Article
Computer Science, Information Systems
Himadri Bhuyan, Jagadeesh Killi, Jatindra Kumar Dash, Partha Pratim Das, Soumen Paul
Summary: This paper presents a method to understand and classify Bharatnatyam dance motion. By analyzing the captured dance performances, it can assist in cultural heritage preservation and tutoring systems. Machine learning algorithms are used to recognize motion based on different features, achieving an overall accuracy of over 90% using Support Vector Machine (SVM) and Convolutional Neural Network (CNN) techniques.
Article
Computer Science, Artificial Intelligence
Vittorio Mazzia, Simone Angarano, Francesco Salvetti, Federico Angelini, Marcello Chiaberge
Summary: This research introduces an attention-based Action Transformer (AcT) architecture that outperforms current mix networks in human action recognition, leveraging small temporal windows of 2D pose representations for low-latency real-time performance. Additionally, a large-scale dataset called MPOSE2021 has been open-sourced to serve as a benchmark for training and evaluating real-time, short-time HAR. The proposed methodology was extensively tested on MPOSE2021, showcasing the effectiveness of the AcT model and setting the groundwork for future HAR research.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Artificial Intelligence
Yuan Tian, Yichao Yan, Guangtao Zhai, Guodong Guo, Zhiyong Gao
Summary: Efficiently modeling spatial-temporal information in videos for action recognition is crucial. This paper proposes a unified action recognition framework that adapts to the dynamic nature of video content by introducing dynamic-scale spatial-temporal kernels and sparse interactions among selected foreground objects. The framework also includes a novel latent motion code module to improve performance.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2022)
Article
Computer Science, Artificial Intelligence
Adrian Nunez-Marcos, Gorka Azkune, Ignacio Arganda-Carreras
Summary: This article explores the development and research status of egocentric action recognition (EAR) field, including the increase of egocentric video data and the challenge of action recognition. A taxonomy is proposed to classify methods more accurately, a review of zero-shot approaches is provided, and datasets used by researchers are summarized.
Article
Computer Science, Artificial Intelligence
Santosh Kumar Yadav, Aayush Agarwal, Ashish Kumar, Kamlesh Tiwari, Hari Mohan Pandey, Shaik Ali Akbar
Summary: This research proposes an expert-level yoga asanas recognition system based on two-stream deep learning, achieving impressive accuracy through real-time correction of multi-person yoga postures.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Mathematics, Interdisciplinary Applications
Rui Ma, Zhendong Zhang, Enqing Chen
Summary: This paper introduces a deep convolutional generation confrontation network for human motion pose recognition, utilizing stacked hourglass network to accurately extract key joint points and designed generator and discriminator to show spatial relationships among human body parts.
Article
Chemistry, Multidisciplinary
Tao Zhang, Yifan Wu, Xiaoqiang Li
Summary: This paper proposes a novel dilated multi-temporal (DMT) module for modeling multi-temporal information in action recognition. It uses dilated convolutions with different dilation rates in different feature map channels to capture information at multiple scales. The DMT module can be integrated into existing 2D CNNs, making it a straightforward solution for multi-temporal modeling.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Interdisciplinary Applications
Pawan Kumar Singh, Soumalya Kundu, Titir Adhikary, Ram Sarkar, Debotosh Bhattacharjee
Summary: This survey provides an overview of the various approaches proposed for Human Action Recognition (HAR) in the past decade, focusing mainly on the development of methods for unimodal HAR using concepts of machine learning and deep learning models. It also includes discussions on different feature extractors, majorly used video and still-image datasets, and offers insights into future work scope.
ARCHIVES OF COMPUTATIONAL METHODS IN ENGINEERING
(2022)
Article
Computer Science, Information Systems
Nan Ma, Zhixuan Wu, Yiu-ming Cheung, Yuchen Guo, Yue Gao, Jiahong Li, Beijyan Jiang
Summary: This paper provides a comprehensive review of the recent developments in human action recognition and posture prediction. It discusses the application of deep learning technologies and emphasizes the importance of these techniques in intelligent interaction and human-computer cooperation, using interactive cognition in self-driving vehicles as an example.
TSINGHUA SCIENCE AND TECHNOLOGY
(2022)
Article
Computer Science, Information Systems
Jian-Wei Cui, Han Du, Bing-Yan Yan, Xuan-Jie Wang
Summary: This article proposes an upper limb action intention recognition method based on the fusion of posture information and visual information, in order to address the problems of decoding information and low recognition rate with EMG and EEG signals. By combining inertial sensors and visual information, the method achieves accurate recognition of upper limb motion and target objects, enabling effective control of prosthetic hands. The experimental results demonstrate the feasibility and practicality of the proposed method.
Review
Computer Science, Hardware & Architecture
Xiankai Huang, Zhibin Cai
Summary: This paper summarizes and analyzes existing video action recognition methods based on 3D convolution to help new researchers understand this field. Firstly, it introduces the classical methods and points out the problems. Then, it summarizes the existing improved methods and compares the experimental results on benchmarks, discussing current challenges and future trends.
COMPUTERS & ELECTRICAL ENGINEERING
(2023)
Article
Computer Science, Information Systems
Tasweer Ahmad, Syed Tahir Hussain Rizvi, Neel Kanwal
Summary: In this paper, a novel idea of action embedding with a self-attention Transformer network is proposed for skeleton-based action recognition, which effectively models the latent information of skeleton data and captures both spatial and temporal features of joints. Experimental results on SYSU-3D, NTU-RGB+D, and NTU-RGB+D 120 datasets demonstrate that our method outperforms other state-of-the-art architectures.
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
(2023)
Review
Chemistry, Analytical
Md Golam Morshed, Tangina Sultana, Aftab Alam, Young-Koo Lee
Summary: Human action recognition systems use sensor data to accurately identify and interpret human actions. Recent developments in computer vision, particularly deep learning-based features, have significantly improved action recognition. This study provides in-depth analysis of human activity recognition, including the latest research on acquiring human action features and deep learning techniques.
Article
Computer Science, Artificial Intelligence
Andrea Bandini, Jose Zariffa
Summary: Egocentric vision applications utilizing wearable cameras have seen significant progress recently due to the availability of affordable equipment and large annotated datasets. The unique perspective offered by these cameras mounted on the head allows for studying and localizing hands, understanding their actions, and developing human-computer interfaces based on hand gestures. This survey categorizes existing approaches into localization, interpretation, and application, and also provides a list of prominent datasets with hand-based annotations.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Review
Biology
Sara Sardari, Sara Sharifzadeh, Alireza Daneshkhah, Bahareh Nakisa, Seng W. Loke, Vasile Palade, Michael J. Duncan
Summary: Performing prescribed physical exercises plays a crucial role in rehabilitation for people with physical disabilities. However, without medical experts, patients cannot assess their performance accurately. Vision-based sensors and advancements in computer vision and deep learning have enabled the development of automatic activity monitoring models to assist patients and physiotherapists. This paper provides an up-to-date review on skeleton data acquisition processes and AI-based methodologies for rehabilitation monitoring, highlighting the challenges and suggesting future research directions.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Engineering, Mechanical
Muhammad Abdul Ahad, Nadeem Iqbal, Sarvat M. Ahmad, Masroor Khan
Summary: This paper presents detailed modeling of a self-developed active magnetic bearing test rig's rotor, electromagnetic actuators, and accompanying electronics in a unified manner. The study extends the previous work by investigating the challenging problem of levitating and spinning a 2 DOF rotor on a pair of opposing electromagnetic actuators. Classical three-term PID controllers are designed to stabilize the 2-DOF open-loop unstable system in the frequency domain, and their performance is compared with modern state-space control strategies, such as Linear Quadratic Gaussian (LQG) and combined LQG-Loop Transfer Recovery (LQG\LTR) compensators.
JOURNAL OF THE BRAZILIAN SOCIETY OF MECHANICAL SCIENCES AND ENGINEERING
(2021)
Article
Computer Science, Artificial Intelligence
Md Atiqur Rahman Ahad, Masud Ahmed, Anindya Das Antar, Yasushi Makihara, Yasushi Yagi
Summary: This research proposes a method to improve action recognition performance by extracting Kinematics Posture Features (KPF) from 3D joint positions based on skeleton data. By combining Linear Joint Position Feature (LJPF) and Angular Joint Position Feature (AJPF), encoding the variation of motion in the temporal domain, and utilizing different classification models, the study demonstrates prominent performance in both statistical machine learning and deep learning-based models.
PATTERN RECOGNITION LETTERS
(2021)
Article
Computer Science, Artificial Intelligence
Anindya Das Antar, Masud Ahmed, Md Atiqur Rahman Ahad
Summary: The recognition of daily human activities in various locomotion and transportation modes has important applications for behavior modification and urban transportation planning. The study explored smartphone sensor-based datasets, demonstrated preprocessing of sensor data, and proposed an approach to improve activity recognition through statistical classifiers and deep learning methods. The proposed method achieved high accuracy rates in two benchmark datasets, showing the effectiveness of the approach compared to existing methods.
PATTERN RECOGNITION LETTERS
(2021)
Review
Computer Science, Artificial Intelligence
Sejuti Rahman, Syeda Faiza Ahmed, Omar Shahid, Musabbir Ahmed Arrafi, M. A. R. Ahad
Summary: Autism Spectrum Disorder (ASD) is a neuro-developmental disorder that limits social and cognitive abilities, with no cure currently available. Researchers are working on developing automated diagnosis systems to reduce misdiagnosis and improve diagnostic accuracy.
COGNITIVE COMPUTATION
(2022)
Editorial Material
Computer Science, Artificial Intelligence
Upal Mahbub, Md Atiqur Rahman Ahad
PATTERN RECOGNITION LETTERS
(2022)
Review
Mathematical & Computational Biology
Khondoker Murad Hossain, Md. Ariful Islam, Shahera Hossain, Anton Nijholt, Md Atiqur Rahman Ahad
Summary: In the past decade, advancements in central nervous system bioinformatics and computational innovation have led to significant developments in brain-computer interface (BCI), making it a forefront of applied science and research. BCI revitalization enables neurorehabilitation strategies for physically disabled patients and patients with brain injury. Recent availability of large, high-quality EEG datasets and the potential of deep learning techniques have shifted the popular methods in BCI applications. Deep learning shows great promise in solving complex tasks using EEG data, and researchers are actively exploring its applications in the BCI field. This review study aims to introduce recent deep learning-based approaches in BCI using EEG data, highlighting their differences, merits, drawbacks, applications, current challenges, and future research directions.
FRONTIERS IN COMPUTATIONAL NEUROSCIENCE
(2023)
Article
Multidisciplinary Sciences
Md Ahasan Atick Faisal, Farhan Fuad Abir, Mosabber Uddin Ahmed, Md Atiqur Rahman Ahad
Summary: Hand gesture recognition is a widely explored field in human-computer interaction. Recent advances in hardware and deep learning algorithms have given new momentum to this research area. This paper evaluates the effectiveness of a low-cost dataglove for hand gesture recognition using deep learning. The study demonstrates promising performance of a generalized hand gesture recognition technique, with high accuracy for both static and dynamic gestures.
SCIENTIFIC REPORTS
(2022)
Article
Computer Science, Artificial Intelligence
Allam Shehata, Yasushi Makihara, Daigo Muramatsu, Md Atiqur Rahman Ahad, Yasushi Yagi
Summary: In this paper, an uncertainty-aware estimation framework for gait relative attributes is proposed. The framework includes a two-stream network model that takes a pair of gait videos as input and outputs Gaussian distributions of gait absolute attribute scores and annotator-dependent gait relative attribute label distributions. Two uncertainty layers are introduced to estimate the gait relative attribute score distribution and the uncertainty on the gait relative attribute labels. Experiments on two gait relative attribute datasets demonstrate the effectiveness of the proposed method.
PATTERN RECOGNITION
(2023)
Article
Engineering, Multidisciplinary
Md Atiqur Rahman Ahad, Israt Jahan
Summary: In this paper, state-of-the-art segmentation algorithms applied on left ventricular segmentation on cardiac cine-MR images were explored. The global thresholding algorithm for cardiac MRI has proved to be more efficient and robust than the adaptive thresholding approach, achieving an accuracy of more than 92%. The use of entropy or histogram to characterize segmentation has a vital effect on segmentation efficiency. Clustering and region-based segmentation have produced more than 93% segmentation accuracy in the absence of boundary information in cardiac cine-MRI.
JURNAL KEJURUTERAAN
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Yeasin Arafat Pritom, Md Sohanur Rahman, Hasib Ryan Rahman, M. Ashikuzzaman Kowshik, Md Atiqur Rahman Ahad
Summary: Automated Japanese packed-meal or Bento preparation or packaging by a robot is a recent challenge in human activity recognition systems. This relies on physical gesture recognition techniques using sensor-based datasets, particularly motion capture or skeleton data. The team implements a method using handcrafted features obtained from filtered data and applies various pre-processing steps to handle dataset inconsistency. After selecting the most significant features, support vector machine, random forest, and extra tree classifiers are tested, with a maximum accuracy of 64.6% achieved using the extra tree classifier.
SENSOR- AND VIDEO-BASED ACTIVITY AND BEHAVIOR COMPUTING, ABC 2021
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
A. H. M. Nazmus Sakib, Promit Basak, Syed Doha Uddin, Shahamat Mustavi Tasin, Md Atiqur Rahman Ahad
Summary: This article discusses the use of skeleton-based motion capture systems in the game and film industry and highlights the challenges of working with smaller datasets and the lack of data for industrial activities. The authors propose an ensemble-based machine learning methodology and conduct experiments on the MoCap data from the Bento Packaging Activity Recognition Challenge 2021, achieving impressive accuracy.
SENSOR- AND VIDEO-BASED ACTIVITY AND BEHAVIOR COMPUTING, ABC 2021
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Faizul Rakib Sayem, Md Mamun Sheikh, Md Atiqur Rahman Ahad
Summary: Due to the advancements of low-cost sensors, human action recognition has become an important research topic and gained attention in human-robot collaboration. This work presents a machine learning paradigm to recognize different Bento packaging activities in real-time. It is a challenging task without lower-body marker information.
SENSOR- AND VIDEO-BASED ACTIVITY AND BEHAVIOR COMPUTING, ABC 2021
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Adrita Anwar, Malisha Islam Tapotee, Purnata Saha, Md Atiqur Rahman Ahad
Summary: This paper discusses the challenge of food packaging activity recognition in industrial automation systems and proposes a classical machine learning approach to address this issue. By preprocessing the dataset, extracting features, and training different models, the authors achieved a certain level of accuracy.
SENSOR- AND VIDEO-BASED ACTIVITY AND BEHAVIOR COMPUTING, ABC 2021
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Shiqi Yu, Yongzhen Huang, Liang Wang, Yasushi Makihara, Edel B. Garcia Reyes, Feng Zheng, Md Atiqur Rahman Ahad, Beibei Lin, Yuchao Yang, Haijun Xiong, Binyuan Huang, Yuxuan Zhang
Summary: The Competition on Human Identification at a Distance 2021 aims to promote research in human identification at a distance and provide a benchmark for evaluating different methods. The results from the top teams demonstrate state-of-the-art performance in gait recognition, with improvements seen compared to the previous year. Useful conclusions were drawn from comparisons and analysis, paving the way for further advancements in future competitions.
2021 INTERNATIONAL JOINT CONFERENCE ON BIOMETRICS (IJCB 2021)
(2021)
Proceedings Paper
Computer Science, Information Systems
Purnata Saha, Malisha Islam Tapotee, Md Atiqur Rahman Ahad
Summary: Social interaction plays a crucial role in overcoming Autism Spectrum Disorder (ASD), and robots can serve as interaction partners during therapy sessions to enhance treatment effectiveness. The DREAM dataset provides valuable data for evaluating Robot Enhanced Therapy and improving social communication skills of children with ASD. Through statistical and gaze-based analysis, ensemble methods like Random Forest and XGBoost classifiers yield satisfactory results in assessing the feasibility of using robots as substitutes for humans in therapy sessions for children with ASD.
13TH INTERNATIONAL CONFERENCE ON MOBILE COMPUTING AND UBIQUITOUS NETWORK (ICMU2021)
(2021)
