Random forest based-biometric identification using smart shoes

This study presents a biometrie identification based on gait (with shoe wearable sensors). Biometrie identification is an excellent method to often alternate inconvenient interaction such as PIN and patterns in smart device. To help elderly person who cannot control smart devices by themselves, it is required to assist automatic personalization by identifying users sharing a device. In this study, we proposed an algorithm combined the discrete cosine transform for detecting frequency feature and random forest which classifies subjects. We performed an experiment for 8 subjects by walking with the smart shoes. Finally, the result demonstrates a user recognition accuracy of 97.9 % and an equal error rate of 2.4%.