摘要
语音信号在实际场合中通常容易受噪声的干扰,较强的噪声会降低语音识别的效率,为了使语音信号可懂,需要对噪声进行处理。本文先分析了语音和噪声的特性,介绍了基本的几种语音降噪的算法,以及这些算法实现的仿真结果对比。然后提出了一种对语音信号进行分帧处理,然后采用谱熵和维纳法进行减噪和语音增强的方法,在算法处理的同时,还需考虑语音质量的问题。
Speech signals are usually subject to noise interference in practical situations. Strong noise will reduce the efficiency of speech recognition. In order to make speech signals understandable, noise processing is required. This paper first analyzes the characteristics of speech and noise, introduces several basic algorithms for speech noise reduction, and compares the simulation results of these algorithms. Then, a frame processing method for speech signal is proposed, and spectral entropy and wiener method are used to reduce noise and enhance speech.
引文
[1]易克初,田赋,付强.语音信号处理[M].北京:国防工业出版社,2000.
[2]赵力.语音信号处理[M].北京:机械工业出版社,2003.
[3]刘淑华.低信噪比下的语音增强技术研究[D].广西:广西大学.
[4]S. Boll. A spectral subtraction algorithm for suppression of acoustic noise in speech[C].International Conference on Acoustics, Speech, and Signal Processing, Washington, USA, 1979,ASSP-27(4):200-203.
[5]M.Berouti, R. Schwartz, J. Makhoul. Enhancement of speech corrupted by acoustic noise[C]. International Conference on Acoustics, Speech, and Signal Processing, Washington, USA, 1979, ASSP-27(4):208-211.
[6]李晔,崔慧娟,唐昆.基于谱减的语音增强方法的改进[J].清华大学学报(自然科学版).2006, 46(10):1685-1687.
[7]张宗平,刘贵忠等,“基于二进制小波变换的信号去噪”,西安交通大学研究报告,1999.
[8]马丽丹.基于小波和OM-LSA的语音增强[D].长春:吉林大学, 2008.
[9]RAOCVR,MURTHYMBR,RAOKS.Speech enhancement using perceptual Wiener filter combined with unvoiced speech-A new Scheme[J].IEEE. Intelligent Computational Systems.2011,9,688-691.
[10]VDB Tim, D Simon, W Jan, et al. Speech enhancement with multichannel Wiener filter techniques in multimicrophone binaural hearing aids[J]. Journal of the Acoustical Society of America, 2009, 125(1):360-371.
[11]Ephraim Y, van Trees H L.A signal subspace approach for speech enhancement[J].IEEE Transaction Speech and Audio Processing, 1995,3(4):251-266.