Research On Dynamic Gesture Recognition Algorithm Based On Smoothed Pseudo Wigner-Ville Distribution

With the development of mobile communication,the concept of Software Define Radio(SDR)began to be popular in the beginning of 1990s.The traditional hardware equipment of wireless communication is the basic platform of Software Define Radio,and most of communication function is realized by the software,so this flexibility and the machine learning theory give the possibility of the of human body pose(gesture)recognition to us.In this paper,the extraction of SDR signals,improved harmony search algorithm and gesture recognition technology based on support vector machine are studied.The main work is as follows:Using the characteristics of SDR,communication function is realized by the software,we make it working in the WiFi mode.The effective information extraction from OFDM defined by 802.11a protocol is accomplished by PC,and the system's communication functions will not be influenced in this way.We call this way reuse of wireless signal already exist.For gesture recognition based on support vector machine(SVM),this paper makes a classification of 9 cases,including 8 gestures.In order to extract the effective datum of the corresponding classification,the feature extraction method based on the smoothed pseudo Wigner-Ville distribution(SP/WVD)is analyzed.The instantaneous power spectral density of the received signal is acquired by SPWVD,and the feature vector of SVM is obtained by using the frequency margin distribution of SPWVD.In order to achieve the identification of dynamic gestures,the sliding window is used to obtain the statistical data.The results show that,the feature extraction method has good effect,it can effectively improve the accuracy of gesture recognition.Some key parameters of support vector machine have great influence on the classification.In order to obtain better SVM parameters,the improved harmony search algorithm is used to optimize the support vector machine.In the process of tuning,standard harmony search algorithm does not consult the best accord in the harmony memory,in order to solve this problem,this paper proposes two kinds of improved Harmony Search,which is based on the idea of optimization method of artificial fish swarm algorithm.The idea of artificial fish feeding is used to obtain the new harmony through adjustment of optimal harmony,so as to speed up the convergence of the optimal solution.At the same time,the original method of tuning is kept to avoid local optimal solution in the process of tuning.Test results show that the improved algorithm accelerates the optimization of the key parameters of SVM.Based on the above ideas,this paper carries on the analysis to the new application of the software radio,and gradually analyzes the key points of the communication model,the support vector,and recognizes human hand posture without affecting the normal communication.In order to verify the feasibility and validity of this method,the simulation model and the actual measurement are used,the results prove the feasibility of the idea in this paper.