基于时频阈值的小波包语音增强算法  

Wavelet Package Based Speech Enhancement Algorithm Using Time-Frequency Threshold

作　　者：徐耀华[1] 王刚[1] 郭英[1] 

机构地区：[1]空军工程大学电讯工程学院

出　　处：《电子与信息学报》2008年第6期1363-1366,共4页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(60601016);陕西省自然科学基金(2006F40);辽宁省高校重点实验室开放基金(2006-04)资助课题

摘　　要：该文考虑小波域应用语音降噪中听觉掩蔽效应,提出了一种基于时频阈值的小波包语音增强算法。新算法首先通过频域增强方法得到语音粗估计,通过跟踪估计语音时频特性的细节变化,及时调节降噪阈值,然后利用时频阈值对小波包系数进行处理,以达到语音降噪的目的。实验表明,较传统小波域语音降噪方法,新算法在抑制平稳白噪声的同时减小了语音信息的损失,其增强语音的MOS(Mean Opinion Score)评分、输出信噪比、MBSD(Modified Bark Spectral Distortion)测度性能均有明显提高。Incorporating masking properties of the human auditory system in wavelet domain, this paper proposed a new algorithm of wavelet package speech enhancement based on the time-frequency threshold. New algorithm first obtains speech pre-estimation by frequency-based de-noising method, then, via tracing the variation of time-frequency information of the speech pre-estimation, the threshold is modulated adaptively. Finally, the noisy speech is de-noised by means of time-frequency thresholding the coefficients of the wavelet package. With comparing to the traditional wavelet algorithms, the proposed algorithm offers more pleasant enhanced speech with less distortion and residual noise in the additive Gaussian noise environments, and the experimental results demonstrate its better performance in Subjective test, input and output SNR test, and Modified Bark Spectral Distortion （MBSD） measurement tests.

关 键 词：语音增强 小波 阈值 时频 

分 类 号：TN912.3[电子电信—通信与信息系统]