一种自适应语音增强均衡器的研究

Research on an adaptive equalizer for speech enhancement

在参考滤波器组的基础上,提出一个用于语音和音频信号进行时不变或自适应谱修正的数字滤波器结构,主要用于音频信号均衡和降噪。在频域计算滤波器系数的同时,信号在时域滤波。与通常频域处理相比,在信号延迟、原型滤波器设计、复杂性等方面,该结构有良好的特性。该算法既适用于均匀频率分辨率,也适用于非均匀频率分辨率。

In this paper a digital filter structure is proposed for fixed or adaptive spectral modification of speech and audio signals, which is derived from a reference filter bank. Main applications are audio equalization and noise reduction. The signal is filtered in the time domain, while the filter coefficients are calculated in the frequency domain. In comparison to the usual spectral-domain processing, the proposed structure exhibits distinct advantages with respect to signal delay, prototype-fiher design, and complexity. The algorithm allows for uniform, as well as for non-uniform spectral resolution.