基于时域滤波多频段谱减法的语音增强

传统的多频段语音增强是将频域分割成各个频段,并对每个频段单独进行语音增强,但是因为相邻频带的共振峰影响清晰语音估计,导致了语音识别效果不理想,为了减少相邻频带的共振峰带来的影响,提出了一种时域滤波中多频带语音增强的方法,通过将时域中的未处理语音过滤成各种等效的基于矩形带宽的子带,然后在每个频带中使用基于离散余弦变换(DCT)谱减法来估计清晰语音,并结合使用各个频带信噪比(SNR)获得频带特定加权因子。在SNR为0~10d B的汽车噪声、餐厅噪声、列车噪声、白色噪声和工厂噪声的环境下基于时域多频段语音增强算法增强效果优于现有的技术。

Speech Enhancement Based on Time Domain Filtering Multi-band Spectrum Subtraction

The traditional multi-band speech is enhanced due to the problem of the pole interaction between the various frequency bands,which leads to the unsatisfactory speech recognition effect. A method of multi - band speech enhance-ment in time domain filtering is proposed,clean speech is estimated by filtering unprocessed speech in the temporal do-main into various equivalent rectangular bandwidth based subbands followed by discrete cosine transform (DCT) based spectral speech enhancement in each band using spectral subtraction and incorporates band-specific weighting factor ob-tained using respective band signal-to-noise ratio (SNR). It is observed that DCT-derived spectrum based tempo-ral-domain multiband speech enhancement algorithm outperforms the existing techniques for car, babble, train, white, and factory noise in the 0–10 dB SNR levels.