Gait based biometric personal authentication by using MEMS inertial sensors

Walking is one of the major human activities, and walking pattern is unique for each individual. Thus, human gait can be applied in biometric personal authentication. The traditional method for gait recognition is based on one or multiple cameras. With the rapid development of Micro-Electro-Mechanical System (MEMS), small light inertial sensors have been used for human identification so far. In this study, a gait based personal authentication method is proposed using MEMS inertial sensors. They are fixed in the smart shoes, collecting motion signals and transmitting them to the server. Then, gait parameters such as step length, cadence, stance phase, swing phase and the pitch angular are calculated and used as features for personal identification. A probabilistic neural network is proposed as a classification mechanism to uniquely identify different users. Experiments are conducted to validate the proposed method. By using two cross-validation techniques, the overall mean classification rate for 22 persons is up to 85.3 and 85.7% respectively, which demonstrates the effectiveness of the method.