Gait analysis for fall prediction using hierarchical textile-based capacitive sensor arrays

Falls are a major cause of injuries in adults above the age of sixty-five. The economic aftermath of falls and their consequent hospitalization can be huge, totaling more than 30 billion dollars in 2010 alone. A plausible way of mitigating this problem is accurate prediction of future falls and taking proactive remedial action. Spatio-temporal variation in gait is a reliable indicator of a future fall, however, existing systems focus on gait analysis in clinical settings and are not tuned towards continuous gait analysis. In this paper, we present the design of a novel textile capacitive sensor array-based system built into clothing that can reliably capture spatio-temporal gait attributes in a home setting. A key novel research contribution of our work is a context-aware hierarchical signal processing architecture that breaks down the signal processing algorithm into a hierarchy of processing elements. The lower power processing components perform generic feature extraction using observations derived from the capacitor plates, while the higher-level processors aggregate features to infer gait attributes such as stride speed and inter-leg spacing. The system activates the higher power processing elements only when it detects walking. We have prototyped our system using textile capacitive plates built into an ace-bandage and a custom FPGA-based system and show that our system can accurately detect gait attributes that have high correlation with falls, while consuming minimal energy as estimated for a multi-clock-domain 180-nm IC.