论文部分内容阅读
在实际应用中,噪声干扰导致语音识别性能急剧下降。针对该问题,本文分析传统方法并提出相应的系统解决方案:采用小波变换对语音信号进行前端处理,以MFCC声道特征结合基频(F0)韵律特征来提高识别系统的鲁棒性。实验结果表明:小波变换能有效地消除噪声影响,经小波降噪处理后,使得F0-MFCC联合模型能更好的识别语音。可以看出在噪声环境下系统的综合性能得到很大改善。
In practical applications, the noise interference leads to a sharp drop in speech recognition performance. In order to solve this problem, this paper analyzes the traditional methods and proposes corresponding system solutions: the wavelet transform is used to front-end the speech signal, and the robustness of the recognition system is improved by combining the MFCC channel features with the fundamental frequency (F0) rhythm features. Experimental results show that the wavelet transform can effectively eliminate the influence of noise. After wavelet denoising, the F0-MFCC joint model can better recognize the speech. It can be seen in the noise environment, the overall performance of the system has been greatly improved.