Research On Gesture Recognition Method Based On Radio Frequency Identification Technology And Its Interactive Application

Since the beginning of the information revolution,human beings have been exploring the possibility of human-computer symbiosis and proposed the concept of human-computer interaction technology,with the aim of improving people's quality of life and work efficiency.Traditional human-computer interaction requires users to communicate with robots through keyboards,mice,touch screens,etc.,and their operational constraints are no longer able to meet the increasing demands of users.With the development of new generation recognition technologies such as voice,image and gesture,the design concept of human-computer interaction has been changed from the operating rules of human adaptation to the adaptation of machines to human communication,which makes human-computer interaction appear in the most natural way possible.As a common way of human communication,gestures are characterized by simple,intuitive and rich semantics.Therefore,the research and application of gesture recognition in the field of human-computer interaction has always been a hot topic in academic research.At present,the research on gesture recognition mainly includes gesture visual image processing and hand wearable sensor data modeling.A gesture recognition research solution based on visual image processing requires a camera to capture gesture images,and the recognition effect is easily affected by illumination,skin color,and device precision.The cost of gesture recognition based on data gloves such as sensors is relatively high,and the sensor carrying the battery is poor in portability,which is not conducive to the popularity of daily life.In view of the above problems,this thesis proposes the use of RFID radio frequency identification technology to realize the recognition of dynamic gestures.The COTS RFID reader is used to collect the raw phase flow of the tag reflection signal,and the data initialization process is performed based on the characteristics of the Kalman filter and the phase periodicity;In the aspect of the feature extraction,the data is windowed for solving the discontinuity of tag reflection signal in the time domain,and using the concept of KL divergence to compare the discrete probability distributions of adjacent windows to determine the start and end endpoints of gesture execution,thereby extracting corresponding gesture fingerprint feature segments;in gesture recognition and feature matching In the link,the dynamic time warping(DTW)algorithm is used to calculate the matching degree of each one-dimensional component in the current gesture segmentation and the priori fingerprint database,and the DTW algorithm is pointed based on the classification tree and the horizontal normalization idea.Sexual optimization is improved,and finally the k-nearest neighbors algorithm is used to realize multi-tag fusion gesture recognition.In order to verify the recognition performance of the dynamic gesture recognition system,this thesis establishes a standard fingerprint database for the predefined nine gestures,and performs real-time dynamic gesture recognition experiments.In addition,for achieving the interactive application of dynamic gesture recognition,this thesis designs a four-legged robot gesture control system,which uses Bluetooth serial communication to realize the user's gesture control of the robot.The experimental results show that the profile of dynamic gesture can be effectively segmented and the error offset is acceptable.The average accuracy of the gesture recognition algorithm can reach 90%,which proves that the system has good feasibility and robustness in the real environment.