English

ISS2: An Extension of Iterative Source Steering Algorithm for Majorization-Minimization-Based Independent Vector Analysis

Signal Processing 2022-06-20 v3 Audio and Speech Processing

Abstract

A majorization-minimization (MM) algorithm for independent vector analysis optimizes a separation matrix W=[w1,,wm]hCm×mW = [w_1, \ldots, w_m]^h \in \mathbb{C}^{m \times m} by minimizing a surrogate function of the form L(W)=i=1mwihViwilogdetW2\mathcal{L}(W) = \sum_{i = 1}^m w_i^h V_i w_i - \log | \det W |^2, where mNm \in \mathbb{N} is the number of sensors and positive definite matrices V1,,VmCm×mV_1,\ldots,V_m \in \mathbb{C}^{m \times m} are constructed in each MM iteration. For m3m \geq 3, no algorithm has been found to obtain a global minimum of L(W)\mathcal{L}(W). Instead, block coordinate descent (BCD) methods with closed-form update formulas have been developed for minimizing L(W)\mathcal{L}(W) and shown to be effective. One such BCD is called iterative projection (IP) that updates one or two rows of WW in each iteration. Another BCD is called iterative source steering (ISS) that updates one column of the mixing matrix A=W1A = W^{-1} in each iteration. Although the time complexity per iteration of ISS is mm times smaller than that of IP, the conventional ISS converges slower than the current fastest IP (called IP2\text{IP}_2) that updates two rows of WW in each iteration. We here extend this ISS to ISS2\text{ISS}_2 that can update two columns of AA in each iteration while maintaining its small time complexity. To this end, we provide a unified way for developing new ISS type methods from which ISS2\text{ISS}_2 as well as the conventional ISS can be immediately obtained in a systematic manner. Numerical experiments to separate reverberant speech mixtures show that our ISS2\text{ISS}_2 converges in fewer MM iterations than the conventional ISS, and is comparable to IP2\text{IP}_2.

Keywords

Cite

@article{arxiv.2202.00875,
  title  = {ISS2: An Extension of Iterative Source Steering Algorithm for Majorization-Minimization-Based Independent Vector Analysis},
  author = {Rintaro Ikeshita and Tomohiro Nakatani},
  journal= {arXiv preprint arXiv:2202.00875},
  year   = {2022}
}

Comments

Accepted for publication in the 30th European Signal Processing Conference (EUSIPCO 2022)

R2 v1 2026-06-24T09:15:08.584Z