TR2008-012

Speech Denoising Using Nonnegative Matrix Factorization with Priors


    •  Wilson, K.W.; Raj, B.; Smaragdis, P.; Divakaran, A., "Speech Denoising Using Nonnegative Matrix Factorization with Priors", IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), ISSN: 1520-6149, March 2008, pp. 4029-4032.
      BibTeX Download PDF
      • @inproceedings{Wilson2008mar,
      • author = {Wilson, K.W. and Raj, B. and Smaragdis, P. and Divakaran, A.},
      • title = {Speech Denoising Using Nonnegative Matrix Factorization with Priors},
      • booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
      • year = 2008,
      • pages = {4029--4032},
      • month = mar,
      • issn = {1520-6149},
      • url = {http://www.merl.com/publications/TR2008-012}
      • }
  • Research Areas:

    Multimedia, Speech & Audio


TR Image
A simple example showing the advantage of regularizing with the log likelihood. In each panel, the horizontal axis represents time and the vertical axis represents frequency. Darker colors represent higher intensity. The leftmost column shows the original signals.

We present a technique for denoising speech using nonnegative matrix factorization (NMF) in combination with statistical speech and noise models. We compare our new technique to standard NMF and to a state-of-the-art Wiener filter implementation and show improvements in speech quality across a range of interfering noise types.