Classification of holding hand for dynamic interface adjustment on mobile devices

This work proposes a gesture-based left and right hand recognition algorithm for dynamic interface adjustment by using the information of the accelerometer and the touch screen on a mobile device. We design a warping method that aligns accelerometer signals with their corresponding positions of unlock-screen gestures. By projecting samples onto feature space and using a support vector machine classifier, the proposed method yields high recognition accuracy. The experiments also show that the proposed method is able to make correct decisions for a new user without pre-gathering information from the user.