Adaptive Echo Cancellation Based on a Multipath Model of Acoustic Channel

Abstract

Usually Acoustic Echo Cancellers (AECs) are realized by adaptive Finite duration Impulse Response (FIR) filter having large number of coefficients and Least Mean Square (LMS) as an adaptive algorithm resulting in slow convergence speed and poor tracking performance of these adaptive filters. In this paper, we have proposed a Multiple Sub-filter (MSF) parallel structure based on multipath acoustic echo model using the basis that each sub-filter will compensate the echo contributed by each path of multipath acoustic channel. To realize the MSF, modified Generalized Autocorrelation-based Estimator (MAE) has been used to estimate time delay associated with each path while the order of each sub-filter has been estimated using Power Spectral Density (PSD) method. Accuracy Percentage (AP) performance measure has been used to characterize the performance of the estimator. Simulation results show that the performance of the MAE improves with the increase in SNR and/or decrease in number of multipath. Using these estimates MSF based AEC is constructed. The convergence performance of MSF-based AEC has been studied, via computer simulation, and compared with the conventional Single Long length adaptive Filter (SLF)-based canceller for different SNRs and number of multipath. The results of MSF have been found to be very encouraging in almost all of the various situations considered. Subsequently, the tracking behavior has also been studied with variation in the channel parameters of the multipath model. The proposed MSF can track variations in the channel parameters of the multipath model faster as compared to the conventional echo canceller.