论文部分内容阅读
对基本的谱减算法进行了改进。利用满足听觉感知模型的滤波器组对语音信号进行处理,然后通过自动跟踪信号的低能量部分的包络来估计背景噪声的特性,最后利用改进的谱减技术对子带语音进行滤波并增强。实验表明,改进的谱减技术具有良好的性能。
The basic spectral subtraction algorithm has been improved. The speech signal is processed by a filter bank satisfying the auditory perception model, and then the characteristics of the background noise are estimated by automatically tracking the envelope of the low-energy part of the signal. Finally, the improved spectral subtraction technique is used to filter and enhance the sub-band speech. Experiments show that the improved spectral subtraction technique has good performance.