Embedded Smart Insole for Accurate Gait Event Detection

This paper presents the design, implementation, and validation of a low-power smart insole system for gait event detection, with a focus on stance, swing, and step durations. Plantar pressure signals, captured via force-sensing resistors embedded in the heel and toe regions, were processed using a centered rolling mean and analyzed through a threshold-based detection algorithm. Ground-truth annotations were obtained from a time synchronized Qualisys motion capture system. Across 14 participants, covering more than 1,200 gait cycles at five controlled walking speeds, the insole-based algorithm demonstrated high accuracy in detecting heel-strike (HS) and toe-off (TO) events. Bland-Altman analysis revealed that stance and swing durations exhibited small speed-dependent biases (mean percent error < 10%) that diminished at faster walking speeds, while step durations showed near-zero bias with tight dispersion (ρ ±1%). Correlation analysis confirmed strong agreement with motion capture across all metrics ( r > 0.95). These results indicate that despite simplified sensing hardware and cost-efficient design, the system achieves robust gait characterization sufficient for ambulatory monitoring and longitudinal assessment in real-world environments.