Human Action Recognition: A Survey

Fu, Meixia; Chen, Na; Huang, Zhongjie; Ni, Kaili; Wang, Kun; Sun, Songlin; Ma, Xiaomei

doi:10.1007/978-981-13-7123-3_9

Cited by 5 publications

(3 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Human activity recognition (HAR) has been extensively explored and investigated, evidenced by the plethora of surveys that have been published on this subject (e.g., [17], [26], [9] and [12], among others), covering various domains, applications and methods employed, and have proved to achieve promising results. The kinds of data that are being handled by researchers in this area vary from common activities such as locomotion to specialized ones like nurse care [14].…”

Section: Background and Related Workmentioning

confidence: 99%

Classical machine learning and deep neural network ensemble model for GPS-based activity recognition

Kojima

Legaspi

Mishima

et al. 2021

Adjunct Proceedings of the 2021 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of The

View full text Add to dashboard Cite

Our KDDI Research team proposes an ensemble model for the Sussex-Huawei Locomotion-Transportation (SHL) recognition challenge using GPS data. During preprocessing, we corrected the GPS dataset that contains errors and missing values using a Kalman smoother. Since this smoother can be processed offline, it can be used to correct backforward using the recorded future locations in the dataset. Further, by using features that have similar distributions in the train subject's data and the other subjects' datasets, we could achieve robust feature selection across multiple subjects. The first stage of our ensemble model employs classical machine learning and deep neural network approaches independently, specifically, LightGBM and LSTM, respectively. The second stage calculates the weighted average of the outputs of both approaches. Our results show the improved accuracy contributed by our ensemble, suggesting that it effectively makes use of both statistical and nonstatistical features given the suitable base models. We confirmed that the use of Kalman-smoother, selection of features with similar distributions across subjects and ensemble modeling contributed to improving the accuracy of both train and validation datasets.

show abstract

Section: Background and Related Workmentioning

confidence: 99%

Classical machine learning and deep neural network ensemble model for GPS-based activity recognition

Kojima

Legaspi

Mishima

et al. 2021

Adjunct Proceedings of the 2021 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of The

View full text Add to dashboard Cite

show abstract

“…Yu et al [37] conducted a survey about human behavior recognition through two parts: action recognition and action prediction. Ramasamy et al [38] reviewed machine learning-based human behavior recognition methods and the latest technology and development of human behavior recognition, and discussed the future research and development directions. Besides, Fu et al [39] studied human behavior recognition-based on deep learning, and also presented four datasets in details.…”

Section: Related Workmentioning

confidence: 99%

A Systematic Survey on Human Behavior Recognition Methods

et al. 2021

View full text Add to dashboard Cite

Human behavior is an essential component of social interaction and is of great significance to identify and analyze human behaviors in a variety of fields. Due to the rapid development of computer vision and machine learning technology, machine with intelligence has started replacing human beings to observe, perceive and analyze the explosive growth of image and video data. Computer vision and machine learning-based human behavior recognition is one of these tasks, which has become a particularly hot research topic in many different fields, such as intelligent monitoring, human-computer interaction, smart home, virtual reality, and medical diagnosis. In this study, we survey systematically the popular methods, algorithms, models and well-known action datasets in human behavior analysis in the past two decades. In addition, the advantages and disadvantages of the methods are discussed and propitious future research directions are also presented. The results of this survey reveal that paradigms of human behavior analysis is being shifted from traditional RGB to RGB-D, from deep learning to more intelligent and automated deep reinforcement learning, and from fixed camera devices to portable devices and channel state information (CSI), and paradigms based on automated deep reinforcement learning and portable devices and CSI would become some hot topics for future research on human behavior analysis.

show abstract

“…In this section, we briefly review the related studies, Including those of using multi-stream neural networks and CNN for skeleton-based action recognition. More detailed analyses of related studies are available in surveys [29], [30].…”

Section: Related Workmentioning

confidence: 99%

Robust Multi-Feature Learning for Skeleton-Based Action Recognition

Wang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

Skeleton-based action recognition has advanced significantly in the past decade. Among deep learning-based action recognition methods, one of the most commonly used structures is a two-stream network. This type of network extracts high-level spatial and temporal features from skeleton coordinates and optical flows, respectively. However, other features, such as the structure of the skeleton or the relations of specific joint pairs, are sometimes ignored, even though using these features can also improve action recognition performance. To robustly learn more low-level skeleton features, this paper introduces an efficient fully convolutional network to process multiple input features. The network has multiple streams, each of which has the same encoder-decoder structure. A temporal convolutional network and a co-occurrence convolutional network encode the local and global features, and a convolutional classifier decodes high-level features to classify the action. Moreover, a novel fusion strategy is proposed to combine independent feature learning and dependent feature relating. Detailed ablation studies are performed to confirm the network's robustness to all feature inputs. If more features are combined and the number of streams increases, performance can be further improved. The proposed network is evaluated on three skeleton datasets: NTU-RGB + D, Kinetics, and UTKinect. The experimental results show its effectiveness and performance superiority over state-of-the-art methods.

show abstract

Human Action Recognition: A Survey

Cited by 5 publications

References 20 publications

Classical machine learning and deep neural network ensemble model for GPS-based activity recognition

Classical machine learning and deep neural network ensemble model for GPS-based activity recognition

A Systematic Survey on Human Behavior Recognition Methods

Robust Multi-Feature Learning for Skeleton-Based Action Recognition

Contact Info

Product

Resources

About