摘要
说话人识别系统在实际应用中面临的主要困难之一是鲁棒性问题,干净语音环境下识别率很高的说话人识别系统,在有噪语音环境下识别性能显著降低。解决这一问题的方法之一是寻找具有鲁棒性的特征参数。本文结合具有多分辨率分析特点的小波变换技术,提出一种基于小波变换的鲁棒型特征提取算法,以提高说话人识别系统在噪声环境下的识别性能。对40个说话人的语音库SUDA2002-D2,在加性高斯白噪声环境下进行的识别实验结果表明,本文提出的特征提取算法可以有效地提高说话人识别系统在噪声环境下的识别性能。
One of difficulties in application of speaker recognition system is robust problem. A speaker recognition system with high performance in relatively clean environment will become deficient with unacceptable recognition performance in noisy environment. One method to solve this problem is to detect robust features against noises. In this paper, a new robust feature-extraction algorithm based on wavelet transform is proposed. Benefit from its multi-resolution analysis abilities, the cepstrum features detected from several different time-frequency channels are integrated with a statistical entropy values. Experiments on SUDA2002-D2 Chinese speech corpus show that the proposed algorithm is quite efficient for speaker identification in noisy environment.
出处
《电路与系统学报》
CSCD
北大核心
2005年第5期129-132,共4页
Journal of Circuits and Systems
关键词
说话人识别
鲁棒型特征
小波变换
矢量量化
speaker identification
robust features
wavelet transform
vector quantization