Extreme learning machine based sEMG for drop-foot after stroke detection

Drop Foot (FD) is the inability to raising the foot due to the weakness of paralysis dorsiflexion muscle, this disability is caused by stroke. Recognize FD of the patient and provide a treatment as they needed is an important requirement for rehabilitation. One of the recognize techniques for FD event is based on using the surface electromyography (sEMG) signal. Utilizing sEMG signal can help to provide patient the specific rehabilitation treatment in specific time since it can detect the FD event before it happen. This paper has investigated the ability of Extreme Learning Machine (ELM) method to classify the sickness and healthy muscles on the leg, based on sEMG, that yield the FD. The performance is compared with Support Vector Machine (SVM) and Neural Network (NN). Classification accuracy with ELM is much better than SVM and NN giving the results with up to 97% classification accuracy using two channels on each side of the leg.