Hierarchical Two-Stream Growing Self-Organizing Maps With Transience for Human Activity Recognition

被引:30
作者
Nawaratne, Rashmika [1 ]
Alahakoon, Damminda [1 ]
De Silva, Daswin [1 ]
Kumara, Harsha [1 ]
Yu, Xinghuo [2 ]
机构
[1] La Trobe Univ, Ctr Data Analyt & Cognit, Melbourne, Vic 3083, Australia
[2] RMIT Univ, Sch Engn, Melbourne, Vic 3001, Australia
关键词
Neurons; Activity recognition; Self-organizing feature maps; Histograms; Video surveillance; Optical flow; Feature extraction; Forgetting; hierarchical learning; human activity recognition (HAR); neural networks; self-organizing maps; HISTOGRAMS;
D O I
10.1109/TII.2019.2957454
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
The rapid growth in autonomous industrial environments has increased the need for intelligent video surveillance. As a predominant element of video surveillance, recognition of complex human movements is important in a wide range of surveillance applications. However, the current state-of-the-art video surveillance techniques use supervised deep learning pipelines for human activity recognition (HAR). A key shortcoming of such techniques is the inability to learn from unlabeled video streams. To operate effectively in natural environments, video surveillance techniques have to be able to handle huge volumes of unlabeled video data, monitor and generate alerts and insights derived from multiple characteristics such as spatial structure, motion flow, color distribution, etc. Furthermore, most conventional learning systems lack memory persistence capability which can reduce the influence of outdated information in memory-guided decision-making resulting in limiting plasticity and overfitting based on specific past events. In this article, we propose a new adaptation of the Growing Self-Organizing Map (GSOM) to address these shortcomings by 1) adopting two proven concepts of traditional deep learning, hierarchical, and multistream learning, applied into GSOM self-structuring architecture to accommodate learning from unlabeled video data and their diverse characteristics, 2) address overfitting and the influence of outdated information on neural architecture by implementing a transience property in the algorithm. We demonstrate the proposed model using three benchmark video datasets and the results confirm its validity and usability for HAR.
引用
收藏
页码:7756 / 7764
页数:9
相关论文
共 29 条
[1]   Dynamic self-organizing maps with controlled growth for knowledge discovery [J].
Alahakoon, D ;
Halgamuge, SK ;
Srinivasan, B .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 2000, 11 (03) :601-614
[2]  
Basavaraj GM, 2017, 2017 2ND IEEE INTERNATIONAL CONFERENCE ON RECENT TRENDS IN ELECTRONICS, INFORMATION & COMMUNICATION TECHNOLOGY (RTEICT), P1516, DOI 10.1109/RTEICT.2017.8256851
[3]  
Bergstra J, 2012, J MACH LEARN RES, V13, P281
[4]  
Chaudhry R, 2009, PROC CVPR IEEE, P1932, DOI 10.1109/CVPRW.2009.5206821
[5]   Histograms of oriented gradients for human detection [J].
Dalal, N ;
Triggs, B .
2005 IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, VOL 1, PROCEEDINGS, 2005, :886-893
[6]   Actions as space-time shapes [J].
Gorelick, Lena ;
Blank, Moshe ;
Shechtman, Eli ;
Irani, Michal ;
Basri, Ronen .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2007, 29 (12) :2247-2253
[7]   Incremental Activity Modeling and Recognition in Streaming Videos [J].
Hasan, Mahmudul ;
Roy-Chowdhury, Amit K. .
2014 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2014, :796-803
[8]  
Kim D, 2017, IEEE IJCNN, P432, DOI 10.1109/IJCNN.2017.7965886
[9]  
Kohonen T., 2001, INFORM SCIENCES
[10]  
Kong Y., 2018, ARXIV180611230CS