Dynamic Gesture Authentication Method Based On Embedded Accelerometer In Mobile Phone

Mobile terminal dynamic gesture authentication scheme based on embedded accelerometer is a hot issue in the field of mobile computing. It is widely used in mobile electronic commerce, remote control, electronic voting and other fields. How to achieve mobile terminal user authentication accurately and rapidly is an important research content in this field. This paper studies dynamic gesture user authentication scheme based on embedded accelerometer in mobile phone, the main research contributions are listed as following:1.In order to detect valid gesture acceleration data without additional noises, a new gesture endpoint detection algorithm based on energy entropy was proposed to achieve the automatic detection of valid gesture endpoint. Endpoint detection method based on energy entropy was used to analyze the energy distribution of acceleration signal in stationary section and gesture section using Short Time Fourier Transform in time-frequency domain. Energy entropy which can fully reflect the signal energy probability distribution uniformity degree was calculated. Then the endpoint of the gesture was determined according to the unified decision threshold. The experimental results show that, compared with the existing algorithms, the proposed method can complete valid signal endpoint detection more accurately. Besides, the acceleration signal is always polluted by random noises in the process of motivation and testing. A new algorithm based on wavelet packet and mutual information entropy was put forward for de-noising. In this new algorithm mutual information entropy was used on the basis of the traditional wavelet packet decomposition to calculate the mutual information entropy between adjacent decomposition components to measure the degree of correlation between high frequency and low frequency parts and choose the reconstruction components according to the turning point of mutual information entropy. The experimental results show that the algorithm not only keeps the basic characteristics of the original signal, but also eliminates high frequency noises and random jitter.2.In order to guarantee a better user experience degree, a good real-time performance is required. Aiming at improving the computing speed of DTW, in this paper, a new algorithm called half dynamic time warping HDTW was proposed. This method cuts the time sequence into two parts, calculate local optimal paths respectively and trace back to get the global optimal path according to the global minimum distortion sum value. Experimental results show that HDTW and DTW produce the same optimal path but the new algorithm effectively reduces the time complexity of computing. Besides, in order to improve the precision of dynamic hand gesture authentication algorithm based on DTW, two improved methods were put forward. The first improved method takes Mahalanobis distance as the measure of distance between different acceleration signals. Calculate the covariance matrix according to authentication templates and optimize time complexity according to the characteristics of covariance matrix. The second improved method combined with a classical matching algorithm Longest Common Subsequence LCS is proposed on the basis of dynamic warping. The experimental results show that this method can improve authentication precision effectively. The second improved method takes in Longest Common Subsequences method on the basis of global dynamic warping. It can measure the similarity of gestures from both global and local aspects, thus improving the authentication system’s ability to resist imitation attack effectively.3.A dynamic gesture authentication system was designed and implemented on the smart phone with Android platform. More than 5000 gesture samples performed by 20 persons were collected in the experiment. The results show that in the proposed DTW-STLCS algorithm, its false rejection rate(FRR) is 1.518% while its false acceptance rate(FAR) is 0%, and the average time for the authentication process takes about 60 ms. Which demonstrates that the accuracy and efficiency of this method are very high. Therefore, this method has great practical and promotional value.