Improved Laplacian Factor Estimation for Noisy Speech Enhancement

Two novel approaches for Laplacian factor estimation are proposed for noisy speech enhancement employing Laplasian-Gaussian mixture model in discrete cosine transform domain. Based on the property of generalized Gaussian distribution model, the proposed approaches indirectly attain the estimation of Laplacian factor using its relationship with the variance of clean speech components under the Laplacian distribution assumption. The estimated results for Laplacian factor are accurate and uneffected by the noise energy, while keeping the resulting method simple. In simulations with speech signals degraded by different kinds of noise, the experimental results show the improved performance of the proposed algorithms.