Speech and Audio Signal Application

INTRODUCTION Speech and audio signal processing has been a topic of active research during the last two decades, given as a result of the development of several efficient algorithms for echo and noise cancelling, active noise cancelling, speech enhancement, etc., most of these applications depending on transversal adaptive filters. Although, recently, the multirate implementations of those systems have proved to be a desirable alternative when large filters order with fast convergence rates are required. In multirate adaptive filtering the input signals are divided into N subbands signals using a filter bank or an orthogonal transformation. Subsequently, an Nf/M order adaptive filter is inserted in each subband, where Nf is the full band adaptive filter order, whose coefficients are updated to minimize either the output errors in each subband given by the difference between the subband adaptive filter output and the subband reference signal, or a full band common output error. Finally, some of the signals involved in the adaptation process are used to synthesize the adaptive filter output signal. Multirate adaptive filtering presents several advantages over the conventional full band filter structure: a) Because each band is downsampled by a factor M, the computational complexity of the subband system is approximately reduced by a factor M, compared with the full band one. This fact allows the development of more sophisticated and efficient adaptive algorithms, as well as 701 E. Chocolate Avenue, Hershey PA 17033-1117, USA Tel: 717/533-8845; Fax 717/533-8661; URL-http://www.idea-group.com ITB7209 IDEA GROUP PUBLISHING