Research And Application Of Dynamic Gesture Recognition Based On PE-DTW Algorithm

In recent years,With the rapid development of advanced science and technology and the increasing popularity of smart home equipment,smart TVs have gradually become the control center of the home Internet of Things.However,in daily use,the traditional remote control has certain dependence and limitations.Gestures,as a human-computer interaction method that fits people's various communication habits in daily life,are natural and comfortable.Compared with wearable devices,gesture operations through cameras are more in line with people's daily needs.Therefore,The human-computer interaction system based on visual dynamic gesture recognition has high research value for smart TVs.At present,vision-based dynamic gesture recognition mainly has problems such as being easily affected by the surrounding lighting environment,inaccurate gesture detection,and poor recognition effect.This paper proposes a dynamic gesture recognition based on a Prematurely ended-Dynamic Time Warping algorithm,which improves the robustness of the algorithm in low light and the efficiency of gesture recognition.For the accuracy and real-time performance of the smart TV operation,this article combines the Kinect v2 depth camera and uses Kinect dynamic gesture recognition technology to design and implement the smart TV gesture operating system.The main work of this paper are as follows:(1)After fully analyzing the common operations involved in the smart TV gesture operating system,7 dynamic gestures are defined,and a standard gesture template library is established through the Kinect v2 depth camera collection.However,the gesture operations commonly used in smart TV have certain similarities,and users are prone to problems such as deviations in actual operations.Therefore,this article sets the recognizable area of gesture operations,which not only reduces the invalid HOG gesture feature direction,but also avoid misrecognition of gestures in adjacent areas,and improve the comfort of the user's gesture operation.(2)Aiming at the problem of judging multiple gestures in a continuous image sequence,this paper sets the identification flags for the start and end of the gesture operation,which reduces redundant trajectories at the start and end points,and avoids the phenomenon that the test sequence and the reference sequence will have delayed data characteristics at both ends.(3)This paper proposes dynamic gesture recognition based on the PE-DTW algorithm.In view of the complex calculations of the traditional DTW algorithm,in the case of increasing the parallelogram global constraint path,combined with the LB?NKeogh lower bound function to eliminate part of the reference sequence in advance,thus greatly reducing the amount of calculation.The experimental results show that the accuracy of the gesture recognition system reaches 94.7%,and the average recognition rate is 0.225 s,which meets the real-time operation requirements of smart TV.(4)This article successfully applies the dynamic gesture recognition based on the PE-DTW algorithm to the smart TV system,and simulates the dynamic gesture operation of the smart TV.The Kinect v2 depth camera acquires and recognizes gesture operations,and operates the smart TV system in real time through the corresponding mapping relationship with the smart TV system.The experimental results show that the smart TV can be operated more easily and efficiently through dynamic gestures,which meets the needs of people's smart home interactive experience,and has important practical significance in the future home Internet of Things.