A Wireless Smart Shoe for Gait Analysis Using Heuristic Algorithms with Automated Thresholding

This paper proposes a novel decision system for segregation of five normal gait phases (stance, heel-off, swing 1, swing 2 and heel-strike) by using a real-time wireless smart shoe. The classification method employed four force sensitive resistors to measure the force underneath a foot, together with an inertial measurement unit that is attached at the back of the shoe to determine the magnitude of acceleration and the inclination angle of the foot with respect to the ground. Data acquisition was through the XBee wireless network protocol to allow serial processing by a computer. The state transition theorem and threshold-based classification were used to distinguish the gait phases according to received data, where the thresholds were optimized by heuristic algorithms, such as the genetic algorithm and particle swarm optimization. Ground truthing was determined by marker-tracking using image processing. Video recording with real-time data and embedded interfacing was used to verify the output of the proposed state transition algorithm under indoor testing on a stationary treadmill. Experimentation and verification were conducted on a subject with a normal gait cycle. The smart shoe was able to ascertain the gait phase with 96.07% accuracy after optimized threshold values had been determined.