一种基于语音端点检测的维纳滤波语音增强算法

摘要

在基于先验信噪比的维纳滤波语音增强算法的基础上，结合语音端点检测算法，本文提出一种新算法。新算法在语音端点检测的基础上，通过平滑处理更新噪声信号功率谱以适应噪声不稳定的环境；通过计算有声段噪声信号估计值，将有声段的噪声影响纳入考虑范围；通过每个语音段自适应调节噪声功率谱，实时的计算出先验信噪比。最后将该算法与改进前算法进行仿真比较验证，该算法有更好的语音增强效果，在非稳定噪声环境中较好的抑制了噪声残留，提高了语音的可懂度。%Based on the Wiener filtering speech prior SNR enhancement algorithm, this paper proposes a new algorithm combined with the speech endpoint detection algorithm. Based on the speech endpoint detection algorithm, the new algorithm updates the power spectrum of the noise signal by smoothing the power spectrum in order to adapt to the noise and unstable environment; The new algorithm takes the influence of the noise into account by calculating the estimate value of the sound segment noise, and calculates the priori SNR in real time by each sound segment adaptive noise power spectrum. Finally, compare the algorithm and improved algorithm by simulation verification, the improved algorithm suppresses the residual noise in the unstable noise environment, and improves the speech intelligibility.