Adapted Compressed Sensing with Incremental Encoder and Deep Performance Predictor for Low-Power Sensor Node Design

A. Marchioni, F. Martinini, L. Manovi, S. Cortesi, R. Rovatti, Gianluca Setti, M. Mangia

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Wearable solutions, especially within the realm of health, are in high demand. Meeting this demand is possible thanks to energy-efficient sensors that acquire signals and transmit them as digital messages. Since transmission often heavily contributes to the overall energy consumption, compression can be beneficial. Inspired by Compressed Sensing, we propose an iterative compression scheme able to reduce the energy needed to transmit a signal. To meet a constraint in the quality of service, our novel scheme adapts the number of transmitted digital words for each signal instance. Adaptation is based on the output of a performance predictor embedded in the receiving decoder, that at every iteration estimates the quality of service. We test our novel compression paradigm over ECG signals and BLE communication protocol. Our method significantly reduces computation and transmission energy consumption compared to other already established techniques based on Compressed Sensing.
Original languageEnglish (US)
Title of host publication2023 IEEE International Instrumentation and Measurement Technology Conference (I2MTC)
PublisherIEEE
DOIs
StatePublished - Jul 13 2023

Fingerprint

Dive into the research topics of 'Adapted Compressed Sensing with Incremental Encoder and Deep Performance Predictor for Low-Power Sensor Node Design'. Together they form a unique fingerprint.

Cite this