On the Enhancement of Dereverberation Algorithms Based on a Perceptual Evaluation Criterion

摘要

This paper describes an optimization strategy based on a perceptual assessment criterion for dereverberation algorithms. The complete procedure is applied to the adaptive inverse-filtering (AIF) and spectral subtraction (SS) stages of a given dereverberation algorithm using the so-called QAreverb quality measure. Experimental results, using a 204-signal speech database, indicate that the associated algorithm can be greatly simplified (in about 97% of the overall computational complexity) by removing the AIF stage. In addition, a fine tuning of the SS stage is able to improve in 6% the algorithm's QAreverb score, resulting in a much simpler and more efficient algorithm in a perceptual point of view.