Publication: Enhanced single and multi-channel blind source separation based on fractional time-frequency analysis with synchrosqueezing and sparsity approaches
Loading...
Date
2024-11-01
Authors
Li Yangyang
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Blind Source Separation (BSS) is vital in signal processing for isolating and
extracting source signals from mixed observation without prior information. This
study addresses challenges in mixed matrix estimation in single and multi-channel
BSS systems. A synchrosqueezing instantaneous frequency method to concentrate
the energy in mixed signal spectra is suggested to improve the performance of single
channel BSS. For multi-channel BSS, an overcomplete chirp-based dictionary to
improve the time-frequency resolution and signal clustering is developed to enhance
the mixed matrix estimation accuracy. For the single channel, the accuracy improved
by 77.58% compared to existing algorithms whereas the reduction of 5 seconds in
processing time at sampling rates above 10,000. For the multi-channel, the accuracy
of the estimated mixed matrix and recovered source signal improved by 27.22% and
29.85%, with a reduction of the processing time of more than 1 second and 0.3
seconds, respectively. These advancements prove the effectiveness of the proposed
algorithms in advancing the BSS technology.