scholarly journals Multifeature Fusion Human Motion Behavior Recognition Algorithm Using Deep Reinforcement Learning

2021 ◽  
Vol 2021 ◽  
pp. 1-11
Author(s):  
Chengkun Lu
Keyword(s):  
Reinforcement Learning ◽  
Human Behavior ◽  
Performance Test ◽  
Recognition Algorithm ◽  
Human Motion ◽  
Identification Accuracy ◽  
Data Sets ◽  
Behavior Recognition ◽  
Data Set ◽  
Human Behavior Recognition

Through the recognition and analysis of human motion information, the actual motion state of human body can be obtained. However, the multifeature fusion of human behavior has limitations in recognition accuracy and robustness. Combined with deep reinforcement learning, multifeature fusion human behavior recognition is studied and we proposed a multifeature fusion human behavior recognition algorithm using deep reinforcement learning. Firstly, several typical human behavior data sets are selected as the research data in the benchmark data set. In the selected data sets, the behavior category contained in each video is the same behavior, and there are category tags. Secondly, the attention model is constructed. In the deep reinforcement learning network, the small sampling area is used as the model input. Finally, the corresponding position of the next visual area is estimated according to the time series information obtained after the input. The human behavior recognition algorithm based on deep reinforcement learning multifeature fusion is completed. The results show that the average accuracy of multifeature fusion of the algorithm is about 95%, the human behavior recognition effect is good, the identification accuracy rate is as high as about 98% and passed the camera movement impact performance test and the algorithm robustness, and the average time consumption of the algorithm is only 12.7 s, which shows that the algorithm has very broad application prospects.

Research on Human Behavior Recognition in Video Surveillance

2011 ◽  
Vol 255-260 ◽  
pp. 2276-2280
Author(s):  
Wen Qiu Zhu ◽  
Qian Qian Li ◽  
Jun Feng Man ◽  
Xiang Bing Wen
Keyword(s):  
Knowledge Base ◽  
Video Surveillance ◽  
Human Behavior ◽  
Dirichlet Process ◽  
Human Motion ◽  
Viterbi Decoding ◽  
Hierarchical Dirichlet Process ◽  
Behavior Recognition ◽  
Human Behavior Recognition ◽  
Current Behavior

For effectively solving human behavior recognition in video surveillance, a novel behavior recognition model is presented. New behaviors may be produced in the process of human motion, hierarchical Dirichlet process is used to cluster monitored feature data of human body to decide whether unknown behaviors occur or not. The infinite hidden Markov model is used to learn unknown behavior patterns with supervised method, and then update the knowledge base. When knowledge base reaches a certain scale, the system can analyze human behaviors with unsupervised method. The Viterbi decoding algorithm of HMM is adopted to analyze current behavior of the human motion. The simulation experiments show that this method has unique advantage over others.

3D Human behavior recognition based on binocular vision and face–hand feature

Optik ◽  
2015 ◽  
Vol 126 (23) ◽  
pp. 4712-4717 ◽  
Author(s):  
Qing Ye ◽  
Junfeng Dong ◽  
Yongmei Zhang
Keyword(s):  
Human Behavior ◽  
Binocular Vision ◽  
Behavior Recognition ◽  
Human Behavior Recognition

Based on spatio-temporal graph convolution networks with residual connection for intelligence behavior recognition

2021 ◽  
pp. 002072092199660
Author(s):  
Yinong Zhang ◽  
Shanshan Guan ◽  
Cheng Xu ◽  
Hongzhe Liu
Keyword(s):  
Human Behavior ◽  
Behavior Recognition ◽  
Convolutional Network ◽  
Skeleton Model ◽  
Basic Technology ◽  
Temporal Model ◽  
Temporal Graph ◽  
Spatio Temporal ◽  
Human Behavior Recognition ◽  
Behavior Characteristics

In the era of intelligent education, human behavior recognition based on computer vision is an important branch of pattern recognition. Human behavior recognition is a basic technology in the fields of intelligent monitoring and human-computer interaction in education. The dynamic changes of human skeleton provide important information for the recognition of educational behavior. Traditional methods usually use manual information to label or traverse rules only, resulting in limited representation capabilities and poor generalization performance of the model. In this paper, a kind of dynamic skeleton model with residual is adopted—a spatio-temporal graph convolutional network based on residual connections, which not only overcomes the limitations of previous methods, but also can learn the spatio-temporal model from the skeleton data. In the big bone NTU-RGB + D dataset, the network model not only improved the representation ability of human behavior characteristics, but also improved the generalization ability, and achieved better recognition effect than the existing model. In addition, this paper also compares the results of behavior recognition on subsets of different joint points, and finds that spatial structure division have better effects.

A Comparative Study of Genetic Algorithms and Gradient Methods for RM12 Turbofan Engine Diagnostics and Performance Estimation

2004 ◽  
Author(s):  
Tomas Gro¨nstedt ◽  
Markus Wallin
Keyword(s):  
Gas Turbine ◽  
Performance Test ◽  
Gradient Methods ◽  
Performance Testing ◽  
Performance Estimation ◽  
Data Sets ◽  
Data Set ◽  
Turbofan Engine ◽  
Local Optima ◽  
Bypass Ratio

Recent work on gas turbine diagnostics based on optimisation techniques advocates two different approaches: 1) Stochastic optimisation, including Genetic Algorithm techniques, for its robustness when optimising objective functions with many local optima and 2) Gradient based methods mainly for their computational efficiency. For smooth and single optimum functions, gradient methods are known to provide superior numerical performance. This paper addresses the key issue for method selection, i.e. whether multiple local optima may occur when the optimisation approach is applied to real engine testing. Two performance test data sets for the RM12 low bypass ratio turbofan engine, powering the Swedish Fighter Gripen, have been analysed. One set of data was recorded during performance testing of a highly degraded engine. This engine has been subjected to Accelerated Mission Testing (AMT) cycles corresponding to more than 4000 hours of run time. The other data set was recorded for a development engine with less than 200 hours of operation. The search for multiple optima was performed starting from more than 100 extreme points. Not a single case of multi-modality was encountered, i.e. one unique solution for each of the two data sets was consistently obtained. The RM12 engine cycle is typical for a modern fighter engine, implying that the obtained results can be transferred to, at least, most low bypass ratio turbofan engines. The paper goes on to describe the numerical difficulties that had to be resolved to obtain efficient and robust performance by the gradient solvers. Ill conditioning and noise may, as illustrated on a model problem, introduce local optima without a correspondence in the gas turbine physics. Numerical methods exploiting the special problem structure represented by a non-linear least squares formulation is given special attention. Finally, a mixed norm allowing for both robustness and numerical efficiency is suggested.

Sign in / Sign up

Export Citation Format

Share Document