Saved in:
Bibliographic Details
Main Authors: Koldovský, Zbyněk, Málek, Jiří, Čmejla, Jaroslav, O'Regan, Stephen
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2407.09259
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913428727136256
author Koldovský, Zbyněk
Málek, Jiří
Čmejla, Jaroslav
O'Regan, Stephen
author_facet Koldovský, Zbyněk
Málek, Jiří
Čmejla, Jaroslav
O'Regan, Stephen
contents Non-Gaussianity-based Independent Vector Extraction leads to the famous one-unit FastICA/FastIVA algorithm when the likelihood function is optimized using an approximate Newton-Raphson algorithm under the orthogonality constraint. In this paper, we replace the constraint with the analytic form of the minimum variance distortionless beamformer (MVDR), by which a semi-blind variant of FastICA/FastIVA is obtained. The side information here is provided by a weighted covariance matrix replacing the noise covariance matrix, the estimation of which is a frequent goal of neural beamformers. The algorithm thus provides an intuitive connection between model-based blind extraction and learning-based extraction. The algorithm is tested in simulations and speaker ID-guided speaker extraction, showing fast convergence and promising performance.
format Preprint
id arxiv_https___arxiv_org_abs_2407_09259
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Informed FastICA: Semi-Blind Minimum Variance Distortionless Beamformer
Koldovský, Zbyněk
Málek, Jiří
Čmejla, Jaroslav
O'Regan, Stephen
Signal Processing
Sound
Audio and Speech Processing
Non-Gaussianity-based Independent Vector Extraction leads to the famous one-unit FastICA/FastIVA algorithm when the likelihood function is optimized using an approximate Newton-Raphson algorithm under the orthogonality constraint. In this paper, we replace the constraint with the analytic form of the minimum variance distortionless beamformer (MVDR), by which a semi-blind variant of FastICA/FastIVA is obtained. The side information here is provided by a weighted covariance matrix replacing the noise covariance matrix, the estimation of which is a frequent goal of neural beamformers. The algorithm thus provides an intuitive connection between model-based blind extraction and learning-based extraction. The algorithm is tested in simulations and speaker ID-guided speaker extraction, showing fast convergence and promising performance.
title Informed FastICA: Semi-Blind Minimum Variance Distortionless Beamformer
topic Signal Processing
Sound
Audio and Speech Processing
url https://arxiv.org/abs/2407.09259