Research And Design Of Gesture Recognition System Based On Touch Perception

Gesture recognition will be important in the future.Gesture recognition technology,on the one hand,as a new human-computer interaction mode,is widely used in entertainment,intelligent control and other fields.On the other hand,the sign language is converted into speech signals through gesture recognition technology,which helps the deaf-mute to communicate with normal people.Gesture recognition system based on tactile sense has the advantages of low cost and easy to be wearable.Therefore,this paper studies and designs gesture recognition system based on tactile sense.The main research contents are as follows:(1)Investigate the bending sensor,piezoresistive sensor and optical waveguide sensor that can realize tactile sensation,and analyze the sensing principle and parameters of each sensor.According to the characteristics of fingers,piezoresistive sensors are selected and corresponding sensor arrays are customized to make tactile gloves for gesture perception.(2)Complete the circuit system design according to the requirements of gesture recognition system.The gesture recognition system mainly completes the tasks of data collection,data classification and output classification results.Based on this,the data acquisition circuit is designed with the analog switch ADG731 and the integrated operational amplifier AD8651 as the core,the data processing circuit is designed with the advanced neural network integrated circuit,and the data transmission circuit of bluetooth and serial port are designed.Each circuit module works in order under the control of the single chip microcomputer.Finally,the results of gesture recognition are presented by the upper computer software and the manipulator which can be developed by the secondary programming.(3)Study gesture recognition algorithm.In this paper,I study the principle of decision tree,support vector machine,neural network and binary neural network algorithm,and analyzes the characteristics of each algorithm.Taking the digits 0 to 9 in American sign language gestures as the recognition target,the data set was collected,the algorithm model was trained,and the recognition accuracy and execution time of each algorithm were evaluated.The experimental results show that the above algorithms can achieve more than 98% recognition accuracy in this design,which proves that this design has good gesture perception ability.The decision tree algorithm has the fastest computing speed and the neural network has the slowest computing speed when each algorithm model is run on the platform of microcontroller,and the computational speed of neural network algorithm can be improved effectively by the neural network integrated circuit.