Abstract
Functional Electrical Stimulation (FES) is an effective and developing method used to restore functions for paraplegic patients. In this research, we focus on the switching problem of FES, which is one of the obstacles that prevents FES from further practical use. Namely, in most of the current FES systems, patients have to make a superfluous action by themselves, or rely on someone else to turn on/off the stimulation instead. To release patients from such a switching action, we have been developing an adaptive switching system for FES control for the lower limb activities of hemiplegic patients, based on the consideration that lower limb activities need the synchronization of limbs on both sides. We used electromyogram (EMG) signals detected from the normal side to recognize the activities that the patients intend to do, and utilized the recognition results as the switching signals. However, motion patterns to be represented and analyzed by EMG are distinctive of individual variations and characteristic alternation, which inevitably results in classification errors in EMG analyzing. Moreover, EMG analyzing for FES switching should be able to cope with the contamination of FES pulse. We first compared three methods to decide the suitable feature extraction for EMG analyzation for FES systems. Then, in order to enable the analyzing system to recognize the correct timings in the dynamical processes of activities, a practical training-set construction method that utilizes additional reference data was proposed. Accordingly, the problem-oriented feature extraction and the training-set construction were incorporated with an Artificial Neural Network (ANN)-based online learning system to form an adaptive switching system for FES. The proposed switching system was applied to an FES system that supports the standing and walking of a hemiplegic subject to verify the effectiveness.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have