Paper co-written by Joseph L. Betthauser, PhD ’20, featured on IEEE Transactions Biomedical Engineering homepage
A paper that was co-written by Joseph L. Betthauser, who earned his PhD from the Johns Hopkins University ECE Department earlier this year, is being featured on the homepage of the IEEE Transactions on Biomedical Engineering homepage for the month of June. The paper, which is titled “Stable Responsive EMG Sequence Prediction and Adaptive Reinforcement with Temporal Convolutional Networks,” also appears in the current June issue of the organization’s printed journal.
Betthauser co-wrote the paper with Nitish V. Thakor, a professor in the JHU Department of Biomedical Engineering who also holds a secondary appointment with the ECE Department. John T. Krall, Shain G. Bannowsky, Gyorgy Levay, Rahul R. Kaliki and Matthew S. Fifer are also listed as co-authors.
Click here to read the full paper. Below is the paper’s full abstract:
Prediction of movement intentions from electromyographic (EMG) signals is typically performed with a pattern recognition approach, wherein a short dataframe, or window, of raw EMG is compressed into an instantaneous feature-encoding that is meaningful for classification. However, EMG signals are time-varying, and the frame-wise approach may not sufficiently incorporate temporal context into predictions leading to erratic and unstable prediction behavior. We demonstrate that sequential prediction models and, specifically, temporal convolutional networks leverage useful temporal information from EMG to achieve superior predictive performance. We compare this approach to other sequential and frame-wise models predicting 3 simultaneous hand and wrist degrees-of-freedom from 2 amputee and 13 non-amputee human subjects in a minimally constrained experiment. We also compare these models on the publicly available Ninapro and CapgMyo amputee and non-amputee datasets. Temporal convolutional networks yield predictions that are more accurate and stable (p<0.001) than frame-wise models, especially during inter-class transitions, with an average response delay of 4.6 ms (p<0.001) and simpler feature-encoding. Their performance can be further improved with online reinforcement training. Sequential models that incorporate temporal information from EMG achieve superior movement prediction performance and these models allow for novel types of interactive training. We want readers of this work to come away, not just with a specific prediction model, but with a concept: movement decoding from EMG is better represented as a sequential problem requiring a new set of considerations and experiment designs. The models themselves will evolve and today’s leading-edge methods will become obsolete, so encouraging researchers to think about the problem in a new way is our overarching goal. Addressing EMG decoding as a sequential modeling problem will lead to enhancements in the reliability, responsiveness, and movement complexity available from prosthesis control systems, pushing us further toward achieving fully restorative natural upper-limb function for amputees.