Toward Posture Recognition with Touch Screen Biometrics

Touch screen data from smartphones can be used to extract behavioral biometric features and enhance user experience of touch screen applications based on user's current environment, such as body postures. In this paper we propose a new method for selecting biometric features in order to recognize body postures when performing simple swipes on the touch screens of smartphones. We performed a controlled experiment in which participants performed the swipes in six different postures (standing and five variants of sitting postures). From the experimental data we extracted features describing the swipe movements and performed feature selection with our method, which leverages the t-test statistic to compare the similarity of values in each pair of postures for each feature. The results show that several touch pressure-related features that were extracted significantly contribute to body posture recognition of a user.