Design And Implementation Of Gesture Recognition Algorithm Based On Neural Network

In recent years,VR(virtual reality)devices have become more and more popular,leading technology companies around the world have also laid out the VR market.The data glove is the research hotspot at home and abroad.However,most of the current data glove hand attitude acquisition uses ine rtial sensors,which occupy a large size of gloves,wear inconvenient.In addition,due to the problem of gesture recognition algorithm,there are some shortcomings,such as low precision,low real time and long training time.The main work of this paper is to design a gesture recognition algorithm based on BP neural network and implement an experimental platform.In this paper,Forte data glove is used for data glove.Forte data glove finger sensor adopts bending sensor Flex,which has the mapping between angle and resistance.As a result,the size of the finger sensor is greatly reduced.According to the characteristics of Forte data glove,the conduction relationship between the sensor and finger of Forte data glove is calculated by analyzing the biology and mechanics of human palm according to the degree of fit between the glove and finger.This paper selects the fingerThe index finger is used as an example of the hand movement model,and the whole palm motion model is built based on the index finger.The main innovation of this paper is in gesture recognition,based on BP neural network,aiming at the slow convergence speed of the traditional BP neural network gesture recognition algorithm,it will lead to the calculation of easy access to the minimum local value.Especially in the unstable error function curve defects.A method of modifying BP neural network by GA and chaos algorithm is designed,in which the GA algorithm can optimize the weights globally,and the chaos algorithm can modify the reverse le arning process.In the optimization process,the minimum error method is used for gesture matching.This can effectively reduce the interference of the gesture switching process.Therefore,a new CGA-BP neural network algorithm is designed,which has the ch aracteristics of high precision,high real time and short training time.Finally,a simulation experiment platform based on Unity3 D and Visual C is designed and implemented.At the end of this paper,the simulation experiment and test of the designed data glove show that the glove can respond the hand attitude to the computer accurately and in real time.Finally,the experimental results show that the CGA-BP neural network algorithm designed in this paper can effectively solve the ability of hand gesture recognition and processing data,and has a great improvement in precision,real-time and training time.The size of the finger sensor is also greatly reduced,making wearable devices more convenient.