Télécom Paris

Combined Beamforming and Frequency Domain ICA for Source Separation
  • Nilesh Madhu, (Institute of Communication Acoustics, Ruhr-Universitaet Bochum)
  • André Gückel, (Institute of Communication Acoustics, Ruhr-Universitaet Bochum)
  • Rainer Martin, (Institute of Communication Acoustics, Ruhr-Universitaet Bochum)
  • Sound enhancement and sound separation
  • Microphone arrays and array signal processing
  • Adaptive filtering algorithms and structures for echo and noise control
Get the paper in PDF format
Acrobat Reader (version 5 minimum) is necessary to read this document.

In this paper we propose a combined approach using beamforming and frequency domain ICA for the source separation problem in reverberant environments. The proposed method gains, and makes use of, the location of (any one) target source for permutation resolution. This approach is computationally less intensive and yields results comparable to current, state-of-the-art approaches. Additionally, we propose and justify the use of a non-linear post-processor to further improve crosstalk suppression in the recovered signals.

©2006 Télécom Paris/TSI
Edition : Télécom Paris -- 2006