低信噪比语音信号增强处理方法的研究

发布时间：2018-11-07 13:58

【摘要】：语音是人类传播信息和交流的重要媒介,人们可以在多个讲话者的环境中获得自己感兴趣的语音信号,这是人耳特有的一种功能。如何从被噪声环境污染的语音信号中恢复出原始语音信号,是现代语音信号处理的一个非常重要的问题。小波阈值法是语音增强中常见的且效果较好的一种算法。现阶段较多的增强算法在信噪比较高、且噪声平稳的情况语音增强效果较好,但在信噪比较低的非平稳噪声情况下,增强效果并不理想。本文主要针对带噪语音信号进行增强处理,由于高信噪比、平稳噪声污染的语音信号的去噪问题比较容易实现,所以本文算法主要是针对信噪比较低的非平稳噪声污染的语音信号的增强处理,论文的主要工作包括：提出了结合小波阈值的MMSE-LSA低信噪比语音信号去噪算法,这种算法可以有效提高语音信噪比,增强语音可懂度。提出了将原始语音经过上述结合小波阈值的MMSE-LSA低信噪比语音信号去噪算法处理后,作为先验信息,进行优化的卡尔曼滤波处理,进一步提高语音去噪增强的效果,并通过仿真验证了算法性能。仿真实验表明,论文所提算法不仅可以有效去除语音信号中的噪声,而且提高了算法的实用性。
[Abstract]:Speech is an important medium for human communication and communication. People can get their own interesting voice signal in the environment of multiple speakers, which is a special function of human ear. How to recover the original speech signal from the speech signal contaminated by noise environment is a very important problem in modern speech signal processing. Wavelet threshold method is a common and effective algorithm in speech enhancement. At present, more enhancement algorithms have higher SNR and better speech enhancement effect when the noise is stable, but the enhancement effect is not satisfactory in the case of non-stationary noise with low signal-to-noise ratio (SNR). This paper mainly focuses on the enhancement of noisy speech signal. Due to the high SNR and stationary noise pollution, the de-noising problem of speech signal is easy to realize. So this algorithm is mainly aimed at the enhancement of non-stationary noise pollution speech signal with low signal-to-noise ratio. The main work of this paper is as follows: the MMSE-LSA low SNR speech signal de-noising algorithm combined with wavelet threshold is proposed. This algorithm can effectively improve the signal-to-noise ratio of speech and enhance speech intelligibility. After the original speech is processed by the MMSE-LSA low signal-to-noise ratio speech signal denoising algorithm combined with wavelet threshold, the optimized Kalman filter is used as prior information to further improve the effect of speech denoising enhancement. The performance of the algorithm is verified by simulation. The simulation results show that the proposed algorithm can not only effectively remove the noise in the speech signal, but also improve the practicability of the algorithm.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN912.3

【共引文献】