A Novel Template-Matching Method for Extracting Gait Cycles from Underfoot Pressure Data

The isolation of gait cycles from underfoot pressure data is a critical step in human movement analysis, yet existing methods require either expensive equipment or substantial manual effort. We present a generalizable, lowresource algorithm for parsing gait cycles from wearable underfoot pressure sensor data across multiple walking terrains. Compared to a ground-truth dataset manually marked and verified by an expert, our algorithm achieved similar accuracy with significantly reduced processing time. For a dataset of 577 steps, our algorithm parsed gait cycles in 41 seconds with only one false negative, whereas three naïve manual parsers required an average of 29 minutes, with total errors ranging from 6 to 33 gait cycles. The algorithm also outperformed threshold-based parsing, which produced 49-362 false positives, depending on the threshold value. By improving computational efficiency with accuracy, this method enables scalable gait cycle detection in real-world applications, expanding the feasibility of automated gait analysis beyond controlled laboratory environments.