Multimodal Latent Variable Analysis

TL;DR

This paper advances multimodal sensor data analysis by developing an algorithm to extract both common and sensor-specific sources of variability, demonstrated through theoretical analysis and applications including fetal ECG extraction.

Contribution

It introduces a novel algorithm that extends existing methods to identify both shared and unique sources of variability in multimodal sensor data.

Findings

The algorithm successfully extracts common and sensor-specific sources.

Theoretical analysis confirms the method's validity.

Effective application demonstrated on synthetic, toy, and fetal ECG data.

Abstract

Consider a set of multiple, multimodal sensors capturing a complex system or a physical phenomenon of interest. Our primary goal is to distinguish the underlying sources of variability manifested in the measured data. The first step in our analysis is to find the common source of variability present in all sensor measurements. We base our work on a recent paper, which tackles this problem with alternating diffusion (AD). In this work, we suggest to further the analysis by extracting the sensor-specific variables in addition to the common source. We propose an algorithm, which we analyze theoretically, and then demonstrate on three different applications: a synthetic example, a toy problem, and the task of fetal ECG extraction.