Research On Gesture Recognition Algorithms For Interactive Teaching Interface

In order to solve the traditional classroom system,teachers in the teaching process because of the interaction with the teaching equipment to disperse the problem of excessive energy,in this thesis proposed an interactive teaching interface.Teachers can use subconscious and natural gestures to teach students a three-dimensional geometry lesson to reduce the cognitive load and operating load of teachers in operating multimedia devices and improve the efficiency of teacher lectures.Traditional gesture recognition algorithms have low recognition rate,poor promotion,and low robustness for similar gestures.It is difficult to meet the requirements of interactive teaching interfaces.Therefore,this thesis is supported by National Natural Science Foundation of China(No.61472163,61603151)National key R &D projects(2016YFB1001403)and Shandong Province key research and development projects(2015GGX101025).The interactive teaching interface is used as an application platform and the gesture recognition algorithm in the teaching interface is researched.The goal of this dissertation is to reveal the relationship between deep learning network training parameters and the recognition rate of the model,and to explore a new method of hand gesture recognition based on deep learning to solve the problems of poor recognition rate,low robustness,and poor generality in traditional gesture recognition algorithms.The innovation of this thesis is reflected in the following two aspects:(1)The design and Implementation of SCDDF(Shape Context Density Distribution Feature)algorithm.Based on the DDF(Density Distribution Feature)density feature based on geometric features,the shape context feature descriptor algorithm is combined and the SCDDF algorithm is proposed.By combining the advantages of the two algorithms,adding the spatial coordinate features such as the main direction of the gesture and the angle between the center of gravity of the finger and the main direction,the recognition rate of the SCDDF algorithm for the 10 similar gestures is increased to over 97%.The accuracy and robustness of the recognition are significantly improved,which is of great significance for somatosensory applications.(2)Deep learning gesture recognition algorithms are proposed.Based on the large database of gestures established in this thesis,a dynamic gesture sequence synthesis method isproposed in which the first and last frames are fixed and the middle frame traversal combination is proposed for a gesture database dynamic gesture sequence.For static gestures,a static gesture image restoration sequence algorithm is proposed to achieve the fusion of motion and state gesture training and recognition.On this basis,combining the influence of the training parameters of solver files to the depth learning model,the static gesture recognition algorithm based on deep learning,the dynamic gesture recognition algorithm based on deep learning and the dynamic and static fusion gesture recognition algorithm based on the depth learning are realized.In this thesis,we design and create an interactive teaching interface around a geometry class that proves that the vertebral volume is one-third of the volume of the cylinder.Combining gesture recognition algorithms based on deep learning and the SCDDF algorithm proposed by this paper,teachers can use natural and subconscious gestures to directly control the objects in the teaching interface to give students geometrical principles.Let the classroom is no longer a step-by-step process in PPT.