System identification for human motion control

System identification is an important tool to investigate human motion control. The goal of this study was to identify and address issues in system identification for human motor control and to investigate if linear time invariant system identification techniques are applicable to recorded mechanical and physiological data. During a posture maintenance experiment subjects had to minimize the deviations of the wrist, while a manipulator applied disturbances through a handle. We estimated frequency response functions and categorized and quantified the errors influencing the estimates. The relation between the position and force of the wrist was found to be highly linear and variations over time were the dominant source of error. The relation between the position changes at the wrist and the recorded signals from the brain was highly nonlinear. System identification techniques based on periodic multisine perturbation signals are a promising approach to investigate human motion control, even in the presence of inherent noise and nonlinearities.