Classification of Trunk Motion for a Backbone Exoskeleton Using Inertial Data and Surface Electromyography

This paper evaluates a multi-sensor, intent recognition approach towards controlling a powered backbone exoskeleton. Inertial sensors emulated from motion capture recordings, and surface electromyogram (sEMG) readings from the back and abdomen, were investigated for efficacy in recognizing flexion and extension motion of the trunk. A linear discriminant analysis (LDA) based method of pattern classification was used to integrate the two different sensor readings together and identify trunk motion. The classification approach involved multiple classifiers that were generated from different features extracted from the data, which were then combined using a majority voting based classifier fusion scheme. An experimental procedure to record the signals and validate the pattern classification method has been described and results are discussed.