Acceleration of rank-constrained spatial covariance matrix estimation for blind speech extraction

Abstract

In this paper, we propose new accelerated update rules for rank-constrained spatial covariance model estimation, which efficiently extracts a directional target source in diffuse background noise. The naive update rule requires heavy computation such as matrix inversion or matrix multiplication. We resolve this problem by expanding matrix inversion to reduce computational complexity; in the parameter update step, we need neither matrix inversion nor multiplication. In an experiment, we show that the proposed accelerated update rule achieves 87 times faster calculation than the naive one.

Blind Speech Extraction Based on Rank-Constrained Spatial Covariance Matrix Estimation With Multivariate Generalized Gaussian Distribution

“…The conclusions of this paper are presented in Section VII. Note that this paper is partially based on international conference papers [28], [29] written by the authors. Additional contributions of this paper are that we generalize the statistical model using the multivariate GGD, employ the majorization-minimization (MM) [30] and majorizationequalization (ME) [31] algorithms for better parameter optimization, and conduct BSE experiments under extended acoustic conditions including real recorded data.…”