论文部分内容阅读
介绍了一种双通道的噪声抵消系统 .传统的频谱减法的语音增强需要预先知道噪声的特性 .但是 ,噪声的平稳性和可分离性在许多实际应用场合都无法满足要求 .提出采用两路话筒输入 ,以便动态获得噪声频谱的方法 .采用自适应校正滤波的同步算法 ,利用语音信号的相关性同步两路信号中的语音 ,通过抵消其中的语音成分获得噪声的频谱特性 .实验证明该方法能够获得很好的同步效果 .在获得噪声频谱的基础上用频谱减和维纳滤波两种方法进行了噪声抵消的对比实验 .对 0dB信噪比输入的信号可以获得约 9dB信噪比的提高 .用维纳滤波处理后的语音比频谱减法残留较低的“音乐声” .在非平稳条件下对噪声仍具有一定的跟踪能力
A two-channel noise cancellation system is introduced, the traditional spectral subtraction speech enhancement needs to know the characteristics of noise in advance, however, the stability and separability of noise can not meet the requirements in many practical applications. Input, in order to dynamically obtain the noise spectrum.Adopting the adaptive correction filtering synchronization algorithm, the use of the correlation of the speech signal to synchronize the two signals in the speech, by offsetting the speech components to obtain the noise spectral characteristics.Experiments show that the method can Get a very good synchronization effect.On the basis of obtaining the noise spectrum, we use the spectral subtraction and the Wiener filter two methods to do the experiment of noise cancellation.It can improve the signal to noise ratio of about 9dB to the 0dB SNR input signal. The Wiener filtered speech has a lower residual “music sound” than the spectral subtraction, and still has certain tracking ability to noise under non-stationary conditions