Accelerometer-Derived Jerk Metrics for Detecting Intentional Upper-Limb Tremor During a Reaching Task in Healthy Adults

Objective: To investigate whether accelerometer-derived metrics from an upper-limb reaching task can discriminate between normal and intentionally tremulous movements in healthy adults and to examine the influence of low-pass filtering on tremor sensitivity. Design: Experimental repeated-measures study Methods: Nineteen healthy adults (19–35 years old) performed a seated reaching task from a standardized starting position to targets at 120% arm length at eye and shoulder heights. A wireless triaxial accelerometer was secured to the dorsum of the hand, recording signals at 1,500 Hz and synchronized with a footswitch to define movement onset and offset. Participants completed three trials of normal and intentional tremor movements at both heights. Acceleration data were low-pass filtered using a 4th-order Butterworth filter (cutoffs of 5 and 10 Hz). The movement time, peak-to-peak count of jerk magnitude, mean jerk in each axis, and jerk cost were computed and compared between conditions using paired t-tests. Results: With 5 Hz filtering, intentional tremor produced significantly longer movement times, higher peak-to-peak counts, and elevated jerk cost at both heights (all p < 0.001), whereas mean axis-specific jerk values were largely non-discriminative. Increasing the cutoff to 10 Hz increased the absolute peak counts while preserving between-condition differences (p < 0.001 for all comparisons). Conclusions: Peak-to-peak count and jerk cost derived from accelerometer-based jerk magnitude sensitively distinguished normal from intentional tremor during a reaching task and were robust to modest filtering changes. These metrics show promise as quantitative digital biomarkers for upper limb tremor assessment in future clinical populations.