A Hybrid Measurement System for Hand Signs Recognition based on EMG-FMG Measurements

摘要

Human-machine interaction, precision robotic control, and guide advice for the rehabilitation process are scorching topics. For these topics, detecting and accurately recognizing muscle signals for specific gestures is a proven process, and component. In this direction, this paper designs a hybrid measurement system based on EMG and FMG with an 8-channel FMG (Force Myography) band based on nanocomposite pressure sensors and 4-channel EMG (Electromyography) based on the Myoware chip. To construct the database, we selected three healthy right-handed individuals as subjects. Each individual is asked to perform 13 basic hand gestures, and each gesture is repeated 20 times. A linear support vector machine learns the data for each test subject for classification. The system shows high robustness to one subject measurements, where the minimum archived accuracy between the three volunteers is 99 %. In addition, the system is tested against person independence, where all data of all subjects are mixed. The system achieves 98.1 % accuracy.