Speech Signal Enhancement Based on Adaptive Multi-resolution Form of SVD

A speech signal enhancement method based on adaptive multi-resolution form of singular value decomposition (MSVD) is proposed in this paper. The residual noise after spectral subtraction, which results in audible musical noise, is reduced further by MSVD filter. In comparison with the original singular value decomposition filter (SVD), MSVD filter reduces the computational burden, which is very important for real time speech processing situations.