Wavelet-COH在通信语音质量评估算法中的应用研究

将小波变换用于子带分解,对纯净语音信号和受扰语音信号进行特征提取,包括质心、子带能量和带宽等,并进一步与一致性函数(COH)方法相结合对语音客观音质评价方法进行了研究,即Wavelet-COH方法。通过最小二乘多项式拟合模型,对Wavelet-COH方法得到的客观评测和主观评测结果进行相关分析,得出相应算法的相关系数和方差值。通过对比,表明Wavelet-COH语音评估方法比传统的COH客观评价方法有很大改善。

Application of Wavelet-COH in Communication Speech Quality Evaluation Algorithm

With the application of wavelet tranform in sub-band decomposition, the characteristics of both clean and jammed speech signals are extracted,such as centroid, sub-band energy and band width. These characteristics are further combined with COH (Coherence Function), i.e. Wavelet-COH, and applied to objective evaluation of speech quality. Based on least squares polynomial fitting model, the obtained results of both objective and subjective evaluations through Wavelet-COH are analyzed, and the relationship coefficient and error variance also acquired. Comparison indicates that the Wavelet-COH speech evaluation method enjoys even more significant improvement as compared with traditional COH objective evaluation method.