Research on Recognition of Human Motion State Based on Force and Motion Sensor Fusion

Abstract

The exoskeletons, as the wearable mechanical devices with powerful features of physical enhancement on human performance, has gained more and more attention from scientific researchers in the world. It has become a new research hot spot, and has also begun to be gradually applied to the military industry. An exoskeleton can be a coupling system based on human-computer interaction. In order to achieve human-machine motion coordination and adaptation enhancement, the system must have cognitive intelligence. When the exoskeleton is in use, the system needs to operate fast and accurate enough to predict the human intention and movement (such as: walking, standing, sitting or going up and going down stairs) and determine the gait cycle stage. In addition, it must be robust against small errors within the period simultaneously. Therefore, the recognition of human motion status is one of the core technologies in system design for the exoskeleton. In our research, signals from foot force sensors and Inertial Measurement Unit (IMU) sensors were collected, and Support Vector Machine (SVM) was used to identify the human body motion state through adaptive time window segmentation. After that, experimental tests were conducted to verify the algorithm and recognition accuracy.