Correlation-based adaptive filters for channel identification

Ning Yao, Man-Wai Kwan, Chi-Wah Kok

Abstract: A correlation-based recursive least-squares (CRLS) adaptive filter design is proposed. This algorithm is applied to the channel identification problem by minimizing the difference between the correlation functions of the received signal and of the designed filter output in the LS sense. The CRLS algorithm is shown to be much more robust to strong and instantaneous channel disturbance than the traditional RLS algorithm, especially when the disturbance is correlated with the input source signal. This is a common situation when echo cancellation is considered in communication systems with double-talk. The results of simulations to compare the channel identification performance of the CRLS and the traditional RLS algorithm for echo cancellation with double-talk are presented; they show the high robustness of the proposed CRLS algorithm.