摘要
语音识别系统中,语音的特征提取是语音识别的关键技术之一。通过对语音的系统研究,提出一种全新的基于流形学习的特征提取方法。流形算法是近些年才发展起来的非线性降维方法,在人脸识别领域已取得较好效果,但在语音识别领域一直处于空白。现提出的基于流形学习LPP算法的语音特征提取方案,是一次重大的尝试,可以为以后深入研究语音识别技术提供较好参考。仿真实验结果表明,该算法与传统特征提取LPCC、MFCC算法相比,可以取得较好的识别率。
Speech feature extraction is one of the key technologies in speech recognition systems through systematic study of phonetic system, a new feature extraction based on manifold learning is proposed. Man- ifold algorithm, as a non-linear dimension reduction method developed in recent years, has achieved fairly good results in facial recognition field, but is still nonexistent in the field of speech recognition. However, the newly-proposed scheme of speech feature extraction based on LPP algorithm of the manifold learning is a significant try and may provide a good reference for further study of speech recognition technology. The simulation experiment shows that this algorithm has better recognition rate as compared with LPCC, MFCC algorithms of traditional feature extraction.
出处
《通信技术》
2013年第12期15-18,共4页
Communications Technology
关键词
流形学习
语音识别
特征提取LPP算法
manifold learning
speech recognition
feature extraction
LPP algorithm
