帧同步混合小波包变换模拟听觉模型的语音增强的研究

The research on speech enhancement based on the simulation of auditory model using frame-synchronized combined wavelet packet transform algorithms

摘要: 首先介绍了帧同步混合小波包的分析方法。该方法结合了小波包时频窗口可变的特点和STFT的分帧处理形式。它既能够保证语音信号处理中帧长的要求,即可实时处理,义能获得对信号频域上的最佳分解,是一种类似FFT的小波包的快速算法。在此基础上,应用该方法模拟了听觉模型,并运用于语音增强。实验表明,即使在-5 dB低信噪比的条件下,也能获得良好的除噪效果和听觉效果。该方法还可运用于语音的编码、合成和识别等领域。

Abstract: The Frame-synchronized wavelet packet transform algorithms are introduced in the paper. The algorithms not only have the advantages of the variable time-frequency windows of waelet packet, but also can analyse signals in a STFT-like frame-by-frame way. So they not only can reduce the analysis time-window length to fit the requirement of real-time speech analysis, but also can perform the best analysis in frequency domain, and meanwhile they are the fast algorithms as well as the FFT algorithm. We have simulated auditory model by using the algorithms and applied them to speech enhancement. That the better result has been obtained at SNR= -5 dB has been shown in the experiments. The algorithms also can be used for speech coding, synthesis and recognition.