Design And Implementation Of Real-time Gesture Authentication System Based On Millimeter Wave Sensing

With the all-round popularity of 5G,the high-speed data transmission requirements make millimeter wave appear in people’s lives.It can not only be used in the field of data transmission,but also show great advantages in intelligent perception.Because of its high detection accuracy,millimeter wave technology is mature in gesture recognition,but there are few solutions to the problem of how to identify the user who performs gestures.This thesis presents a method of gesture authentication based on millimeter-wave perception,designs and implements a gesture identification system,and verifies its validity.This thesis first uses millimeter-wave radar to detect distances for real-time gesture perception,and provides a certain data cache to ensure the continuity of gesture movements,which provides high-quality raw data for gesture modeling.Secondly,the original signal is preprocessed,and precise gesture reflection point cloud is filtered out based on the information of distance and reflection intensity.Eight local eigenvalues,such as distance,speed,energy,angle of arrival and spatial coordinates,are calculated.Range-Doppler energy map is used as the global transformation feature of gesture in terms of distance and speed,describing the details of gesture from two aspects,eliminating unrelated signal reflection.Finally,a lightweight residual network is designed to realize user identification:at the beginning of the network,two subbranches are used to process local and global features respectively,and at the end,the full connection layer is fused to comprehensively judge user identity;In the last part of authentication,a user similarity computer system is designed to further verify the identity,so as to enhance the security of the system.The results show that the recognition accuracy of the gesture authentication system designed in this paper is 94.42%under the test of 20 users and 86.44%under the real-time authentication environment.Moreover,it can effectively eliminate unrelated actions,and the misjudgement rate of malicious imitation invasion to illegal users is 3.7%;The system authentication time is 1.7 seconds,which guarantees the user experience of real-time authentication.