Android Software Based Multi-touch Gestures Recognition for Secure Biometric Modality

Abstract

Now a days the multitouch interfaces interactions with computing devices are changing rapidly. The common problems related to users for user authentication are selecting weak textual passwords or forgetting password. Hence we used biometric gesture based multitouch authentication system. In a biometric authentication system, the user gesture is given to the system. This system has enrollment and verification stage. In Enrollment stage, user's gestures are registered in the system by acquiring and storing biometric templates corresponding to the users. In verification stage, the users gestures is compared with the stored biometric users gesture in template, where all touch points are correctly labeled and ordered consistently and thus, system authenticate a user. The verification process starts with the user multi-touch gestures, which are mapped with x, y coordinates, time-stamps and labels. The 5 touch points from 5 fingertips are serially stored by the device. When all 5 touch point get correctly labeled, the fingertip points are normalized to length, location and orientation. The Dynamic Time Warping matching algorithm is used to compute the distance between each of the normalized stored user gestures and current user gesture. Finally, a dissimilarity score is calculated from the distances obtained and a decision is then made by comparing the dissimilarity score with a threshold to authenticate a user. Multitouch gesture based authentication system is implemented by using android OS in Samsung tablet. We created android app with login, user name, register three gestures trial as input and then gesture verification. With the euclidean or jaccard distance, only one distance is measured at a time with 80% accuracy.