Optimal Channel-set and Feature-set Assessment for Foot Movement Based EMG Pattern Recognition

Electromyography (EMG) -based control is the most convenient and robust way to classify body movements for controlling prosthetic as well as orthotic devices. Its translation from lab-based approach to assistive devices demands a problem-centric and cost-effective solution. This paper demonstrates its utility for the classification of four foot movements, viz Plantar flexion, Dorsi flexion, Eversion and Inversion. For the experimental study, four superficial muscles (viz. Tibialis Anterior, Extensor Hallucis Longus, Gastrocnemius Medial and Fibularis Longus) were identified as electrode positioning locations for the EMG data acquisition. This work is aimed to minimize the number of electrode locations without significantly affecting the classification performance. Channel-set CH2,4 corresponding to the combination of Hallucis Longus and Fibularis Longus muscles is found to be the most optimal. The maximum classification accuracy obtained for the given set with the selected feature-set has been (91.85 ± 3.57)%. The classification performance has been assessed on the basis of parameters such as the type of classifier, window length, data sampling and also the body mass index of the participants. The developed technique can be applied for control of ankle exoskeletons for healthy as well as person with certain disabilities.