语音增强用于抗噪声的汉语说话人识别

Speech Enhancement Applied to Chinese Speaker Identification in Noisy Environments

大多数实际应用环境中总是存在各种各样的噪声,由于训练环境与识别环境不匹配,现有的绝大多数说话人识别系统在噪声环境中的性能都不可避免的急剧下降。为了让说话人识别系统在强噪声环境中,有较好的识别效果,研究一个将语音增强器和说话人识别系统级连起来的系统,该系统中将语音增强作为前端处理来提高输入的信噪比,实验证明,该系统具有很好的抗噪声性能。

This paper presents our recent work on a noisy speaker identification system. There are all kinds of noises in practical environments. Because of the unmatchment of identifying environments with training environments, the performance of most speaker identification systems degrades quickly, in our system, speech enhancement as the frontend processing module is used to improve the Signal-Noise Ratio of the input signal for speaker identification in the latter stages. Experimental results show that the system is satisfactory in noisy environments.