Research On Hand Gesture Segmentation And Recognition Based On Deep Learning

With the rapid development of information technology,human-computer interaction has become an important part of people's daily life.It is widely used in smart home,medical education and virtual reality.Because of its natural and direct,hand gesture interaction has become a hot research area in the field of human-computer interaction.Due to the influence of light,complex background and hand shape changes,hand gesture segmentation and recognition are still the main research contents of gesture interaction.In recent years,the great success of deep learning in computer vision tasks has brought new research methods and ideas to hand gesture interaction.This paper makes a deep research of hand gesture interaction based on computer vision.A new hand gesture segmentation and recognition algorithm based on deep learning is proposed,which realizes accurate and fast hand gesture recognition.The main work of this paper includes the following aspects:?.In response to hand gesture segmentation in complex scenes,the Full Convolutional Network is introduced into hand gesture segmentation field,and a FCN gesture segmentation model based on VGG-16 is proposed and constructed.Firstly,the full connection layer of VGG-16 is replaced with convolution layer,and the upsampling layer is increased.Then the skip structure is used to extract the details.Finally,the algorithm is tested on the OUHANDS dataset.The experimental results show that FCN gesture segmentation model is robust to light,complex background and facial shade.?.In response to the problems of losing edge details and slow segmentation speed of FCN segmentation algorithm,a kind of hand gesture segmentation algorithm based on the Full Convolution Residual Network is proposed.Using the idea of removing fully connection layer in FCN algorithm,the Residual Network is used as the basic network,and the atrous spatial pyramid pooling module is added to realize hand gesture segmentation that fusion context and multi-scale information.The algorithm is tested on the OUHANDS dataset.The experimental results show that the algorithm not only ensure the segmentation speed,but also retain the gesture details better,and the average accuracy is improved to 91.45%.?.Through a brief analysis of each classification algorithm,a deep convolutional neural network model for hand gesture recognition is built,the model has the characteristics of fewer parameters and less computation.In order to improve the performance of gesture recognition,the batch normalization layers and the global average pooling layer are added to the model.Using the hand gesture segmentation results in this paper,the model is tested.The experimental results show that the hand gesture recognition algorithm proposed in this paper has high accuracy and fast recognition speed.A good strategy for real-time hand gesture recognition is provided.