Motion Gesture Delimiters for Smartwatch Interaction

Abstract

Smartwatches are increasingly popular in our daily lives. Motion gestures are a common way of interacting with smartwatches, e.g., users can make a movement in the air with their arm wearing the watch to trigger a specific command of the smartwatch. Motion gesture interaction can compensate for the small screen size of the smartwatch to some extent and enrich smartwatch-based interactions. An important aspect of motion gesture interaction lies in how to determine the start and end of a motion gesture. This paper is aimed at selecting gestures as suitable delimiters for motion gesture interaction with the smartwatch. We designed six gestures (“shaking wrist left and right,” “shaking wrist up and down,” “holding fist and opening,” “turning wrist clockwise,” “turning wrist anticlockwise,” and “shaking wrist up”) and conducted two experiments to compare the performance of these six gestures. Firstly, we used dynamic time warping (DTW) and feature extraction with KNN ( K -nearest neighbors) to recognize these six gestures. The average recognition rate of the latter algorithm for the six gestures was higher than that of the former. And with the latter algorithm, the recognition rate for the first three of the six gestures was greater than 98%. According to experiment one, gesture 1 (shaking wrist left and right), gesture 2 (shaking wrist up and down), and gesture 3 (holding fist and opening) were selected as the candidate delimiters. In addition, we conducted a questionnaire data analysis and obtained the same conclusion. Then, we conducted the second experiment to investigate the performance of these three candidate gestures in daily scenes to obtain their misoperation rates. The misoperation rates of two candidate gestures (“shaking wrist left and right” and “shaking wrist up and down”) were approximately 0, which were significantly lower than that of the third candidate gesture. Based on the above experimental results, gestures “shaking wrist left and right” and “shaking wrist up and down” are suitable as motion gesture delimiters for smartwatch interaction.

Exploring gesture generation for smartwatches: is user elicitation enough?

“…Expert generation involves experts (designers, engineers, ergonomists, etc.) to generate and evaluate gestures for interfaces as in Cannanure, et al (2016), andZhao, et al (2022). Where as, during Gesture Elicitation participants are typically shown the effect of a certain action, and are then prompted to perform the gesture that would cause this action.…”