Subtle finger motion recognition based on the cots smartwatch
Jiahao Wang, Liquan Qian, Yingzi Xie, Qiuling Long · 2017
Current researches on muscle activity reorganization always require specialized instruments like Biosensor or SignAloud, which usually have strict limitations and only fit for scientific research. With the development of sensing and computing capabilities of COTS mobile devices, such wearable devices represented by smartwatches, provide a promising solution to recognition of human finger gestures. Besides arm gesture, this paper puts forward an elaborate mechanism to extract subtle finger motion segment precisely through accelerometers in the COTS smartwatch. The first step is to extract the weak features representing the behaviors of the finger gesture through an IAMSE algorithm by sliding windows. After the motion segments obtained, a motion model can be constructed by different classification algorithms, which can identify the subtle motion of the human finger precisely. The experimental results show that the accuracy of the system can reach as high as 88% when identifying difference subtle finger gestures of protruding the fist state from 1 to 5 fingers, which provides a viable reference scheme to achieve more fine motion recognition and its related applications by using the smartwatch.