Decoding Simultaneous Multi-DOF Wrist Movements From Raw EMG Signals Using a Convolutional Neural Network

Pattern recognition (PR) methods are commonly utilized in the extraction of motion intentions from myoelectric signals, which is realized by relating several electromyogram (EMG) patterns to specific types of motion. Researchers have reported that the hand-engineering features widely used in PR-based methods can be significantly affected by external confounding factors that diminish their accuracy and robustness in clinical settings. Moreover,...