MMSE-Based speech enhancement using the harmonic model

Modern speech and audio signal processing applications deal with signals contaminated with high noise level. Thus, the problem of speech and audio enhancement is of great importance in speech and audio signal processing. In this paper, a new method for speech enhancement using the harmonic model for voiced frames is proposed. In this method, the fundamental frequency is estimated via the maximum a-posteriori probability (MAP) tracking, while the harmonic amplitudes are estimated via the minimum-mean-square error (MMSE) estimator, which utilizes prior statistical information of the speech phonemes. The MMSE estimator is implemented using the Monte Carlo method. The performance of the proposed method is evaluated by the perceptual evaluation of speech quality (PESQ) algorithm. It is shown that the method outperforms other speech enhancement methods in low signal-to-noise ratios (SNRs).