Group sparsity for MIMO speech dereverberation

Ante Jukić1, Toon van Waterschoot2, Timo Gerkmann1, Simon Doclo1

1 University of Oldenburg, Dept. of Medical Physics and Acoustics, Oldenburg, Germany
2 KU Leuven, Dept. of Electrical Engineering Leuven, Belgium

Description

In this paper we present a batch algorithm employing a signal model based on multi-channel linear prediction in the short-time Fourier transform domain. Aiming to achieve multiple-input multiple-output (MIMO) speech dereverberation in a blind manner, we propose a cost function based on the concept of group sparsity. The samples given below demonstrate the application of the algorithm for binaural dereverberation.


Configuration

  • M = 4 microphones: 2 behind-the-ear (BTE) hearing aids, 2 microphones per hearing aid
  • RIRs: from Database of multichannel in-ear and behind-the-ear head-related and binaural room impulse responses, University of Oldenburg.
  • Environment: Cafeteria
  • Source: position E, position F (see below). Noise: cafeteria ambient noise

  • Parameters

  • Sampling frequency fs = 16 kHz
  • STFT: 64 ms square root Hamming window, 16 ms shift
  • MCLP: length of the filters Ls = 15, number of iterations itmax=10

  • Position E

    Clean RSNR Mic 1 Out, p=0 Out, p=1/2 Out, p=1 Out, p=2
    40 dB
    20 dB
    10 dB

    Position F

    Clean RSNR Mic 1 Out, p=0 Out, p=1/2 Out, p=1 Out, p=2
    40 dB
    20 dB
    10 dB

    Configuration detail

  • M = 4 microphones: 2 behind-the-ear (BTE) hearing aids, 2 microphones per hearing aid
  • Environment: Cafeteria
  • Source: position E, position F. Noise: cafeteria ambient noise

  • Cafeteria