Exploring the influence of reverberation on the perceived quality of time-scale modified speech

Abstract

One of the methods to enhance speech intelligibility is to use time-scale modification (TSM) algorithms. This approach is especially useful in challenging acoustic environments caused by noise or reverberation, for instance. Unfortunately, TSM algorithms tend to introduce distortions into speech, and these distortions become more audible as the slowdown ratio increases. The purpose of the experiment was to determine whether adding reverberation to slowed speech could improve perceived quality. To explore this, MUSHRA tests were conducted with two types of reverberation, three slowdown ratios, and sentences in both Polish and English. Under the experimental conditions investigated, the results indicated that at a slowdown ratio of 2.5, participants rated the speech with reverberation as higher in quality than the same speech without reverberation. No significant differences in quality perception were found between languages, although the characteristics of the reverberation itself were significant.