Non linear adaptive filters for echo cancellation of speech coded signals

Adaptive filters have been widely used for echo cancellation, channel equalization, adaptive beamforming and many other applications. Polynomial adaptive filters are gaining wide acceptance because of the improvement in performance shown by these filters when compared to the traditional linear filters.; In this research we investigate the effect of using polynomial filters for echo cancellation of speech coded signals. Speech codecs are usually present in the echo path and create a significant amount of non-linearity. In this thesis we consider the Adaptive Multi Rate codec (AMR) and G.729 codec because of their application in the present day cellular technologies. We consider the acoustic echo produced within the body of the handset. This echo path is usually non-linear and is of short duration (∼15ms). We analyze the performance of the Volterra polynomial filters and the bilinear polynomial filters in reducing the non-linearity introduced by the speech codecs present in the echo path. The results show that by choosing an appropriate filter length and a sufficient number of filter coefficients, the polynomial filters show a significant improvement in performance when compared to the linear filters.