Musical-Noise-Free Speech Enhancement with Low Speech Distortion by Biased Harmonic Regeneration Technique

摘要

Many noise reduction techniques generate two frequently occurring problems: speech distortion and musical noise. A number of methods have been proposed to solve these problems. One, which addresses the former problem, is harmonic regeneration noise reduction (HRNR). Using restored signals, HRNR regenerates the harmonics that were eliminated. Another, which addresses the latter problem, is a musical-noise-free noise reduction method based on the minimum mean-square error short-time spectral amplitude estimator. The method can suppress noisy speech without generating musical noise. This paper describes a new noise reduction technique we propose that combines these two methods. The technique is shown to be effective in suppressing both speech distortion and musical noise generation.