Adaptive template adjustment for personalized gesture recognition based on a finger-worn device

Abstract

Wearable device based gesture recognition has become a hot topic in healthcare research fields. Effective gesture recognition is helpful to not only provide services for health monitoring, but also develop various applications like appliance control and emergency call. However, significantly individual difference of gesture performance brings challenge for accurate gesture recognition. In this paper, a personalized method of gesture recognition is proposed, which can analyze personal gesture features and automatically adjust gesture templates to improve recognition accuracy. The method was evaluated on a finger-worn device named Magic Ring that collected eight gestures from three subjects for one week testing. Results show the effectiveness of the method that average improvement of 16% in recognition accuracy has been achieved.