Development of EMG-based mode and intent recognition algorithms for a computer-controlled above-knee prosthesis

Abstract

Conventional above-knee prostheses are unable to replace normal leg function due to the lack of adaptive control. Computer-controlled prostheses offer the possibility to implement such adaptive control. For a particular locomotion mode, control algorithms can generate appropriate damping profiles as a function of selected sensory inputs and stored information on normal walking. However, it is essential in such systems that the locomotor modes can be determined accurately from the inputs of the control system. Recognition of the intent to change from one mode to another is also a necessity because the control system has to account for such transitions. The possibility of using EMG signals from hip muscles and muscle residuals of the stump for this purpose is investigated. The modes tested are level walking, ramp ascent and ramp descent with slopes of 6 degrees and 9 degrees. EMG activity curves for three muscles: gluteus maximus, gluteus medius and tensor fasciae latae, are presented for the different modes. The results are obtained from two reference groups of 12 normal individuals and from a prosthetic patient. The results show the possibility of discriminating between modes. A discussion is made of the implementation of the results obtained by a finite state approach and the difficulties relating to the use of EMG signals for control purposes.