A EMG SIGNAL BASED GAIT PHASE RECOGNITION METHOD USING FEATURE SELECTION ADAPTIVE TO MUSCLES FOR STAIRS ASCENDING AND STAIRS DESCENDING

摘要

The present invention relates to a walking step recognition method based on electromyogram signals in an upward or downward stair walking operation. The walking step recognition method comprises: (a) a step of receiving training signals of electromyogram signals in a plurality of channels measured during an uphill or downhill walking operation, and generating a classifier to classify walking steps in a walking operation from the electromyogram signals for each of the channels; (b) a step of receiving testing signals of the electromyogram signals to test the classifiers and obtain accuracy thereof, and selecting classifiers in accordance with the accuracy; (c) and a walking step of recognizing the walking steps of the walking operation with the selected classifiers. The classifiers extract two or more feature values from the electromyogram signals and classify the walking steps in accordance with the extracted feature values. The walking step recognition method is able to combine two or more suitable properties of muscles used in the walking operation to generate the classifiers and classify and recognize the walking steps, thereby improving a rate of recognition of the walking steps when compared with the recognition rate obtained when only a classifier for a single feature is used.