Evaluation of Objective Intelligibility Prediction Measures for Speech Enhancement in Mandarin

In this paper, we evaluate the performance of several state-of-the-art objective measures in terms of predicting speech intelligibility in Mandarin of the processed noisy signals by speech enhancement algorithms. The speech signals were first corrupted by three types of noises at two signal-to-noise ratios, followed by four classes of speech enhancement algorithms. The objective intelligibility prediction measures were then performed. The subjective intelligibility ratings were obtained by performing a comprehensive investigation of the unprocessed noisy signals and the processed signals by various single-channel noise reduction algorithms through listening tests. Based on the subjective intelligibility scores, in this paper, we focus on examining the capability of objective intelligibility prediction measures using the correlation analysis and the standard deviation of error. The analysis results reported in this paper do provide valuable hints for analyzing and optimizing noise-reduction algorithms.