基于CSE的SRMR语音可懂度客观评价算法的研究

Speech intelligibility index (SII) commonly used in speech intelligibility objectively evaluate algorithms。Depending on whether pure voice as input，Algorithms are divided into invasive and non-invasive two categories. Traditional intelligibility evaluation algorithms based mainly intrusive,and objective and subjective evaluation results of the evaluation results are highly correlated. However, in view of its dependence clean speech features, In this paper, speech-to-reverberation modulation energy ratio (SRMR) algorithm for extracting speech after cochlea-scaled entropy. The results show that non-intrusive SRMR intelligibility algorithm with the traditional invasive method similar results, Because of dependence rid pure voice, non-invasive intelligibility evaluation algorithmshave more practical use.