总体最小二乘的改进与信道估计应用

Total Least Squares' Improvement and Application in Channel Estination

总体最小二乘法同时考虑了数据矩阵和观察值的误差,在参数估计中得到广泛应用。然而总体最小二乘法没有针对具体问题利用误差的先验信息。总体最小二乘的改进算法—结构总体最小二乘法设定了误差矩阵的结构,通过迭代运算估计参数及误差。基于叠加训练序列的时不变信道估计中,信息序列均值构成的误差矩阵为Toeplitz矩阵。结构总体最小范数法作为一种结构总体最小二乘法,可以设定误差矩阵具有Toeplitz结构,有效提高叠加训练信道估计性能。论文对比了最小二乘,数据最小二乘,总体最小二乘和结构总体最小范数在叠加训练序列信道估计中的应用,仿真结果表明,基于结构总体最小范数的估计算法的归一化信道均方误差最小。

Total least squares（TLS）method,considering errors both in data matrix and observation values,is widely used in parameter estimation.However,it doesn’t utilize prior knowledge of error under certain conditions.The improved TLS method sets the structure of error matrix,and calculates parameters and errors through iteration calculation.In the frequency selective channel estimation based on superimposed training（st）,the error matrix constructed by the information sequence mean value is Toeplitz matrix.The structured total least norm（STLN）method,one of the STLS,can make error matrix have the structure of Toeplitz mafrix to effectively improve the poficiency of superimposed training channel estimation. Application of least squares（LS）,data least squares（DLS）,TLS,STLN in ST channel estimation are compared to find that the error of least normalized channel with STLN algorithm is the least one.