Leveraging Textile Capacitive Sensor Arrays for Body Gesture Classification Applications

Abstract

The objective of this theses is to explore textile capacitive sensor arrays (CSAs) as a means for capturing human body gestures. Given their low profile, these fabric sensors can be integrated into the user's clothing or environment to provide a unique low-cost solution for capturing information-rich kinetic signals. However, our attempts to infer motion and user intent indirectly from an array of irregular deformable sensors revealed computational challenges to implementation of the low-cost smart sensor. This theses includes a discussion of challenges for use of fabric capacitive sensors, as well as implementation of low-cost embedded systems utilizing distance-based and statistical classifiers. The usage of CSAs is motivated through three unique applications; namely, an environmental control system for individuals with limited upper-body mobility, a distracted driving detector, and a restless leg syndrome sleep monitoring system. The three applications are compared to their commercial equivalents in terms of power, cost, and usability to highlight the advantages of CSAs. The fabrication of the CSAs, and the design of the custom data acquisition module are described in detail. Algorithms presented for classifying gestures include K-Nearest Neighbor and Hidden Markov Models. Finally, for sensor validation and assisting design of future CSA applications, initial work on a low-cost tool based on finite-element simulation is presented.