摘要
在分析MFCC提取原理的基础上,结合小波包分析理论,得到新的特征参数.提出了一种新的特征参数提取方法,用动态时间规整法,分别测试了MFCC的识别率和新的特征参数的识别率.研究证明新的特征参数不仅具有较高的识别率,而且有一定的抗噪声能力.
In the paper, a method of new feature factor extraction is proposed. Based on the MFCC extraction theory and combining with wavelet packet analysis, a new feature factor is obtained. Using DTW, we have tested the speech recognition rate of MFCC and the new feature factor. The experiment suggests that the new feature factor achieves not only high recognition rate, but also anti-noise capacity.
出处
《宁波大学学报(理工版)》
CAS
2007年第1期51-54,共4页
Journal of Ningbo University:Natural Science and Engineering Edition
基金
浙江省自然科学基金(104144)
宁波市博士基金(2005A610003).
关键词
小波包
美尔倒谱系数
特征参数
动态时间规整
wavelet packet
Mel frequency cepstral coefficient
feature coefficient
dynamic time warping