Adaptive real-time tool for human gait event detection using a wearable gyroscope

Abstract

The development of robust algorithms for human gait analysis are essential to evaluate the gait performance, and in many cases, crucial for diagnosing gait pathologies. This work proposes a new adaptive tool for human gait event detection in real-time, based on the angular velocity recorded from one gyroscope placed on the instep of the foot and in a finite state machine with adaptive decision rules. The signal was segmented to detect 6 events: Heel Strike (HS), Foot Flat (FF), Middle Mid-Stance (MMST), Heel-Off (HO), Toe-Off (TO), and Middle Mid-Swing (MMSW). The tool was validated with healthy subjects in ground-level walking using a treadmill, for different speeds (1.5 to 4.5 km/h) and slopes (0 to 10%). The results show that the tool is highly accurate and versatile for the detection of all events, as indicated by the values of accuracy, average delays and advances (HS: 99.96%,-7.95 ms, and 9.85 ms; FF: 99.48%,-4.95 ms, and 9.35 ms; MMST: 98.26%, 36.54 ms, and 16.38 ms; HO: 98.87%,-22.71 ms, and 18.62 ms; TO: 95.95%,-6.80 ms, 14.38 ms; MMSW: 96.06%,-3.45 ms; 0.15 ms, respectively). These findings suggest that the proposed tool is suitable for the real-time gait analysis in real-life activities.