Accelerated Convolutive Transfer Function-Based Multichannel NMF Using Iterative Source Steering

TL;DR

The paper tackles overdetermined blind source separation in reverberant acoustics by improving the computational efficiency of convolutive transfer function-based MNMF (CTF-MNMF). It replaces the expensive matrix-inversion-based iterative projection (IP) updates with an inversion-free iterative source steering (ISS) approach, while keeping a MM-based update for the PSD via NMF. The proposed CTF-MNMF-ISS achieves comparable or better separation performance to the original IP-based method and dramatically reduces computational cost, reducing runtime by about 40% and offering increased numerical stability under challenging reverberant conditions. This makes real-time or resource-constrained deployment of CTF-MNMF more practical for multi-microphone arrays in reverberant environments.

Abstract

Among numerous blind source separation (BSS) methods, convolutive transfer function-based multichannel non-negative matrix factorization (CTF-MNMF) has demonstrated strong performance in highly reverberant environments by modeling multi-frame correlations of delayed source signals. However, its practical deployment is hindered by the high computational cost associated with the iterative projection (IP) update rule, which requires matrix inversion for each source. To address this issue, we propose an efficient variant of CTF-MNMF that integrates iterative source steering (ISS), a matrix inversion-free update rule for separation filters. Experimental results show that the proposed method achieves comparable or superior separation performance to the original CTF-MNMF, while significantly reducing the computational complexity.

Figures (3)

• Figure 1: Illustration of the simulation setup.
• Figure 2: Average SDR improvement under different reverberation and microphone setups.
• Figure 3: Average runtime under different microphone setups.