Objective and subjective intelligibility evaluations of noise-reduction algorithms in Mandarin

Abstract

In this paper, we summarize the recent work that we have done on the objective and subjective evaluations of single-channel noise-reduction algorithms in Mandarin Chinese. In the evaluations, clean Mandarin speech signals were first corrupted by three types of noises at two signal-to-noise ratios and then processed by five typical single-channel noise-reduction algorithms. The processed signals were presented to normal-hearing listeners for recognition in subjective evaluations, and passed to eight intelligibility prediction measures in objective evaluations. Subjective evaluation results showed that the majority of noise-reduction algorithms did not improve Mandarin speech intelligibility, and the objective evaluation results indicated that of all tested objective measures, the short-time objective intelligibility (STOI) measure provided the highest abilities in predicting Mandarin speech intelligibility in all conditions and in predicting the effect on speech intelligibility due to non-linear noise-reduction processing. These evaluation results reported here do provide valuable hints for analyzing and optimizing noise-reduction algorithms for Mandarin.