A Modified ICA Approach for Signal Separation in CMB Maps

TL;DR

This paper introduces a modified ICA method tailored for separating signals in CMB maps, effectively utilizing prior information and improving stability and quality over traditional ICA techniques.

Contribution

It presents a novel modification of the ICA algorithm that incorporates prior knowledge, enhancing signal separation in CMB data analysis.

Findings

Modified ICA yields more stable separation results

Improved quality of signal extraction in CMB maps

Better exploitation of prior information in signal processing

Abstract

AIMS: One of the most challenging and important problem of digital signal processing in Cosmology is the separation of foreground contamination from cosmic microwave background (CMB). This problem becomes even more difficult in situations, as the CMB polarization observations, where the amount of available "a priori" information is limited. In this case, it is necessary to resort to the "blind separation" methods. One important member of this class is represented by the "Independent Components Analysis" (ICA). In its original formulation, this method has various interesting characteristics, but also some limits. One of the most serious is the difficulty to take into account any information available in advance. In particular, ICA is not able to exploit the fact that emission of CMB is the same at all the frequencies of observations. Here, we show how to deal with this question. The connection of the proposed methodology with the "Internal Linear Composition" (ILC) technique is also illustrated. METHODS: A modification of the classic ICA approach is presented and its characteristics are analyzed both analytically and by means of numerical experiments. RESULTS: The modified version of ICA appears to provide more stable results and of better quality.