论文部分内容阅读
针对目前语音增强存在较大时延的问题,提出一种低时延语音增强算法。在传统的先验信噪比估计和噪声估计的基础上,对判决引导算法进行低时延的改进,并提出了一种基于语音存在概率和语音激活检测相结合的噪声估计方法,本文的算法采用对数MMSE估计器结合语音存在概率。采用ITU-T P.826 PESQ、分段信噪比、总信噪比和对数谱失真对该算法进行了测试,并与其他几种算法进行了对比,实验结果表明,该算法有效降低了时延,可以很好的跟踪非平稳噪声,在信噪比较低的情况小可以取得很好的增强效果,且音乐噪声和残留背景噪声也可以得到很好的抑制。
Aiming at the problem of large delay in speech enhancement, a low-latency speech enhancement algorithm is proposed. Based on the traditional a priori SNR estimation and noise estimation, the decision-directed algorithm is improved with low delay, and a noise estimation method based on the combination of speech presence detection and voice activation detection is proposed. The proposed algorithm Logarithmic MMSE estimator is used in conjunction with speech presence probability. The algorithm is tested by ITU-T P.826 PESQ, segmented signal-to-noise ratio, total signal-to-noise ratio and logarithmic spectrum distortion, and compared with other algorithms. Experimental results show that the proposed algorithm effectively reduces The delay can well track the non-stationary noise, and can obtain very good enhancement effect when the signal-to-noise ratio is low, and the music noise and the residual background noise can also be well suppressed.