RESCU System for Robust Upper Limb Prosthesis Control

Retrospectively Supervised Classification Updating (RESCU) is founded on two innovations that promise significant improvement in performance and outcome. The first is a highly robust machine intelligence algorithm, an Extreme Learning Machine with Adaptive Sparse Representation (EASRC), and the second is a novel adaptive learning algorithm and communication interface we call Nessa. We contend that these two technologies allow the prosthetic device to adapt to its user from the initial fitting through continuing, long-term use in the activities of daily living, shifting the paradigm of training from the current prospective data gathering methods to a more dynamic retrospective application.

Pattern recognition prostheses associate the patterns of activity of multiple EMG sites to the action of a prosthesis. Such strategies have historically required prospective calibration of the EMG activation patterns.