earticle

영어

This study proposes a post-processor to reduce the effect of musical residual noise which is annoying to the human ear. Initially, a state-of-the-art speech enhancement algorithm is performed as the first stage to reduce background noise for noisy speech. Hence the enhanced speech is post-processed by a hybrid-mean filter to reduce the musical effect of residual noise. In the case of a vowel-like spectrum, directional-mean filtering is performed to slightly reduce the musical effect of residual noise, where the harmonic spectrum can be well maintained at an acceptable level. Conversely, block-mean filtering is performed to heavily reduce the spectral variation for noise-dominant spectra, enabling musical tones to be significantly smoothed. The musical effect of residual noise is therefore reduced. Finally, the pre-processed, the directional-mean filtered and the block-mean filtered spectra are fused according to speech-presence probability. Experimental results show that the proposed hybrid-mean filter can efficiently improve the performance of a speech enhancement system by reducing the musical effect of residual noise.