TechRxiv
An_Linf_norm_Based_Criterion_for_BCA_Brotto_Nose_Romano.pdf (1.96 MB)
Download file

An ℓ∞ norm-Based Criterion for Bounded Component Analysis

Download (1.96 MB)
preprint
posted on 02.05.2022, 18:36 authored by Renan BrottoRenan Brotto, Kenji Nose FilhoKenji Nose Filho, João Romano
In the last decade, there has been a great interest in finding new priors for Blind Source Separation (BSS). Some of the most interesting of them are sparsity, non-negativity and boundedness of the sources, since they consist in weaker assumptions when compared to the classical statistical independence hypothesis. In this paper, we propose a new criterion for Bounded Component Analysis (BCA), using the $\ell_\infty$ norm, with an associated algorithm based on Givens Rotations. We analyze our proposal and evaluate it with numerical simulations for three kinds of bounded signals. From both theoretical and experimental results we show that the $\ell_\infty$ norm is a suitable contrast function for BCA, presenting very good results when compared to the state-of-the-art algorithm.

History

Email Address of Submitting Author

renanbrotto@gmail.com

Submitting Author's Institution

UNICAMP

Submitting Author's Country

Brazil