Latent Sensor Fusion: Multimedia Learning of Physiological Signals for Resource-Constrained Devices

TL;DR

This paper introduces a sensor-latent fusion method using autoencoders to efficiently analyze multimodal physiological signals on resource-limited devices, achieving faster and scalable performance without losing accuracy.

Contribution

It presents a novel modality-agnostic encoder leveraging meta-embeddings and compressed sensing for biosignal analysis on constrained hardware.

Findings

Faster and lighter than modality-specific methods

Maintains high representational accuracy

Scalable to multiple physiological signals

Abstract

Latent spaces offer an efficient and effective means of summarizing data while implicitly preserving meta-information through relational encoding. We leverage these meta-embeddings to develop a modality-agnostic, unified encoder. Our method employs sensor-latent fusion to analyze and correlate multimodal physiological signals. Using a compressed sensing approach with autoencoder-based latent space fusion, we address the computational challenges of biosignal analysis on resource-constrained devices. Experimental results show that our unified encoder is significantly faster, lighter, and more scalable than modality-specific alternatives, without compromising representational accuracy.