Gait Learning Based Authentication for Intelligent Things

Identity authentication plays an important role for the safety of smart terminals. Most existing schemes use biological features such as the iris and the fingerprint for identity authentication, which can not implement real-time and continuous identification of user identity. In light of this, we propose a feature extraction and fine-grained authentication scheme based on gait data in this paper. The proposed scheme contains a comprehensive data preprocessing mechanism for human gait data based on the mutual information model and Principal Component Analysis (PCA) model, as well as an identification mechanism using the Support Vector Data Description (SVDD) model and Long Short-Term Memory (LSTM) model, which is convenient for data collection and easy deployment. To evaluate the performance of the proposed scheme, we conduct experiments with human gait data collected by smartphones, which shows that our authentication scheme possesses a higher identification accuracy compared with other existing schemes.