A Speech Recognition Approach with MFCC and Fractal Dimension
A modified approach present here combines Mel-frequency cepstral coefficients (MFCC) and fractal dimension as mixed feature parameter to carry out the speech recognition.Due to the respective advantages on expressing speech signal of MFCC and fractal dimension, fractal dimension denotes the self-similarity, periodicity and randomness of speech signal; meanwhile MFCC feature parameter describes speech nonlinearity. Besides, BP neural network is introduced into the whole speech recognition training procedure. Experimental results demonstrate fractal dimension is able to reflect the speech feature to some extent,makeup the shortcoming of traditional speech feature. The recognition performance is then improved by introducing fractal dimension feature.
Speech Recognition MFCC Fractal Dimension neural network mixed parameter
Minghai Yao Jing Hu
College of Information Engineering, Zhejiang University of Technology Hangzhou, 310032, China
国际会议
杭州
英文
349-351
2006-10-12（万方平台首次上网日期，不代表论文的发表时间）