ED-TCN

Stable Responsive EMG Sequence Prediction and Adaptive Reinforcement with Temporal Convolutional Networks

Author(s)3: Joseph L. Betthauser, John T. Krall, Shain G. Bannowsky, Gyorgy Levay, Rahul R. Kaliki, Matthew S. Fifer, Nitish V. Thakor
Stable Responsive EMG Sequence Prediction and Adaptive Reinforcement with Temporal Convolutional Networks 170 177 IEEE Transactions on Biomedical Engineering (TBME)
Movement prediction from EMG can be performed by compressing a short window of EMG into a feature-encoding that is meaningful for classification— an approach that can cause erratic prediction behavior. Temporal convolutional networks (TCN) leverage temporal information from EMG to achieve superior predictions for 3 simultaneous degrees-of-freedom that are more accurate and stable, have a very low response delay, and allow for novel types of interactive training. Addressing EMG decoding as a sequential prediction problem requires a new set of considerations that will lead to enhancements in the reliability, responsiveness, and movement complexity available from prosthesis control systems. read more