Algorithm Research Of Gesture Recognition And Implementation By FPGA

With the rapid development of information technology and hardware performance,the intelligent level of equipment is getting higher and higher,and the way of human-computer interaction is becoming more and more simple and humanize.Gestures,as one of human's natural communication methods,are very suitable for the development needs of human-computer interaction.So,gestures are used for human-computer interaction increasingly.And the algorithm research of gesture recognition has also become a hot topic.However,the complex environment and changeable gestures have increased the difficulty of gesture recognition,and the practical application of human-computer interaction also puts forward higher requirements on the real-time performance of the algorithm.In this paper,the gesture recognition algorithm is used as the research object,and the algorithm is implemented by hardware.The main research contents are as follows:1.Research on gesture segmentation algorithm under complex background.Aiming at the problem that it is difficult for the existing gesture segmentation methods to efficiently and completely segment the static gesture from the image under the complex background,a gesture segmentation algorithm based on the skin color centroid and edge self-growing is proposed.First,the algorithm uses the centroid of skin color to divide the gesture area,which reduces the computation amount of subsequent edge detection.Then,the complete edge of the gesture is obtained by using the edge self-growing algorithm,and then difference operation between skin color and edge information,so the skin color background is separated out.Finally,the connected domain and morphological processing can be used to segment the gesture image.Through the verification of segmentation effect under different scenes,the segmentation effect of this algorithm is better.2.Research on gesture recognition algorithm based on convolutional neural network.In view of the low resolution and low complexity of gesture images,this paper optimizes a convolutional neural network model which is easy to implement by FPGA.This model is a lightweight modification of VGGNet,which keeps the depth of the network,but reduces the size of the input layer and the number of feature maps of the network,so the calculated amount and the number of parameters is also reduced.the recognition rate of the network model can reach 98.93% by verifying the 4321 pictures of 15 gestures.3.FPGA implementation of gesture recognition algorithm.Gesture recognition algorithm involves digital image processing,and it has large data throughput and high real-time requirements,so it is very feasible to use FPGA which can compute in parallel.Through modular design of gesture recognition algorithm,this paper uses FPGA to judge 15 network classification results,and 11 of them are defined as 11 static gestures.At the same time,a state machine with multiple delay judgments is designed.The remaining 4 classifications are used to achieve Definition and recognition of 7 dynamic gestures.The results show that the response time of gesture recognition using FPGA can reach 17 ms,which can meet the real-time requirement of human-computer interaction.