The Approximation of Motor Speech Instruction based on Wavelet Transform and FVEZ
Considering processes of speech recognition in a moving automotive, we need to deal with complicated background noises. A number of normal de-noising preprocessing and speech extraction method is not appropriate for such a low SNR condition. In this paper a new automotive speech instruction approximation approach based on wavelet transformation and FVEZ was presented. Its a good solution to this kind of problems. The method is first of all to choose a different wavelet function such as Haar, dbN, coifN and so on, then to decompose the noisy speech signal with level six, then according to FVEZ choose special decomposition level of wavelet coefficients, and to reconstruct the speech using the inverse wavelet function. The experimental results show that the method can be used to approximate the useful speech instruction signal from the lower SNR motor speech signal well, thus improve the EPD accuracy rate and the speech recognition rate.
wavelet transform automotive speech instruction De-noising FVEZ approximation
Haiyan SHI
Computer center, Shaoxing University, Shaoxing, 312000 P. R. China
国际会议
The Third International Conference on Modelling and Simulation(第三届国际建模、计算、仿真、优化及其应用学术会议 ICMS 2010)
无锡
英文
151-154
2010-06-04（万方平台首次上网日期，不代表论文的发表时间）