Research And Design Of Authentication System Based On Human Motion Recognition

In the current era of information society,smart phones as the carrier of so many sensitive information such as social information, financial information, file information and other data, mobile phone information security is more and more important. The authentication and login technology of mobile phone has become a very important and practical information protection technology. At present, the authentication technology for smart phones mainly includes fingerprint identification,ultrasonic fingerprint identification, common character password and swipe gesture password authentication. In the aspect of hardware cost,fingerprint recognition and ultrasonic fingerprint recognition technology is higher than other technologies, while the defects of common character password and swipe gesture lie on that if the user set simple password, it will be easy to break; on the other hand, if the password is too complex, it will reduce the usability of the authentication system.Aiming at the security and usability of smart phone authentication technology, this paper proposes an authentication scheme based on fine grained human acctivity recognition. Users only need to wave the smart phone to draw a three-dimensional gesture motion, and the system extracts the real motion data by using ending points detection algorithm and implements dynamic time warping algorithm with restricted slope of the search path to authenticate login. This will greatly enhance the user experience.This paper mainly does the following work:1. Study the background and significance of this research. Research and summarize the current development of the biological authentication technology.2. Research and study the biological authentication technology, signal processing and matching technology. To improve the security of the authentication system, this paper also studies the technology of data encryption storage and transmission.3. Study the motion data starting and ending points detection and extraction algorithms and human motion gesture recognition algorithms.The sampled motion gesture raw data has redundant data at the motion beginning and ending positions, which will affect the accuracy of the algorithm greatly. This paper presents a combination of discrete data difference method, linear interpolation method and threshold processing algorithm, which can effectively extract the real motion data, and insensitive to the fluctuation that near the motion starting and ending positions. The algorithm first implements differential operation on the original motion data, and then extracts the peak points by using linear interpolation algorithm, finally uses the threshold method for extracting the real motion data.4. Study authentication algorithm and pattern updating scheme. Use the restricted DTW algorithm to adapt the efficiency of the algorithm on mobile phone.5. Design and implement the motion authentication system. This project consists of two parts: client side and server side. The client side authentication system based on Android application, which is the key point of this research; the server side is mainly used for training model parameters for client, data storage and authentication function simulation.6. Test the motion authentication system. This paper conducts experiments to verifying the efficiency of ending points detection algorithm, and tests how the different pre-treatment strategy affect the accuracy of the authentication algorithm and then using 6 evaluation parameters to choose the optimal theshhold for the authentication system.In addition, the function of the system has been tested, to ensure the system running normally.