Single channel speech enhancement using MDL-based subspace approach in Bark domain

We present a novel algorithm for single channel speech enhancement. It is based on a subspace approach in the Bark domain and an optimal subspace selection by the minimum description length (MDL) criterion. The processing in the Bark domain allows us to take into account in an optimal manner the masking properties of the human auditory system. The subspace selection provided by the MDL criterion overcomes the limitations encountered with other selection criteria, like the overestimation of the signal-plus-noise subspace or the need for empirical parameters. Together, the resulting MDL-subspace approach in the Bark domain provides maximum noise reduction while minimizing the signal distortions. The performance of our algorithm is assessed in white and colored noise. It shows that our algorithm provides a high performance for a large scale of input signal-to-noise ratio