摘要
经验模态分解是一种新的信号分解方法,该方法可将非线性非平稳信号分解成若干个单分量的本征模态函数,使得每个本征模态函数都具有一定的物理意义。本文探索了该方法在语音增强方面的应用.在文献[8]的基础上,对其方法进行了有效改进。首先将带噪语音进行经验模态分解,得到六个本征模态函数和一个余量信号,对这七个信号分别进行小波阈值滤波,并由滤波后的七个信号重构语音。结果表明,该方法的滤波效果明显优于对带噪语音直接采用小波阈值滤波的方法,并且较之文献[8]的滤波方法也具有一定的优势。
Empirical mode decomposition is a new method for signal analysis. Using this method,nonlinear and nonstationary signals can be decomposed into several intrinsic mode functions, and each IMF has its own physical meaning. The paper discusses the application of this method in speech enhancement. Based on literature[ 8 ] ,this paper has improved the method in literature [ 8 ]. Firstly, decompose the noisy speech into six IMFs and a residual signal with the EMD method;secondly, apply soft threshold with wavelet transform to each signal and get seven new signals;at last rebuild the speech from the seven new signals. Experiments show that this method is more effective than using soft threshold method with wavelet transform and the similar method proposed in paper[ 8 ].
出处
《信号处理》
CSCD
北大核心
2008年第2期237-241,共5页
Journal of Signal Processing
基金
国家高技术研究发展计划(2006AA01Z146)专项经费资助
关键词
经验模态分解
本征模态函数
小波变换
Empirical Mode Decomposition
Intrinsic Mode Function
Wavelet Transform
