基于噪声估计和听觉掩蔽效应的语音增强

Speech enhancement based on noise estimation and auditory masking properties

针对非平稳噪声环境和低信噪比的情况,提出了一种基于低频区语音特性的非平稳噪声估计方法,通过构造一个时变的权值,实现对噪声的实时估计,同时结合人耳听觉掩蔽效应,利用估计出的噪声自适应设定增强系数。仿真实验表明,该方法能够较好地抑制背景噪声,提高信噪比,减少语音失真。

Aiming at non-stationary noise environment and low SNR,a non-stationary noise estimation method is proposed by using low-frequency regions of noisy speech,a weighting factor is used to track the time variation of noise.Then,the estimated noise can be used to self-adaptively adjust enhancement parameters according to auditory masking properties.The experiment result shows that the method is effective to reduce background noise,improve SNR and decrease speech distortion.