Research Of Gesture Recognition Based On Data Glove

With the development of the human interaction technology,gesture as a kind of natural,intuitive and easy to learn means of human-computer interaction has become a research hotspot.With the continuous improvement of sensor technology,gesture recognition technology of based on data glove has been widely used.On the basis of the research status at home and abroad,this paper designed the three static gesture recognition methods,the method based on BP neural network,based on support vector machine(SVM)method and template matching method.This article uses 5DT company research and development of 5DT Data Glove type Ultra 14 Data Glove.According to the method of gesture recognition designed in this paper and the characteristics of the signal,Collected for the sensor signal is processing,it mainly includes data normalization,filtering processing of data.To analyze the gestures,to determine gestures sample set,select 7 kinds of suitable for static gestures as experimental objects.Secondly,this design mainly researches 3 kinds of algorithm of gesture recognition in view of the feature.The BP neural network method mainly includes two aspects of training and recognition.Applying the sampling data during processing as training data of BP algorithm,test data were sent to trained neural network to validate.Gesture recognition based on support vector machine(SVM)model mainly includes the creation of model and gesture recognition.After the model algorithm is studied,support vector machine(SVM)mode and least squares support vector machines(LS-SVM)model are set up respectively by One-on-one classification method,and compare the recognition rate through the establishment of the model for gesture recognition.Template matching method mainly includes the analysis of characteristics of gestures,to set up the gesture template and test,including establishing gestures template to determine the threshold range of each sensor of hand joint through many experiments.Finally,by comparing three kinds of gesture recognition time,the recognition accuracy and feasibility of the recognition method,template matching method is used to realize real-time gesture recognition,anduse the Visual Studio development platform,using C#language programming complete control aircraft by gesture,to verify the feasibility of the method.Gesture recognition of off-line analysis experiment by MATLAB programming software,the simulation is verified the possibility of three kinds of gesture recognition algorithm,to obtain a high recognition rate.In experiment,the average recognition rate of gesture recognition based on BP neural network is 87.1%,the average recognition rate of gesture recognition based on support vector machine(SVM)including the LS-SVM model recognition rate is 88.6%,the SVM model recognition rate is 91.4%,and Template matching method of gesture average recognition rate of 85.7%.The SVM model system recognition rate is the highest,the rate of template matching method is the lowest,however the recognition time of the template matching method is the shortest,and it programming easy to implement.