The Problem of Linear Predictive Algorithms for Blind Multichannel Identification

摘要

Traditionally, blind channel identification/equalization techniques have been based on higher-order statistics, which are known to suffer from many drawbacks. Later some methods using only second order statistics have been proposed, and it's a major breakthrough. The main methods based on second order statistics include linear prediction algorithm (LPA), outer product decomposition algorithm (OPDA), multi-step linear prediction algorithm (MSLP), least square smoothing algorithm (LSS), and constrained minimum output energy algorithm (CMOE). But the simulations of these algorithms show that the channel couldn't be identified because there still needs other condition, which couldn't be got for knowing nothing about the channel in the whole blind condition. This paper will analyze these algorithms and point out the reason that results in the failure of channel identification.