Echo Cancellation Using Predictive Deconvolution Based on High Order Statistics

In this paper, we present a new method of predictive deconvolution based on high order statistics, and apply it on echo cancellation. The traditional predictive deconvolution is based on the assumption that the observed signal and the estimated signal are orthogonal. Since the speech signals have the property of nongaussianity, we can design an improved predictive deconvolution based on high order statistics to achieve a better result. Experiment result shows that the proposed method can remove the echo from the observed signals better than using the traditional predictive deconvolution.