Gait Pattern Recognition based on Multi-sensors Information Fusion through PSO-SVM Model.

This study describes a gait analysis system that classifies gait phases using three pressure sensors and one inertial measurement unit (IMU) sensor. The pressure sensors are placed on the insole to measure ground reaction forces (GRF), while the IMU sensor is placed on the tongue to monitor foot angle changes. Gait phases, such as initial contact (IC), mid-stance (MS), terminal stance (TS), and swing (SW), are classified using a support vector machine (SVM). To tackle the problem of low accuracy caused by incorrect SVM model parameters, the study utilized particle swarm optimization (PSO) to optimize the SVM parameters. Three classifiers were constructed using the collected gait dataset while walking, such as a KNN classifier, a SVM classifier, and a PSO-SVM classifier. The experimental results demonstrate that the PSO-SVM algorithm outperforms the others in gait phase classification, achieving an accuracy of 95%. The PSO-SVM approach outperforms both the SVM. This finding illustrates the PSO-SVM approach's superiority in gait phase categorization and indicates its potential utility for classifying gait phases. [ABSTRACT FROM AUTHOR]