TR2007-062

Supervised and Semi-Supervised Separation of Sounds from Single-Channel Mixtures

    •  Paris Smaragdis, Bhiksha Raj, Madhusudana Shashanka, "Supervised and Semi-Supervised Separation of Sounds from Single-Channel Mixtures", Tech. Rep. TR2007-062, Mitsubishi Electric Research Laboratories, Cambridge, MA, July 2006.
      BibTeX TR2007-062 PDF
      • @techreport{MERL_TR2007-062,
      • author = {Paris Smaragdis, Bhiksha Raj, Madhusudana Shashanka},
      • title = {Supervised and Semi-Supervised Separation of Sounds from Single-Channel Mixtures},
      • institution = {MERL - Mitsubishi Electric Research Laboratories},
      • address = {Cambridge, MA 02139},
      • number = {TR2007-062},
      • month = jul,
      • year = 2006,
      • url = {https://www.merl.com/publications/TR2007-062/}
      • }
  • Research Areas:

    Artificial Intelligence, Speech & Audio

Abstract:

In this paper we describe a methodology for model-based single channel separation of sounds. We present a sparse latent variable model that can learn sounds based on their distribution of time/frequency energy. This model can then be used to extract known types of sounds from mixtures in two scenarios. One being the case where all sound types in the mixture are known, and the other being being the case where only the target or the interference models are known. The model we propose has close ties to non-negative decompositions and latent variable models commonly used for semantic analysis.