Research of Speech Enhancement Based on Selection of Wavelet Base Function
Due to a new multi-resolution characteristics, wavelet transform can be carried out in time domain and frequency domain analysis at the same time, it has characteristics of localized time-frequency and multiresolution, is particularly suited to handle the nonstationary signals. Wavelet is got by which wavelet function was translated and stretched, it has the characteristics of simple, flexible and random and has the function of multi-resolution analysis. Since different wavelets have different orthogonal degrees, and vanishing moments, support and so on are not the same, different selections of the wavelet functions will result in different effects. This paper describes the researches of property and screening for wavelet function by means of computer simulation experiments of speech enhancement, and studies the effects in the different treatment produced by different wavelet function. We had more understandings with the performance of each wavelet function through these experiments. This paper also proposes a solution to realize variable threshold based on the voiceless and the voiced of voice. The simulation results show that the use of these methods can not only raise the SNR of voice signal, but also improve the degrees of voice distortion to a large extent. It has a certain guiding significance in practice.
XU YAN LIU JUN
School of Electronic and Information Engineering Lanzhou Jiaotong University Lanzhou ,China
国际会议
上海
英文
278-282
2010-10-20（万方平台首次上网日期，不代表论文的发表时间）