一种鲁棒的基于子空间分解的盲信噪比估计方法

摘要

To solve the poor robustness problem of the signal and noise subspace dimension estimation in the subspace based algorithm, it uses the knowledge of oversampling rate and proposes a new method to built the autocorrelation matrix, which reduces the relevance of data in the matrix,thus improving the precision of the estimation. For the fact that the minimum distance length (MDL) criteria can only accurately estimates the dimension in limited signal-to-noise ratio (SNR) scope, a noise power method is introduced to estimate the signal subspace dimension, which improves the performance of the original algorithm when the SNR is too low or too high, so that the SNR bound of the estimation is augmented. Simulation results show that the new method can directly deal with the intermediate frequency (IF) signal with a better estimation performance and it is not sensitive to the shaping filter roll-off factor and modulation mode.%为解决原子空间分解算法中对信号和噪声子空间维数估计鲁棒性差的问题,根据信号的过采样率信息提出了一种新的构造自相关矩阵的方法,以减少矩阵内部数据间的相关性,从而达到提高信号和噪声子空间维数估计精度的目的.同时,针对利用最小描述距离(MDL)准则估计维数时,只能在有限信噪比范国内实现精确估计这一情况,采用噪声功率(NP)法对信号子空间维数进行估计,进一步改善了原算法在信噪比过低或过高时估计性能变差的问题,增大了信噪比的估计范围.仿真结果表明:新算法能对中频信号直接处理,具有较低的估计偏差和均方误差,且对成型滤波器滚降系数及调制方式均不敏感.