EMG-based pattern recognition with kinematics information for hand gesture recognition

Predicting movement intention based on pattern recognition of biosignals are becoming increasingly important in many clinical and research applications. The main idea of these technologies is to use a human-in-the-loop system for controlling several hardware devices, in which the identification of the user intention is the most challenging aspect. Electromyography (EMG) provides biological signals that can be used in various applications including identifying neuromuscular diseases, control for prosthetic and assistive devices, human-machine interaction, etc. Decoding upper-limb gestures from EMG could help to develop human-computer interfaces that increase the quality of life of the disabled or aged people. Implementation of the EMG-based pattern recognition is not easy to be accomplished due to some difficulties, among them that EMG are time-varying and highly nonlinear. Kinematics information (i.e. position, velocity, acceleration) could provide valuable information in an EMG-based pattern recognition process to improve classification. This work describes a comparative study of two EMG-based pattern recognition algorithms aimed at decoding upper-limb gestures. Results show that algorithm using kinematics information and EMG data was better in classification of five movements at the upper-limb level that algorithm using only EMG data.