2006, Vol.9, No.4, pp.370-379

This paper deals with the analysis of a near-end speech detector which controls the system for reducing of echo in hands-free telephones sets. In the first part of the paper it was analyzed a detector based on calculating of coherent (ordinary and partial) functions and consisting of two microphones and one loudspeaker. The simulation algorithm was described which is verifying the possibility of detecting of the near-end speech signal. Further in the paper a detector was proposed that consists of one microphone and one loudspeaker and where an ordinary coherence was used. Simulation results are shown and comparative analysis with near-end speech detector with two microphones was performed. Verifying of the functioning of the detector was realized in conditions of superimposed ambient disturbances that were analyzed for various values of SNR. On the base of the limiting values of the ordinary coherence function thresholds for deciding on the echo-signal existence were determined. In the second part of this paper the speed of the algorithm convergence for calculating of coherent functions in near-end speech detector was analyzed. The analysis was made for cases of implementing of some classic, time symmetric window functions. Times were defined in which the process of algorithm convergence runs. By the comparative analysis it was shown the effect of some window functions on the algorithm of coherence calculation from [6].
Key words: acoustic echo cancelation, detection, adaptive filtering

Full text:  Acrobat PDF  (286KB)  

Copyright © Nonlinear Phenomena in Complex Systems. Last updated: January 4, 2007