Classification of Abnormal Gaits with Machine Learning Algorithms using Sensor-Inherited Insoles

The digital health industry's interest in gait analysis has driven research into sensor-enabled insoles for practical, everyday gait monitoring. Traditional methods, such as 3D motion capture systems, are costly and time-consuming. To address this, we propose an efficient method to evaluate gait performance. Our study involved 54 subjects performing various gait patterns, with data collected from six insole pressure sensors. Rigorous data processing resulted in 36 significant parameters. These parameters were used to build a classification model using Support Vector Machine, Random Forest, Extreme Gradient Boosting, and k-Nearest Neighbors. Promising results were observed, with Extreme Gradient Boosting showing high classification performance. The model achieved accuracies of 0.76 at the sample level and 0.85 at the subject level. This study contributes to digital health by providing an alternative for gait analysis, which will improve patient care in orthopedics and rehabilitation.