Adaptive Dereverberation of Speech Signal

  • Praveen P B, Dr. Aravind H S

Abstract

The quality of the speech signal degrades mainly due to additive noise and/or reverberation. This in turn degrades the performance of many speech-based applications like automatic speech recognition, telecommunications, speech recordings, etc. This paper deals with reducing the reverberation effect without prior knowledge of room reverberation response (RIR). In most cases,it is unknown, therefore, much research done on blind dereverberation [1][2]. This paper proposes an adaptive method to perform dereverberation of an unknown RIR. This method first estimates the impulse response of the echo path using the adaptive filter estimation and using these estimated weights, it cancels the echo. Meaning this method doesn’t extract the clean signal but the direct signal. Performance and convergence of the estimated weights/coefficients of an adaptive filter purely depend on the type of algorithm used for estimation in the adaptive filter. Since this method adaptsto time-varyingimpulses, this means it also adaptsto the speaker’s position change. This method is implemented and tested using the MATLAB tool.

Published
2022-06-14
How to Cite
Praveen P B, Dr. Aravind H S. (2022). Adaptive Dereverberation of Speech Signal . Design Engineering, (1), 3517 - 3527. Retrieved from http://thedesignengineering.com/index.php/DE/article/view/9488
Section
Articles