Post-processing in masking-based β-order MMSE speech enhancement

A post-processing technique for improving the performance of the masking-based β-order MMSE method in speech enhancement is proposed in this letter. After the enhancement process, weak spectral components in the high-frequency band are usually over-attenuated. A non-linear high-frequency regeneration technique is used to re-synthesize the upper-band signal based on the lower-band speech.