Identification of intrinsic and reflexive components of human arm dynamics during postural control

In this study a new methodology to quantify reflexive feedback gains from the mechanical behavior of the human arm during posture maintenance is proposed. Disturbance experiments were carried out on human subjects using continuous random force inputs. The task instruction was 'minimize displacements', prescribing a maximum performance task. For the separation of intrinsic and reflexive components, system identification in the frequency domain is applied. From the time records of position and force, frequency response functions (FRFs) are estimated. Given a model structure and an appropriate estimate of the intrinsic component, an estimate of the reflex gains for length and velocity are obtained from the FRFs. The feedback gains vary considerably with the frequency content of the disturbance signal. The results show that reflexive dynamics are substantial for narrow-band and especially low-frequency input signals. It is likely that high reflex gains are most effective for low-frequency inputs (< 3 Hz) that do not excite the closed-loop system's eigenfrequency. Also significant negative reflex gains are estimated for near-sinusoidal inputs (> 1.5 Hz). It is concluded that this new methodology can offer interesting insights into the ability of the central nervous system to modulate reflexive feedback gains.