文献类型：期刊文献

英文题名：Attention Mechanism Based on Improved Spatial-Temporal Convolutional Neural Networks for Traffic Police Gesture Recognition

作者：Wu, Zhixuan[1,2];Ma, Nan[3];Gao, Yue[4];Li, Jiahong[2];Xu, Xinkai[5];Yao, Yongqiang[1];Chen, Li[6]

第一作者：Wu, Zhixuan

通讯作者：Ma, N[1]

机构：[1]Beijing Union Univ, Beijing Key Lab Informat Serv Engn, Beijing 100101, Peoples R China;[2]Beijing Union Univ, Coll Robot, Beijing 100101, Peoples R China;[3]Beijing Univ Technol, Fac Informat Technol, Beijing 100124, Peoples R China;[4]Tsinghua Univ, Sch Software, Beijing 100084, Peoples R China;[5]Beijing Union Univ, Demonstrat Ctr Expt Teaching Comprehens Engn, Beijing 100101, Peoples R China;[6]Beijing Inst Technol, Sch Mech Engn, Beijing 100081, Peoples R China

第一机构：北京联合大学北京市信息服务工程重点实验室

通讯机构：[1]corresponding author), Beijing Univ Technol, Fac Informat Technol, Beijing 100124, Peoples R China.

年份：2022

卷号：36

期号：08

外文期刊名：INTERNATIONAL JOURNAL OF PATTERN RECOGNITION AND ARTIFICIAL INTELLIGENCE

收录：;EI(收录号：20222812355449);Scopus(收录号：2-s2.0-85133887213);WOS:【SCI-EXPANDED(收录号:WOS:000821707100012)】；

基金：We really thank the anonymous reviewer's constructive suggestions. This work was supported by Beijing Natural Science Foundation (No. 4222025), the National Natural Science Foundation of China (Nos. 61871038 and 61931012)

语种：英文

外文关键词：Intelligent interaction; attention mechanism; human action recognition; spatial-temporal features; traffic police gestures

摘要：Human action recognition has attracted extensive research efforts in recent years, in which traffic police gesture recognition is important for self-driving vehicles. One of the crucial challenges in this task is how to find a representation method based on spatial-temporal features. However, existing methods performed poorly in spatial and temporal information fusion, and how to extract features of traffic police gestures has not been well researched. This paper proposes an attention mechanism based on the improved spatial-temporal convolutional neural network (AMSTCNN) for traffic police gesture recognition. This method focuses on the action part of traffic police and uses the correlation between spatial and temporal features to recognize traffic police gestures, so as to ensure that traffic police gesture information is not lost. Specifically, AMSTCNN integrates spatial and temporal information, uses weight matching to pay more attention to the region where human action occurs, and extracts region proposals of the image. Finally, we use Softmax to classify actions after spatial-temporal feature fusion. AMSTCNN can strongly make use of the spatial-temporal information of videos and select effective features to reduce computation. Experiments on AVA and the Chinese traffic police gesture datasets show that our method is superior to several state-of-the-art methods.