Biometric gait recognition based on wireless acceleration sensor using k-nearest neighbor classification

Due to the explosive growth in the number of users who rely on their phones and tablets for more and more of their daily interactions, protecting user's private information in mobile devices is extremely important in these days. To address the limitations of conventional authentication methods such as PIN or password-based security schemes, there has been a growing interest in developing authentication methods based on characteristic biometric features such as fingerprint, iris, face, voice, and gait. In particular, much attention has been devoted to the use of human gait patterns as a biometric due to its unobtrusive nature. In this paper, we propose six new gait signature metrics to represent characteristics of the gait of a user. These new metrics derive from the rate of changes of acceleration data (jerk). They consist of two parts: dynamic and static portions. We identified that the dynamic part clearly illustrates the characteristic of body movement from walking. After storing all users' reference gait metrics in the mobile device, the system applies a k-Nearest Neighbor (KNN) algorithm to find out the best match of the current gait signature metrics from the list of reference gait metrics. To validate the usefulness of the proposed metrics, we conducted a number of experiments and measured the accuracy of the gait signature authentication system. The results of our experimental study show that the proposed metrics are quite effective and the system can identify or authenticate individuals.