Identification of quadriceps-shank dynamics using randomized interpulse interval stimulation

Model structures for artificially stimulated paralyzed muscle-limb system dynamics were developed and experimentally evaluated in paraplegic patients. The examined system consisted of the quadriceps, electrically stimulated using surface electrodes, and a freely swinging shank. The interpulse interval of the stimulation sequence was randomized to obtain persistent system excitation. The outputs of the system were the angular position, velocity, and acceleration, measured by externally mounted sensors. The authors especially report on model identification of the active quadriceps dynamics and the angle prediction performance of the total quadriceps-shank model. Second-order modeling of the twitch dynamics with delay did not significantly improve the prediction results in comparison to a zero-order model with delay (α=0.05). Nonlinear torque-angle and torque-angular velocity relations in combination with a zero-order model (with delay) only slightly improved the prediction results for large prediction intervals (α=0.05). The delay between stimulation input and resulting knee joint acceleration appeared to be joint angle dependent and was estimated to be largest in the knee angle range near knee extension, i.e., when quadriceps muscle is shortest