说话人识别中语音切分算法的研究

Research on Speech Segmentation Algorithm in Speaker Recognition

论文针对说话人识别中语音能量变化和噪声对提取有效语音数据的影响,在传统时域语音切分算法犤1,3犦的基础上,提出了三种孤立词的精确切分算法和一种连续语音的非精确切分算法。实验表明,新算法较好地克服了语音能量变化对切分的影响,在原始语音具有较高信噪比(≥10dB)的情况下,能够切除某些短时噪声和白噪声犤2犦。

In this paper,three word-segmentation algorithm and one continual speech segmentation algorithm are pre-sented to avoid the effect caused by speech energy shifting and noise in speech data.The novel methods can overcome the limitation of the conventional speech segmentation method,and also have some effect on cutting off short-time noise and white noise.