Recognition of lower limb movements by artificial neural network for restoring gait of hemiplegic patients by functional electrical stimulation

This study focused on man-machine interface of FES system for restoring gait of hemiplegic patients. A method of recognition of lower limb movements using an artificial neural network (ANN) was examined in monitoring restored motions and in giving control command with normal subjects and a hemiplegic patient. Acceleration signals were measured with a three-axis accelerometer attached to the heel of the normal side (right side) during walking for using in the recognition. Subjects performed some specific movements by their normal lower limbs supposing control command input in the walking measurements. The ANN recognized three different walking patterns, which were level floor walking, going up and down stairs, based on the acceleration waveforms with about 80% of recognition rate for normal subjects and above 70% for the patient. A similar structure of the ANN discriminated four specific movements by the lower extremity with more than 90% of recognition rate after the third performance of the movement simulated by using recognition and mis-recognition rates for experimentally measured data. The method was found to be useful in monitoring FES movements for safety and in giving control commands to the FES system without using upper limbs.