CS134 Project Proposal

Audio Source Separation

My email: Xiaoyu.Zhao@dartmouth.edu

Goal of Project, Method, Data

My project will focus on blind audio source separation. The data will be collected by two microphones at different locations in a room when two audio source is simultaneously available. The two audio source can be a background music and a person talking, or two person talking. The machine learning method I will use to achieve the goal is Independent Component Analysis. The key idea of ICA is to find a linear representation of non-gaussian data so that the components are statistically as independent as possible. This method is used to capture the structure of the audio signal. The key algorithm I will use is FastICA. The fastICA is based on a fixed-point iteration scheme for finding the maximum of the nongaussianity of the linear combination of the microphone record if considering them as a vector. I will try Image segmentation with this algorithm if it works well.

Timeline:

  • Before May 12, 2009 Read papers about ICA and implement FastICA algorithm.
  • Before Final Debugging the algorithm and test the correctness of the algorithm.
  • Collect data. Do preprocessing (Sampling, centering, whitening).
  • Apply FastICA to the data to separate the source. Analysis results. Write the final report.

    • Reference

  • Aapo Hyvarinen and Erkki Oja, Neural Networks Research Center Helsinki University of Technology, Independent Component Analysis: Algorithm and Applications.
  • Hiroshi Saruwatari , Blind Source Separation Combining Independent Component Analysis and Beamforming
  • H.Farid E.H.Adelson, Separating reflections from images by use of independent component analysis. Jornal of the optical society of america. 16(9):2136-2145, 1999