Paper Info
Title
Gesture Recognition Network Design Based On Deep Convolutional Networks And Spatial Enhancement Block
Abstract
Gesture interaction is a natural way of human computer interaction. In this paper, a deep convolution network is designed to recognize the gesture in complex scenes. The Gesture Recognition Network consists of basic network, adapted SE blocks, and two identification modules, which achieves fast recognition with high accuracy. Besides, we build a data set with complex background and varying illumination for train this network. This data set contains 3289 photos and 15 gestures. In this data set, the number of anchors in practical application scenarios is analyzed in order to balance the recognition speed and accuracy. Data augmentation techniques are used to increase data diversity. We have released this data set on the Internet for public use1. The learning rate is adjusted based on the loss value to improve the convergence speed. Our proposed network only needs a two-dimensional picture from a low-cost camera and achieves 42.8 frames per second, which can be useful for some practical applications, like a mobile robot.
Year
DOI
Venue
2018
10.1109/ROBIO.2018.8665217
2018 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND BIOMIMETICS (ROBIO)
Field
DocType
Citations 
Convergence (routing),Computer vision,Network planning and design,Gesture,Gesture recognition,Feature extraction,Control engineering,Artificial intelligence,Frame rate,Deep learning,Engineering,Mobile robot
Conference
0
PageRank 
References 
Authors
0.34
0
4
Name
Order
Citations
PageRank
Qinglin Li100.68
Biao Hu2299.98
Junkuan Li300.34
Cao Zhengcai44216.38