Selection of an optimal adaptive filter for speech signal noise cancellation using C6455 DSP

Abstract

This article discusses the implementation of the various adaptive filters for the filtering of the noisy speech signal, whose spectrum overlaps or is close to the information signal. Three types of adaptive filters are compared: LMS, NLMS and RLS. The computational load of the C6455 digital signal processor is monitored for different order filters and the efficiency of the filtering is measured using signal-to-noise ratio off the output signal. The results determined for the most suitable family of filters for application in real situations Ill. 8, bibl. 10 (in English; abstracts in English and Lithuanian).