Research On 3D Static Gesture Recognition Algorithm Based On Space Coordinates

Gesture,as a kind of body language,has been the most widely used communication method since the development of society,and it is the most suitable way of communication among people.The technology of gesture recognition is to let the computer recognize gestures and understand the meaning of diverse gestures,and make intelligent feedback.With the continuous development of science and technology,people have entered an unprecedented information age in production and lifestyle,which has made people put forward higher expectations for the performance and use of media equipment,and increasingly pursue more natural human-computer interaction.Gestures have gradually become an intelligent human-computer interaction trend in addition to classic interaction methods such as mouse,keyboard and buttons.In this paper,the research of 3D static gesture recognition based on spatial coordinates is oriented to system control applications in the field of human-computer interaction,using depth camera Leap Motion and machine learning theories,starting with feature extraction and model building,establishing effective gesture feature sequences and model of gesture recognition,to realize the accurate and rapid recognition of multiple types of custom static gestures,and provide theoretical support for the wide application of the technology in the field of system control.The main work of this paper:(1)Obtain gesture skeleton data based on Leap Motion,establish standard gesture feature vectors and gesture databases by analyzing different gesture features,and perform visual macroscopic analysis of gesture data sets based on PCA algorithm;(2)In establishing standards Based on the gesture database,considering whether there is an obvious feature learning process on the work efficiency of the gesture recognition model,the PSO-kNN algorithm and the PGA optimized Multiclass-SVC gesture recognition model are built respectively,and the hyperparameter optimization,training and testing of the model are carried out.Comparison of classification model gesture recognition accuracy.The experimental results show that the PGA-optimized Multiclass-SVC model has the best gesture recognition effect.The gesture recognition accuracy rate can reach 97%,and the gesture recognition rate is 0.001 seconds.The gesture recognition effect of the PSO-kNN model is the second,and the gesture recognition accuracy rate is good.Up to 93%,the gesture recognition rate is 0.007 seconds.(4)Design and develop a music player system based on gesture recognition.In this case,the volley interaction between gestures and music player is realized,which preliminarily verifies the feasibility and effectiveness of the gesture recognition method studied in the paper in the system control application field.