"Kinematic features Labeled with Sympathetic Arousal "

"Time-varying acceleration (TvA), a common motion pattern in modern transportation systems, evokes compensatory responses that couple postural adjustments with sympathetic arousal. Applying an interpretable machine learning pipeline to a multi-cohort dataset, we select an consensus optimal feature subset of kinematic features that predict event-related skin conductance responses (ER-SCRs) with an area under the receiver operating characteristic curve (AUROC) of 0.79.  This study uncovers a hierarchical postural control strategy under threat, providing a theoretical foundation for monitoring autonomic states via body kinematics in dynamic settings. Participants' ER-SCRs and kinematic data (resultant acceleration and angular displacement from right temporal region, L5 lumbar vertebra, and right fibular head) were simultaneously recorded. The EDA were time-locked to the onset of each TvA event, and kinematic data were segmented within the corresponding acceleration intervals. This yielded 181 observations in the development cohort and 94 observations in the external cohort. "