Human Action Recognition Method Based on Multi-Attention Mechanism and Spatiotemporal Graph Convolution Networks

Abstract

Human action recognition has become a challenging task in computer vision because it is difficult to combine spatiotemporal information. A multi-attention spatiotemporal graph convolution network is proposed.The core idea is to construct a connected graph according to the time series information and natural connection of human skeleton, and use the spatiotemporal graph convolution network with multi-attention mechanism to automatically learn spatial and temporal features and optimize the connected graph to realize prediction. Graph attention module is introduced, the topological structure of the graph constructed by the model will be optimized with the process of network training after initialization, then the topological structure which is more suitable for expressing human actions will be obtained. In addition, the channel attention module is added to make the network pay more attention to the important channel information, so as to extract the features of describing actions more effectively. A large number of experiments are carried out on the recognized large datasets: NTU-RGDB and Kinectics, which show that the method has higher recognition accuracy.

Effective factors of icons searching performance based on visual attention capture for the nuclear power plants

“…It has been found that circular shapes contribute to higher cognitive efficiency than other contours; shapes with topological differences are more visually shapes with topological consistency 1 . The purpose of this experiment was to verify the existence of an influential role of different topologies of icons on visual search performance.…”

“…It has been found that circular shapes contribute to higher cognitive efficiency than other contours; shapes with topological differences are more visually shapes with topological consistency. 1 The purpose of this experiment was to verify the existence of an influential role of different topologies of icons on visual search performance. Therefore, this paper taking data icon shapes of nuclear power interface as sample, conducted experiences after designing simple shapes with topological differences in the aspects of shape adjacency, connectivity, and type of constituent elements.…”

“…Thus, the display of information about the injection situation is clearly extremely significant. To explore the identification performances under different time pressures, researchers have studied the influence of complex digital interfaces on color and shape codes 1,2 . The layout design of fighter radar situation‐interface has been evaluated experimentally and has been analyzed through an objective evaluation technology of eye's tracker 3,4 .…”

“…To explore the identification performances under different time pressures, researchers have studied the influence of complex digital interfaces on color and shape codes. 1,2 The layout design of fighter radar situation-interface has been evaluated experimentally and has been analyzed through an objective evaluation technology of eye's tracker. 3,4 Jin et al 5,6 evaluate situation awareness of digital graphical interfaces studied the human-computer interaction of the digital.…”

Key frame extraction method for lecture videos based on spatio-temporal subtitles

“…Finally, a steady-state key frame selection method for video is proposed by combining curve edge detection and subtitle existence threshold. Unlike the key frame extraction method in the deep learning perspective [ 20 , 22 ], which has the defects of complex feature extraction, difficult algorithm framework and large computational amount [ 7 , 9 , 13 , 21 ], the proposed method adopts the spatio-temporal slice technique in the video feature extraction process to achieve the local operation, which greatly reduces the computational amount. Therefore, the method of this paper is helpful to achieve accurate and fast extraction of the lecture video key frames and it is interesting and worthwhile to research.…”