update unit 201.
- the following method is used to equalize the received signal vector:
- the upper corner ' 'Indicating matrix transpose The updated weight vector for the weight update unit 201, For the received signal vector, The signal that is output after equalization processing.
- the weight update unit 201 includes a first weight update module 2011 and a second weight update module. 2012 and weights merge module 2013 :
- First weight update module 2011, using DSE-CMA The algorithm updates the first weight vector, and specifically updates the first weight vector by using an iterative formula as follows:
- the (positive) iteration step size for weight vector update using the DSE-CMA algorithm which controls the convergence speed of the weight vector update using the DSE-CMA algorithm.
- Is the output signal of the blind equalizer (ie, filter 202) For the received signal vector, the upper corner Representing a conjugate operation; Indicates the error term used to update the weight vector using the DSE-CMA algorithm.
- ESE Excess Mean Square Error
- the second weight update module 2012 uses the maximum posterior probability theory as the basis for the decision, using the steepest descent gradient algorithm, The logarithm of the local posterior p.d.f. is maximized, so that the second weight vector is continuously updated until convergence, and the second weight vector is updated by using the following iterative formula:
- the weight vector of the moment Represents the noise variance, which is related to the spread of the channel.
- a partial posterior pdf representing the output of the blind equalizer (ie, filter 202).
- the second weight update module 2012 uses the steepest descent gradient algorithm to adjust By reducing The value thus maximizes the logarithm of the partial posterior pdf.
- the weight combination module 2013 merges the first weight vector after the first weight update unit is updated And the second weight update vector update unit update second weight vector Specifically, the combination of weight vectors is performed by the following formula:
- the number of taps of the first weight update module 2011 and the second weight update module 2012 is ; , , with The selection can be determined based on actual experience. After multiple debugging, find the appropriate value to ensure fast convergence and good steady state performance. For example, for a QAM signal, The value should be between 0 and 1, generally Value ratio The value of the new one to two orders of magnitude can make the second weight update module 2012 effectively improve the output of the blind equalizer.
- FIG. 1 A block diagram of a blind equalization system in which the blind equalizer provided by the embodiment of the present invention is shown in FIG. Where the signal is sent Superimposed white Gaussian noise by adder 30 after channel 10 response , obtaining the input signal of the blind equalizer 20 , It is the output signal of the blind equalizer 20.
- Figure 4 shows the 64QAM local soft decision region, where the boxed region represents the divided local soft decision region, the open dots represent symbol points, and the black dots represent the output symbols of the blind equalizer. If the output of the blind equalizer is in the area Then, the second weight update module 2012 uses the partial posterior pdf of the symbol points in this area to adjust the weight To adjust the output of the blind equalizer. It should be appreciated that the division of soft decision regions of other orders is similar to that of Figure 4. As can be seen from FIG. 4, the second weight update module 2012 can reduce the number of calculations of the posterior probability density function by dividing the decision region, that is, the calculation can be simplified.
- FIG. 5 shows the intersymbol interference of the output signal after the equalization process is performed by the blind equalizer provided by the present invention when the 16QAM signal source is used ( Inter-Symbol Interference, ISI) simulation map, ISI is defined as:
- Channel Impulse response Represents the weight vector corresponding to the blind equalizer.
- the blind equalizer provided by the embodiment of the present invention is used, and only DSE-CMA is adopted. Convergence is faster than blind equalizers that perform weight vector updates; and ISI is smaller at steady state.
- the first weight update module 2011 is based on The steepest descent gradient algorithm divides the CMA error term into two parts: the real part and the imaginary part, then adds the jitter random signal and performs the symbolic operation to form a new error term.
- the jitter symbol error-modified constant modulus algorithm (Dithered Signed-Error - Modified) is used.
- the Constant Modulus Algorithm, DSE-MCMA) algorithm updates the first weight vector, specifically, the first weight vector is updated using the following iterative formula:
- the (positive) iteration step size for weight vector update using the DSE-MCMA algorithm which controls the convergence speed of the weight vector update using the DSE-MCMA algorithm.
- Is the output signal of the blind equalizer (ie, filter 202) For the received signal vector, the upper corner Representing a conjugate operation; Indicates the error term used to update the weight vector using the DSE-MCMA algorithm.
- the rest is the same as the DSE-CMA algorithm.
- the DSE-MCMA algorithm reduces the error term by adjusting the weight until the equalizer converges.
- the number of taps N of the first weight update module 2011 and the second weight update module 2012 The value is the same as in the above DSE-CMA algorithm, and can be taken based on the empirical value after multiple experiments.
- the second weight update module 2012, the weight combination module 2013, and the filter 202 The implementation is the same as in Embodiment 1, except that the algorithm for updating the first weight vector by the first weight update module 2011 is changed.
- FIG. 6 shows an output signal after the equalization process is performed by the blind equalizer provided by the embodiment of the present invention when the 16QAM signal source is used. ISI simulation diagram. It can be seen from FIG. 6 that the blind equalizer provided by the embodiment of the present invention has a convergence speed lower than that of the embodiment 1 of the present invention. The blind equalizer provided is faster and has better steady state performance.
- FIG. 7 is a flowchart showing an implementation process of a blind equalization processing method according to an embodiment of the present invention, which is described in detail as follows:
- step S701 the first weight vector is updated by using a DSE-CMA algorithm
- step S702 the maximum posterior probability theory is used as the judgment basis, and the steepest descending gradient algorithm is used to make the local posterior
- the logarithm of p.d.f. is the largest, and the second weight vector is updated;
- step S703 the updated first weight vector and the second weight vector are merged
- step S704 the received signal vector is equalized and output according to the weight vector obtained after the combination.
- the following method is used to equalize the received signal vector:
- the upper corner ' 'Indicating matrix transpose The weight vector obtained after the combination in step S703, For the received signal vector, The signal that is output after equalization processing.
- step S701 DSE-CMA is adopted.
- the algorithm updates the first weight vector the first weight vector is updated by using the following iterative formula:
- the transmitted signal is related to the output signal of the blind equalizer, and the values related to these three factors are respectively recorded as , with . specifically,
- steady-state EMSE which is usually selected as .
- step S702 when the second weight vector is updated, the second weight vector is specifically updated by using an iterative formula as follows:
- step S703 the combination of the weight vectors is specifically performed by the following formula:
- first weight vector And second weight vector Dimensions are ; , , with The selection can be determined based on actual experience. After multiple debugging, find the appropriate value to ensure fast convergence and good steady state performance. For example, for a QAM signal, The value should be between 0 and 1, generally Value ratio The value of the new one to two orders of magnitude, in order to make the updated second weight vector Effectively improve the performance of blind equalization processing.
- step S701 and step S702 may be exchanged or may be performed simultaneously.
- step S7011 In order to reduce the phase offset and further increase the convergence speed, as a preferred embodiment of the present invention, in step S7011, according to The steepest descent gradient algorithm divides the CMA error term into two parts: the real part and the imaginary part, then adds the jitter random signal and performs the symbolic operation to form a new error term, which is continuously reduced to reduce the new error term.
- the new first weight vector is until convergence, that is, the first weight vector is updated by the DSE-MCMA algorithm. Specifically, the first weight vector is updated by using the following iterative formula:
- the (positive) iteration step size for weight vector update using the DSE-MCMA algorithm which controls the convergence speed of the weight vector update using the DSE-MCMA algorithm;
- DSE-MCMA The algorithm reduces the error term by adjusting the weight until the equalizer converges.
- the second weight vector is updated and the weight is combined.
- the algorithm is unchanged, only the algorithm of the first weight vector update has changed.
- the weight vector is updated, and then the received signal vector is equalized according to the updated weight vector. After processing, the output realizes a blind equalizer. Can improve convergence speed and steady state performance.
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Abstract
本发明适用于通信领域,提供了一种盲均衡器及盲均衡处理方法,所述盲均衡器包括权值更新单元和滤波器;所述权值更新单元包括第一权值更新模块、第二权值更新模块以及权值合并模块。在本发明中,通过使第一权值更新模块的误差项最小且使第二权值更新模块采用的局部后验概率密度函数的对数值最大,更新权值向量,再根据更新后的权值向量对接收到的信号向量进行均衡处理后输出,实现了一种盲均衡器,能够提高收敛速度和稳态性能。
Description
本发明属于通信领域 . ,尤其涉及一种盲均衡器及盲均衡处理方法。
信号在非理想信道中传输会导致失真,均衡是用来减少失真且补偿信号损失的一种技术。由于盲均衡不需要使用训练序列,因而可以提高系统带宽效率。此外,对某些通信系统来说,在当接收端失去同步等时,很难在接收端得到正确的训练序列。而盲均衡则为这类系统提供了一个能消除信道有害影响的实际手段。
从 20 世纪 90 年代开始,研究和使用得最多的盲均衡算法是恒模算法( Constant
Modulus Algorithm , CMA )。 CMA
试图恢复某些通信信号的恒包络特性,通过恢复信号的模值,可能间接恢复信号的其它特性,从而有可能产生合适的性能。盲均衡系统的结构框图如图 1 所示,
为发送信号,
为信道
的冲激响应,
为均值为零的高斯白噪声,
为输入均衡器的接收信号,
表示长度为
的均衡器,
为均衡器的输出信号,盲均衡算法通过调整均衡器的权值减少码间干扰( Inter-Symbol
Interference , ISI ),使均衡器的输出与原发送信号逼近。 CMA 通过最小化一个非凸代价函数来实现,通常采用的代价函数为:
CMA 均衡器的权值更新方程采用最陡下降梯度算法:
其中,均衡器具有 N 个抽头系数,也称为权值;
表示权值向量;
是(正)迭代步长,它控制收敛速度;均衡器输入信号向量为
,输出信号为
;
为误差项,它是通过对代价函数
中的
求偏导得到的,即
; 上角标
表示取共轭运算 。
P. Schniter 和 C. R. Johnson 在' Dithered
signed-error CMA: The complex-valued case , in Proc. Asilomar Conf. Signals ,
Syst. , Comput., Pacific Grove , CA , 1998' 中描述了针对复数值信号的抖动符号误差恒模算法( Dithered
Signed-Error CMA , DSE-CMA
)。抖动是指在量化之前增加一个随机信号,以试图保存量化过程中丢失的信息。从加性噪声的角度来看,抖动是使量化噪声成为均值为零,且与被量化的信号相互独立的白噪声。
复数情况下, DSE-CMA 算法的权值更新方程同样采用最陡下降梯度法:
其中,
表示 DSE-CMA 算法的权向量;
为 DSE-CMA 算法的(正)迭代步长,它控制 DSE-CMA 算法的收敛速度; 上角标
表示取共轭运算;
表示 DSE-CMA 的误差项,
。其中,
表示输出信号
;
为 CMA 的误差项
,
即在量化之前增加的随机信号,
表示抖动幅度,是正常数;
,
和
分别是均匀分布在
上且服从独立同分布的抖动随机信号。
, sgn(.) 表示取符号操作,因此简化了计算。 DSE-CMA
算法通过调节权值以减少误差项,直到均衡器收敛为止。
DSE-CMA 算法通过把大量的更新乘法转变成符号操作,简化了计算。 DSE-CMA
算法的目的是在降低复杂性的同时,保持 CMA 的鲁棒性。但是, DSE-CMA 算法的收敛速度比较慢,稳态性能不够好。
本发明的目的旨在解决现有盲均衡器技术存在的收敛速度慢和稳态性能差的问题 。
本发明是这样实现的,一种盲均衡器,包括权值更新单元和滤波器;
所述权值更新单元包括:
第一权值更新模块,用于 采用抖动符号误差 - 恒模算法更新第一权值向量;
第二权值更新模块 ,用于 以最大后验概率理论作为判决依据,采用 最陡下降梯度算法, 为
使局部后验概率 密度函数 的对数值最大, 更新第二权值向量;以及
权值合并模块 ,用 于合并更新后的所述 第一权值向量和第二权值向量 ;
所述滤波器用于根据所述 权值合并模块 合并后得到的权值向量,对接收到的信号向量进行均衡处理后输出
。
本发明的另一目的在于提供旨在解决现有技术存在的问题,提供一种
盲均衡处理方法,包括下述步骤:
采用抖动符号误差 - 恒模算法更新第一权值向量;
以最大后验概率理论作为判决依据,采用 最陡下降梯度算法, 为 使局部后验概率 密度函数
的对数值最大,更新第二权值向量 ;
合并更新后的第一权值向量和第二权值向量;
根据合并后得到的权值向量,对接收到的信号向量进行均衡处理后输出。
在本发明中,通过
采用最陡下降梯度算法,为使第一权值更新模块的误差项最小且使第二权值更新模块采用的局部后验 概率
密度函数的对数值最大,更新权值向量,再根据更新后的权值向量对接收到的信号向量进行均衡处理后输出,实现了一种盲均衡器, 能够提高收敛速度和稳态性能。
图 1 是现有技术提供的 盲均衡系统的结构框图;
图 2 是本发明 实施例提供的 盲均衡器的结构框图;
图 3 是本发明 实施例提供的 盲均衡器所处的盲均衡系统的框图;
图 4 是本发明 实施例提供的 局部软判决区域划分图 ;
图 5 是本发明 实施例提供的 码间干扰( ISI )仿真图 ;
图 6 是本发明 实施例提供的 码间干扰( ISI )仿真图 ;
图 7 是本发明实施例提供的盲均衡处理方法的实现流程。
为了使本发明的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本发明进行进一步详细说明。应当理解,此处所描述的具体实施例仅仅用以解释本发明,并不用于限定本发明。
实施例 1 :
在本发明实施例中,通过采用最陡下降梯度算法,为使第一权值更新模块的误差项最小且使第二权值更新模块采用的局部后验 概率
密度函数的对数值最大,更新权值向量,再根据更新后的权值向量对接收到的信号向量进行均衡处理后输出,实现了一种盲均衡器。
图 2
示出了本发明实施例提供的盲均衡器的结构,为了便于说明仅示出了与本发明实施例相关的部分。
该盲均衡器可以用于无线移动通信系统的接收机,可以是运行于这些接收机内的软件单元、硬件单元或者软硬件相结合的单元,其中:
权值更新单元 201 ,根据最陡下降梯度算法,为了使第一权值更新模块 201
的误差项最小且使第二权值更新模块 202 采用的局部后验概率密度函数( probability density function , p.d.f.
)的对数值最大,更新权值向量;
滤波器 202 ,根据权值更新单元 201 更新后的权值向量,对接收到的信号向量进行均衡处理后输出
,采用下式对接收到的信号向量进行均衡处理:
具体地,如图 2 所示,权值更新单元 201 包括第一权值更新模块 2011 、第二权值更新模块
2012 和权值合并模块 2013 :
第一权值更新模块 2011 , 采用 DSE-CMA
算法更新第一权值向量,具体采用如下迭代公式对第一权值向量进行更新:
其中,
表示第一权值更新模块 2011 在
时刻的权值向量;
为采用 DSE-CMA 算法进行权值向量更新的(正)迭代步长,它控制采用 DSE-CMA
算法进行权值向量更新的收敛速度,
为盲均衡器(也即滤波器 202 )的输出信号
,
为接收到的信号向量 , 上角标
表示取共轭运算;
表示采用 DSE-CMA 算法进行权值向量更新的误差项。具体地,
其中,
为采用 CMA 算法进行权值向量更新的误差项,
;
即在量化之前增加的随机信号,
表示抖动幅度,为正常数,
为恒模值,
,
和
分别是均匀分布在
上且服从独立同分布的抖动随机信号;
, sgn(.) 表示取符号操作,从而可以简化计算。
第二权值更新模块 2012 ,以最大后验概率理论作为判决依据,采用 最陡下降梯度算法, 为
使局部后验 p.d.f. 的对数值最大,从而不断 更新第二权值向量直至收敛,具体采用如下迭代公式对第二权值向量进行更新:
其中,
表示 第二权值更新模块 2012 在
时刻的 权值向量;
表示进行权值向量更新的(正)迭代步长,它控制进行权值向量更新的收敛速度;
,
为 权值合并模块 2013 在
时刻的 权值向量 ,
表示噪声方差,其取值与信道的散布有关,
表示盲均衡器(也即滤波器 202 )输出的局部后验 p.d.f. 。第 二权值更新模块 2012
使用最陡下降梯度算法调整
,通过减小
的值从而最大化局部后验 p.d.f. 的对数值。
其中, 第一权值更新模块 2011 和第二权值更新模块 2012 的抽头个数都为
;
、
、
和
的选取可以根据实际经验确定,经多次调试找到合适的取值以保证快速收敛和良好的稳态性能。例如,对于 QAM
信号,
的取值应该在 0 到 1 之间,一般
的取值比
的取值大一至两个数量级,才能使 第二权值更新模块 2012 能起到有效改善盲均衡器输出的作用。
本发明实施例提供的盲均衡器所处的盲均衡系 统的框图如图 3 所示。其中,发送信号
经过信道 10 响应后再通过加法器 30 叠加上高斯白噪声
,得到盲均衡器 20 的输入信号
,
为盲均衡器 20 的输出信号。
图 4 展示了 64QAM
局部软判决区域,其中方框区域表示划分后的局部软判决区域,空心点表示符号点,黑点表示盲均衡器的输出符号。如果盲均衡器的输出位于区域
内,则 第二权值更新模块 2012 利用此区域中的符号点构成的局部后验 p.d.f. 调整权值
,从而调整 盲均衡器的输出。应当认识到,其它阶数的软判决区域的划分类似于图 4 。从图 4
可以看出,第二权值更新模块 2012 通过划分判决区域可以减少后验概率密度函数的计算次数,即可以简化计算。
图 5 展示采用 16QAM 信号源时,根据本发明提供的盲均衡器进行均衡处理后,输出信号的码间干扰(
Inter-Symbol Interference , ISI )仿真图, ISI 定义为:
从图 5 可以看到,采用本发明实施例所提供的盲均衡器,与仅采用 DSE-CMA
进行权值向量更新的盲均衡器相比,收敛更快;且稳态时, ISI 更小。
实施例 2 :
为了减少相位偏移,并进一步提高收敛速度,作为本发明的一个优选实施例,第一权值更新模块 2011 根据
最陡下降梯度算法,把 CMA 误差项分为实部和虚部两部分,然后加入抖动随机信号并进行取符号操作,构成新的误差项,
为减小新的误差项从而不断更新第一权值向量直至收敛,即采用抖动符号误差-改进恒模算法( Dithered Signed-Error - Modified
Constant Modulus Algorithm, DSE-MCMA )算法更新第一权值向量,具体地,采用如下迭代公式对第一权值向量进行更新:
其中,
表示第一权值更新模块 2011 在
时刻的权值向量;
为采用 DSE-MCMA 算法进行权值向量更新的(正)迭代步长,它控制采用 DSE-MCMA
算法进行权值向量更新的收敛速度,
为盲均衡器(也即滤波器 202 )的输出信号
,
为接收到的信号向量 , 上角标
表示取共轭运算;
表示采用 DSE-MCMA 算法进行权值向量更新的误差项。具体地,
其余部分与 DSE-CMA 算法相同。 DSE-MCMA
算法通过调整权值减小误差项,直至均衡器收敛。其中, 第一权值更新模块 2011 和第二权值更新模块 2012 的抽头个数 N 、
的取值与上述 DSE-CMA 算法中相同,可以依据多次实验后的经验值取。
在本发明实施例中, 第二权值更新模块 2012 、权值合并模块 2013 和 滤波器 202
的实现与实施例 1 中相同,仅仅是第一权值更新模块 2011 更新第一权值向量的算法有所改变。
图 6 展示采用 16QAM 信号源时,根据本发明实施例所提供的盲均衡器进行均衡处理后,输出信号的
ISI 仿真图。从图 6 可以看到,采用本发明实施例所提供的盲均衡器,由于减小了相位失真,其收敛速度比本发明实施例 1
提供的盲均衡器要快,且稳态性能更好。
另外,虽然为了简便,这里只仅给出了 16QAM 的仿真图,但是可以采用其它阶数的 QAM
来使用本发明提供的盲均衡器。
实施例 3 :
图 7 示出了本发明实施例提供的盲均衡处理方法的实现流程,详述如下:
在步骤 S701 中,采用 DSE-CMA 算法更新第一权值向量;
在步骤 S702 中,以最大后验概率理论作为判决依据,采用 最陡下降梯度算法, 为 使局部后验
p.d.f. 的对数值最大,更新第二权值向量 ;
在步骤 S703 中,合并更新后的第一权值向量和第二权值向量;
在步骤 S704 中,根据合并后得到的权值向量,对接收到的信号向量进行均衡处理后输出
,采用下式对接收到的信号向量进行均衡处理:
具体地,在步骤 S701 中,采用 DSE-CMA
算法更新第一权值向量时,具体采用如下迭代公式对第一权值向量进行更新:
其中,
表示在
时刻的权值向量;为采用 DSE-CMA 算法进行权值向量更新的(正)迭代步长,它控制采用 DSE-CMA
算法进行权值向量更新的收敛速度;
表示采用 DSE-CMA 算法进行权值向量更新的误差项;
表示输出信号
;
为接收到的信号向量;上角标
表示取共轭运算。具体地,
其中,
为采用 CMA 算法进行权值向量更新的误差项,
;
即在量化之前增加的随机信号,
表示抖动幅度,为正常数,
为恒模值,
,
和
分别是均匀分布在
上且服从独立同分布的抖动随机信号;
, sgn(.) 表示取符号操作,从而可以简化计算。
在步骤 S702 中,更新第二权值向量时,具体采用如下迭代公式对第二权值向量进行更新:
其中,
表示 在
时刻的 权值向量;
表示进行权值向量更新的(正)迭代步长,它控制进行权值向量更新的收敛速度;
,
为
时刻进行均衡处理的 权值向量 ,
表示噪声方差,其取值与信道的散布有关,
表示进行均衡处理后输出的局部后验 p.d.f. 。通过使用最陡下降梯度算法调整
,通过减小
的值从而最大化局部后验 p.d.f. 的对数值。
在步骤 S703 中,具体采用下式进行权值向量的合并:
其中, 第一权值向量
和第二权值向量
的维数都为
;
、
、
和
的选取可以根据实际经验确定,经多次调试找到合适的取值以保证快速收敛和良好的稳态性能。例如,对于 QAM
信号,
的取值应该在 0 到 1 之间,一般
的取值比
的取值大一至两个数量级,才能使更新后的第二权值向量
起到有效改善盲均衡处理的性能。
当然,步骤 S701 和步骤 S702 的执行顺序可以交换,也可以同时执行。
实施例 4 :
为了减少相位偏移,并进一步提高收敛速度,作为本发明的一个优选实施例,在步骤 S7011 中,根据
最陡下降梯度算法,把 CMA 误差项分为实部和虚部两部分,然后加入抖动随机信号并进行取符号操作,构成新的误差项,为减小新的误差项从而不断地更
新第一权值向量直至收敛,即采用 DSE-MCMA 算法更新第一权值向量,具体地,采用如下迭代公式对第一权值向量进行更新:
其中,
表示在 n 时刻的权值向量;
为采用 DSE-MCMA 算法进行权值向量更新的(正)迭代步长,它控制采用 DSE-MCMA
算法进行权值向量更新的收敛速度;
表示采用 DSE-MCMA 算法进行权值向量更新的误差项。具体地,
其余部分与 DSE-CMA 算法相同。 DSE-MCMA
算法通过调整权值减小误差项,直至均衡器收敛。
在本发明实施例中, 第二权值向量更新、权值合并
的算法不变,仅仅是第一权值向量更新的算法有所改变。
在本发明中,通过采用最陡下降梯度算法,为减小误差项且使局部后验概率的对数值最大,更新权值向量,再根据更新后的权值向量对接收到的信号向量进行均衡处理后输出,实现了一种盲均衡器,
能够提高收敛速度和稳态性能。
本领域普通技术人员可以理解,实现上述实施例方法中的全部或部分步骤是可以通过程序来指令相关的硬件来完成,所述的程序可以在存储于一计算机可读取存储介质中,所述的存储介质,如
ROM/RAM 、磁盘、光盘等。
以上所述仅为本发明的较佳实施例而已,并不用以限制本发明,凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等,均应包含在本发明的保护范围之内。
Claims (10)
- 一种盲均衡器,其特征在于,所述盲均衡器包括权值更新单元和滤波器;所述权值更新单元包括:第一权值更新模块,用于 采用抖动符号误差 - 恒模算法更新第一权值向量;第二权值更新模块 ,用于 以最大后验概率理论作为判决依据,采用 最陡下降梯度算法, 为 使局部后验概率 密度函数 的对数值最大, 更新第二权值向量;以及权值合并模块 ,用 于合并更新后的所述 第一权值向量和第二权值向量 ;所述滤波器用于根据所述权值合并模块合并后得到的权值向量,对接收到的信号向量进行均衡处理后输出。
- 如权利要求 1 所述的盲均衡器,其特征在于,所述第二权值更新模块进行权值向量更新所采用的迭代步长比所述第一权值更新模块进行权值向量更新所采用的迭代步长大一至两个数量级。
- 如权利要求 1 所述的盲均衡器,其特征在于,所述第一权值更新模块 采用抖动符号误差 - 改进恒模算法更新第一权值向量。
- 一种盲均衡处理方法,其特征在于,所述方法包括下述步骤:采用抖动符号误差 - 恒模算法更新第一权值向量;以最大后验概率理论作为判决依据,采用 最陡下降梯度算法, 为 使局部后验概率 密度函数 的对数值最大,更新第二权值向量 ;合并更新后的第一权值向量和第二权值向量;根据合并后得到的权值向量,对接收到的信号向量进行均衡处理后输出。
- 如权利要求 8 所述的方法,其特征在于, 采用抖动符号误差 - 改进恒模算法更新所述第一权值向量。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201010171519.8 | 2010-05-13 | ||
CN 201010171519 CN101854317A (zh) | 2010-05-13 | 2010-05-13 | 一种盲均衡器及盲均衡处理方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2011140833A1 true WO2011140833A1 (zh) | 2011-11-17 |
Family
ID=42805591
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2011/070029 WO2011140833A1 (zh) | 2010-05-13 | 2011-01-04 | 一种盲均衡器及盲均衡处理方法 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN101854317A (zh) |
WO (1) | WO2011140833A1 (zh) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105162738A (zh) * | 2015-07-30 | 2015-12-16 | 南京信息工程大学 | 一种卫星信道复数神经多项式网络盲均衡系统及方法 |
CN113259284A (zh) * | 2021-05-13 | 2021-08-13 | 中南大学 | 一种基于Bagging和长短期记忆网络的信道盲均衡方法及系统 |
CN114095320A (zh) * | 2021-11-11 | 2022-02-25 | 西安电子科技大学 | 基于动量分数阶多模盲均衡算法的信道均衡方法 |
CN114826842A (zh) * | 2022-03-27 | 2022-07-29 | 西安电子科技大学 | 非高斯噪声下多进制正交调幅信号多模盲均衡方法及系统 |
CN114826841A (zh) * | 2022-03-27 | 2022-07-29 | 西安电子科技大学 | 基于多模算法的低复杂度高斯-牛顿盲均衡方法及系统 |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101854317A (zh) * | 2010-05-13 | 2010-10-06 | 深圳大学 | 一种盲均衡器及盲均衡处理方法 |
CN102255834A (zh) * | 2011-07-05 | 2011-11-23 | 电子科技大学 | 一种快速收敛的半盲均衡方法 |
CN102904643B (zh) * | 2012-10-11 | 2016-03-30 | 复旦大学 | 适用于qdb频谱压缩偏振复用信号的多模盲均衡方法 |
CN103873405B (zh) * | 2014-02-28 | 2017-08-29 | 北京遥测技术研究所 | 一种基于apsk调制信号幅值的恒模盲均衡方法 |
FR3030963B1 (fr) * | 2014-12-18 | 2017-12-29 | Continental Automotive France | Egalisateur aveugle de canal |
CN107786475B (zh) * | 2016-08-26 | 2020-04-10 | 深圳市中兴微电子技术有限公司 | 盲均衡误差计算方法和装置 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1806398A (zh) * | 2003-07-16 | 2006-07-19 | 三星电子株式会社 | 移动通讯系统中用自适应天线阵列接收数据的设备和方法 |
CN101854317A (zh) * | 2010-05-13 | 2010-10-06 | 深圳大学 | 一种盲均衡器及盲均衡处理方法 |
-
2010
- 2010-05-13 CN CN 201010171519 patent/CN101854317A/zh active Pending
-
2011
- 2011-01-04 WO PCT/CN2011/070029 patent/WO2011140833A1/zh active Application Filing
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1806398A (zh) * | 2003-07-16 | 2006-07-19 | 三星电子株式会社 | 移动通讯系统中用自适应天线阵列接收数据的设备和方法 |
CN101854317A (zh) * | 2010-05-13 | 2010-10-06 | 深圳大学 | 一种盲均衡器及盲均衡处理方法 |
Non-Patent Citations (4)
Title |
---|
CHE, WEN ET AL.: "A Hybrid Maximum Likehood and Probability Data Association MIMO Detection Algorithm", JOURNAL OF JILIN UNIVERSITY (ENGINEERING AND TECHNOLOGY EDITION), vol. 38, no. 5, September 2008 (2008-09-01), pages 1175 - 1180 * |
LIU, SHUNLAN ET AL.: "New Blind Equalization Algorithm Based on Concurrent SCA and SDD", JOURNAL OF DATA ACQUISITION & PROCESSING, vol. 23, no. 5, September 2008 (2008-09-01), pages 537 - 541 * |
SCHNITER, P. ET AL.: "The Dithered Signed-Error Constant Modulus Algorithm", PROCEEDINGS OF THE 1998 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING, vol. 6, 15 May 1998 (1998-05-15), pages 3353 - 3356 * |
ZHANG, CHENGYU ET AL.: "Low Complexity Blind Equalization Algorithm Based on Probability Density Function", JOURNAL OF DALIAN MARITIME UNIVERSITY, vol. 34, no. 3, August 2008 (2008-08-01), pages 43 - 50 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105162738A (zh) * | 2015-07-30 | 2015-12-16 | 南京信息工程大学 | 一种卫星信道复数神经多项式网络盲均衡系统及方法 |
CN105162738B (zh) * | 2015-07-30 | 2018-11-27 | 南京信息工程大学 | 一种卫星信道复数神经多项式网络盲均衡系统及方法 |
CN113259284A (zh) * | 2021-05-13 | 2021-08-13 | 中南大学 | 一种基于Bagging和长短期记忆网络的信道盲均衡方法及系统 |
CN113259284B (zh) * | 2021-05-13 | 2022-05-24 | 中南大学 | 一种基于Bagging和长短期记忆网络的信道盲均衡方法及系统 |
CN114095320A (zh) * | 2021-11-11 | 2022-02-25 | 西安电子科技大学 | 基于动量分数阶多模盲均衡算法的信道均衡方法 |
CN114826842A (zh) * | 2022-03-27 | 2022-07-29 | 西安电子科技大学 | 非高斯噪声下多进制正交调幅信号多模盲均衡方法及系统 |
CN114826841A (zh) * | 2022-03-27 | 2022-07-29 | 西安电子科技大学 | 基于多模算法的低复杂度高斯-牛顿盲均衡方法及系统 |
CN114826841B (zh) * | 2022-03-27 | 2023-12-05 | 西安电子科技大学 | 基于多模算法的低复杂度高斯-牛顿盲均衡方法及系统 |
CN114826842B (zh) * | 2022-03-27 | 2024-04-12 | 西安电子科技大学 | 非高斯噪声下多进制正交调幅信号多模盲均衡方法及系统 |
Also Published As
Publication number | Publication date |
---|---|
CN101854317A (zh) | 2010-10-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2011140833A1 (zh) | 一种盲均衡器及盲均衡处理方法 | |
WO2017222140A1 (ko) | Cnn 기반 인루프 필터를 포함하는 부호화 방법과 장치 및 복호화 방법과 장치 | |
US4847797A (en) | Adaptive blind equilization method and device | |
WO2013055046A2 (ko) | 통신/방송 시스템에서 데이터 송수신 장치 및 방법 | |
US6151358A (en) | Method and apparatus, and computer program for producing filter coefficients for equalizers | |
WO2016163855A1 (en) | Method for multiplexing uplink information | |
US20120134407A1 (en) | Adaptive equalization with group delay | |
WO2004077670A1 (en) | Nonlinear filter | |
US20130230092A1 (en) | Sparse and reconfigurable floating tap feed forward equalization | |
CA2529551A1 (en) | Updating adaptive equalizer coefficients using known or predictable bit patterns distributed among unknown data | |
WO2020218769A1 (en) | Reluctance force compensation for loudspeaker control | |
WO2017111362A1 (ko) | 임의의 길이를 가지는 폴라 코드를 이용한 harq 수행 방법 | |
EP1295448B1 (en) | Channel equalizer | |
EP1776814A1 (en) | Method for calculating filter coefficients for an equaliser in a communication receiver using hermitian optimisation | |
JP3625205B2 (ja) | 適応等化器及び受信装置 | |
WO2019027142A1 (ko) | 동적 네트워크에서 네트워크 형성 방법 및 장치 | |
WO2021256843A1 (ko) | 동형 암호문에 대한 통계 연산 수행하는 장치 및 방법 | |
WO2022164095A1 (en) | Method and apparatus for signal processing | |
US20050047779A1 (en) | Post-detection, fiber optic dispersion compensation using adjustable inverse distortion operator employing trained or decision-based parameter adaptation (estimation) | |
WO2023106448A1 (ko) | 직교 진폭 변조 필터 뱅크 다중 반송파 통신 시스템에서 낮은 자기 간섭 및 높은 주파수 효율에 도달하기 위한 송수신기, 송수신 방법 및 수신 원형 필터 설계 방법 | |
Endres et al. | Carrier independent blind initialization of a DFE | |
WO2023219271A1 (ko) | 통신 시스템 또는 방송 시스템에서 미래 제약을 이용한 극부호의 복호화를 위한 장치 및 방법 | |
KR100227804B1 (ko) | 가변 적응 상수를 이용한 채널 등화기 및 채널 등화기의 탭 계수 갱신방법 | |
WO2024019341A1 (ko) | 멀티-뉴머롤로지를 지원하는 mimo 시스템에서 선형 결합기를 포함하는 수신 장치 및 이의 동작 방법 | |
WO2022034967A1 (ko) | 적응형 지연 다이버시티 필터와, 이를 이용하는 에코 제거 장치 및 방법 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 11780055 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 11780055 Country of ref document: EP Kind code of ref document: A1 |