Wavelet Package Based Speech Enhancement Algorithm Using Time-Frequency Threshold
基于时频阈值的小波包语音增强算法

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.