This project investigates the dynamic contribution of visual sensory information to human upright balance control. Maintaining a stable posture is a complex, closed-loop process that relies heavily on the central nervous system's ability to continuously integrate visual, vestibular, and somatosensory inputs. Impairments in these systems can lead to severe balance disorders and an increased risk of falls, which represent a significant global health concern, particularly among older adults. Despite extensive research in the field, quantifying the specific role of vision remains challenging due to the ambiguity of optical flow cues and the inherently nonlinear nature of visually evoked postural responses. To address these gaps, this project focuses on developing a comprehensive experimental and analytical framework designed to accurately model how visual perturbations influence both body sway and neuromuscular activation.

To quantify these dynamics, the study utilized a customized virtual reality (VR) platform to deliver precisely controlled visual perturbations to healthy participants. Subjects were exposed to moving visual scenes using specifically designed waveforms, including a trapezoidal velocity (TrapV) signal and a novel multi-level TrapV (ML-TrapV) waveform, which allowed for independent control over perturbation amplitude and velocity. During the trials, participants wore rigid splints extending from the shank to the waist. This crucial mechanical constraint restricted movement at the knees and hips, effectively isolating the body's motion to the ankle joint. This setup enforced an ankle-dominant postural strategy, allowing the researchers to approximate the body's dynamics as a single-link inverted pendulum while simultaneously recording kinematic data, ground reaction forces, ankle torques, and electromyographic (EMG) muscle activity.

The analytical approach leveraged advanced system identification techniques to translate the raw experimental data into interpretable mathematical models. To address the challenge of non-stationarities and slow postural drifts, a novel detrending method based on the inverse Fourier transform was implemented, successfully preserving the critical low-frequency sway components necessary for balance control. For the neuromuscular analysis, Principal Component Analysis (PCA) was applied to the bilateral EMG signals to extract a unified, low-dimensional representation of overall ankle muscle activation. Nonparametric frequency response functions (FRF) and parametric transfer function models including nonlinear Hammerstein models tailored for the ML-TrapV trials were then utilized to characterize the dynamic input-output relationship between the visual stimuli and the resulting biomechanical and neural responses.

The results revealed that visually evoked postural responses exhibit a delayed, second-order, low-pass filter behavior with pronounced amplitude-dependent nonlinearities; specifically, the system's gain decreased as the amplitude of the visual perturbation increased. Furthermore, the neuromuscular analysis uncovered a consistent two-phase muscle activation strategy primarily driven by visual velocity, where ankle muscles initially generate movement consistent with the visual motion and subsequently reverse activation to stabilize posture. The introduction of the ML-TrapV waveform proved highly effective at efficiently capturing these nonlinear dynamics within a single experimental trial. Ultimately, this project establishes a robust quantitative framework that deepens our mechanistic understanding of sensorimotor integration in upright stance. These findings lay a critical methodological foundation for future clinical applications, offering promising avenues for the targeted assessment, diagnosis, and personalized rehabilitation of balance disorders.