论文部分内容阅读
由于人耳频率分辨率是非线性的 ,用传统的线性信号处理方法 (如FFT)来模拟人耳基底膜的频率分析特性是比较困难的 .小波包算法有灵活的时频分析能力 ,可较好地符合人耳基底膜的频率分析特性 .在模拟人耳的听觉机理方面 ,用动态阈值法成功地对含噪语音进行了去噪处理 ,在去噪处理中引入音乐噪声的问题也较好地得到解决 .实验表明 :在单声道的条件下 ,其语音增强效果比传统的频谱减法有更高的清晰度和可懂度
Because human ear frequency resolution is non-linear, it is difficult to simulate the frequency analysis of human ear basement membrane by traditional linear signal processing methods (such as FFT) .Wavelet packet algorithm has the advantage of flexible time-frequency analysis In line with human ear basement membrane frequency analysis characteristics.In the simulation of the human ear auditory mechanism, the dynamic threshold method to successfully carry out the noisy speech noise processing, noise reduction in the introduction of music noise problem is also better The experimental results show that the speech enhancement effect is more sharp and intelligible than the traditional spectral subtraction under mono conditions