Visual-Inertial Fusion Framework for Isolating Seated Human-Body Vibration in Dynamic Vehicular Environments

Area of Science:

• Biomechanics
• Human-Computer Interaction
• Automotive Engineering

Background:

• Accurate assessment of discomfort, fatigue, and postural control in vehicle occupants requires understanding whole-body vibration (WBV) transmission and active compensation.
• Existing methods often use single-modality sensing, failing to decompose complex motion components like platform vibration, passive transmission, active compensation, and camera jitter.

Purpose of the Study:

• To propose and validate a visual-inertial fusion framework for isolating seated human body vibration in dynamic vehicular environments.
• To differentiate between passive vibration transmission and active postural compensation.
• To separate true body motion from camera jitter in real-world monitoring systems.

Main Methods:

• A visual-inertial fusion framework integrating Inertial Measurement Units (IMUs) and RGB-D data was developed.
• The system synchronized three IMUs with RGB-D landmarks to separate seat, human body, and camera accelerations.
• Body vibration velocity was derived from differential acceleration, incorporating band-pass filtering and spectral integration. Postural Compensation Index metrics were calculated using 3D landmarks.

Main Results:

• The framework successfully distinguished passive ride phases from actively compensated phases.
• Camera jitter was effectively separated from genuine human body motion.
• Anisotropic postural strategies were revealed, with torso vibration dominated by anteroposterior components (40%) and head motion by lateral sway (>50%).
• Anthropometric evaluations showed shoulder width errors within ±10-20 mm.

Conclusions:

• The proposed framework provides a structured basis for analyzing vibration and posture in in-vehicle monitoring.
• It enables a more accurate assessment of occupant comfort, fatigue, and postural control by decomposing complex motion dynamics.
• The system's ability to isolate specific motion components offers advancements for driver monitoring and vehicle ride analysis.