An adaptive β-order MMSE estimator for speech enhancement using super-Gaussian speech model

This paper proposes an adaptive β-order Minimum-Mean-Square-Error (MMSE) estimator for speech enhancement using super-Gaussian speech model (β-SG-MMSE). The spectral amplitude of clean speech is estimated by MMSE estimator under the assumption that the DFT coefficients of clean speech are modeled by super-Gaussian distribution and the DFT coefficients of noise signal are modeled by Gaussian distribution. Then, the speech presence probability under super-Gaussian model is introduced into the proposed estimator. In order to obtain a good trade-off between noise suppression and speech distortion, the order β of estimator is updated adaptively according to the Signal-to-Noise Ratio (SNR) in each sub-band. The result of performance evaluation by ITU-T G.160 shows that the overall performance of the proposed method is better than the reference algorithms in white noise and color noise.