SENSORS: Embedded Software for Distributed, Smart Sensor Networks

University Of Utah, Salt Lake City UT

Investigators

Abstract

This research is aimed at solving one of the fundamental problems in tele-sensing: the interpretation of information, and determination of the system state, based on noisy and sparse data from sensor grid sources, interconnected with interruptible (wired or wireless) communication channels. Such situations include medical sensing on ambulatory subjects when sensor signals are frequently dominated by motion artifacts at the interface between the sensor and the body. The approach explored in this research combines sensor-level signal processing with adaptive, optimal sampling of data from multiple sensors to reduce/eliminate motion artifacts and determine reliability/uncertainty measures for the entire measurement and data communication process. Non-medical applications of the proposed software system include sensor control and interpretation of data from other sensor networks, such as shipboard sensors, automotive sensors or airplane sensor grids where distributed, autonomous sensor ensembles are essential in increasing reliability and operational safety. The objective of this research is to develop a method and software for reducing motion artifacts in biomedical sensors. The method will enable devices that (a) fuse data from multiple, networked, independent sensors in order to derive information about the signal of interest and to characterize interfering noise, (b) use adaptive signal conditioning (adaptive filtering, automatic gain control, etc.) and Optimal Algorithms for data acquisition that are based on characteristics of the signal of interest; and (c) provide quantitative measures of data quality and reliability based on error-minimizing optimal sampling criteria. Motion artifacts, i.e., spurious signals caused by tissue movement and intermittent contact of the sensor with the body surface cause significant errors in the signal being measured and are the most serious and most difficult to eliminate class of errors in medical sensors for use by ambulatory subjects. Information and communication technologies successfully address many telehealth monitoring issues; however the success of telemedicine implementation depends on the availability of non-invasive, minimal-contact, miniature, autonomous and low power processors and software embedded in intelligent sensors, immune to errors caused by motion of the subject wearing them. This research, if successful, will have broad impact, enabling the development of a class of reliable, miniaturized personal telehealth monitoring devices and networks that will be used by the elderly persons, convalescent patients, persons working under hazardous conditions and environments, and by military personnel and athletes for physical conditioning in training and protective monitoring in the battlefield.