Modelling and correcting the effects of tremor on touchscreen interaction

Essential tremor and Parkinson’s tremor are common movement disorders, affecting approximately 7% of adults over 65, increasing with age. Although numerous studies have used keyboard timing, stand-alone sensors, or smartphone sensors to identify and measure tremor severity, relatively little research has isolated and examined how tremor impacts overall mobile interaction and usability models. This study investigated controlled fixed levels of induced tremor in a standardised ISO pointing task to quantify the effect on usability models of different levels of tremor. Tremor was induced using a wired glove delivering 4 Hz stimulation at 0, 15, 19, and 23 mA (approximating the range from slight to marked tremor). The results show that the Fitts’ throughput was 5.43 bps with no induced tremor, with reductions of 9.4%, 20.3%, and 28.9% at the active levels. Fitts’ law models showed high correlations that improved further when applying the Welford correction for tremor, supporting use of the adjustment when tremor is present. The study also indicates that machine-learned correction significantly improves touch accuracy when using smartphone sensors.