2017 年 22 巻 1 号 p. 41-50
In recent years, gesture recognition using surface-electromyogram (sEMG) has become popular and these approaches can classify gestures at a high recognition rate. In particular, easily-mountable sEMG based gesture recognition devices have been developed. When developing such devices for use in daily lives, it is necessary to consider situations where users will be using the device while gripping objects, such as umbrellas, bags and so on. By using sEMG of the forearm, it is possible to still collect data without intruding on what the user is doing in these situations. However, sEMG based gesture recognition while gripping an object still lacks sufficient investigation. Therefore, in this study, in order to investigate the feasibility of gesture input while gripping an object, we performed an experiment to measure recognition accuracy for four hand gestures of users while they are holding a variety of objects. From the results, we discuss both feasibility and problems of gesture input while gripping objects and propose a new approach to resolve those problems.