Facial expression recognition of online learners from real-time videos using a novel deep learning model

Jagadeesh, Madhu; Baranidharan, B.

doi:10.1007/s00530-022-00957-z

Cited by 20 publications

(4 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The performance metrics used for evaluating the IQHPSO algorithm typically include measures of solution quality, convergence speed, and robustness. These metrics can be used to compare the performance of the algorithm with other optimization algorithms, as well as to evaluate its performance on different problem instances [38][39][40][41][42].…”

Section: Details Of Performance Metrics Used For Evaluating the Algor...mentioning

confidence: 99%

Improved Quantum-inspired Hybrid Particle Swarm Optimization Based Resource Allocation for Quantum- Enabled IoT Devices in Ad Hoc Networks

2024

Preprint

View full text Add to dashboard Cite

Ad hoc networks of quantum-enabled IoT devices require efficient resource allocation techniques to optimize energy consumption and data transmission rates. This paper proposes an enhanced version of quantum-inspired hybrid particle swarm optimization called the Improved quantum-inspired hybrid particle swarm optimization (IQHPSO) algorithm for resource allocation in ad hoc networks of quantum-enabled IoT devices. The enhanced QHPSO algorithm incorporates a novel innovation technique based on adaptive parameter tuning and a global best selection strategy that further improves the optimization performance of the QHPSO approach. We evaluate the proposed approach through simulations and compare its performance with the original QHPSO algorithm and other state-of-the-art optimization techniques regarding objective functions such as energy consumption, data transmission rate, and network lifetime. Our results show that the IQHPSO algorithm outperforms the original QHPSO algorithm by 10-15% in energy consumption, 5-10% in data transmission rate, and 3-7% in network lifetime. Furthermore, our results demonstrate that the IQHPSO algorithm outperforms other state-of-the-art optimization techniques regarding these objective functions. We also discuss the practical implementation of the IQHPSO algorithm in a real-world ad hoc network of quantum-enabled IoT devices. This research contributes to the ongoing development of quantum-inspired optimization techniques for wireless networking applications. Furthermore, it demonstrates the benefits of incorporating innovative approaches, such as adaptive parameter tuning and global best selection strategy, to improve the performance of existing optimization algorithms.

show abstract

Section: Details Of Performance Metrics Used For Evaluating the Algor...mentioning

confidence: 99%

Improved Quantum-inspired Hybrid Particle Swarm Optimization Based Resource Allocation for Quantum- Enabled IoT Devices in Ad Hoc Networks

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…This article mainly started from the usage of the video dataset, and the authors also applied some classic network structures of CNN to analyze the practicality of the dataset. In 2022, Jagadesh and Baranidharan [37] introduced their own real-time online learning videos dataset.…”

Section: Related Workmentioning

confidence: 99%

An Optimized CNN Model for Engagement Recognition in an E-Learning Environment

Jiang

Zhu

2022

Applied Sciences

View full text Add to dashboard Cite

In the wake of the restrictions imposed on social interactions due to the COVID-19 pandemic, traditional classroom education was replaced by distance education in many universities. Under the changed circumstances, students are required to learn more independently. The challenge for teachers has been to duly ascertain students’ learning efficiency and engagement during online lectures. This paper proposes an optimized lightweight convolutional neural network (CNN) model for engagement recognition within a distance-learning setup through facial expressions. The ShuffleNet v2 architecture was selected, as this model can easily adapt to mobile platforms and deliver outstanding performance compared to other lightweight models. The proposed model was trained, tested, evaluated and compared with other CNN models. The results of our experiment showed that an optimized model based on the ShuffleNet v2 architecture with a change of activation function and the introduction of an attention mechanism provides the best performance concerning engagement recognition. Further, our proposed model outperforms many existing works in engagement recognition on the same database. Finally, this model is suitable for student engagement recognition for distance learning on mobile platforms.

show abstract

“…Head posture, eye gaze, eye-opening and closing states, and the most used facial movement units are used as visual features and others object detectors [23]. Jagadeesh and Baranidharan [24] develop a deep learning-oriented facial expression recognition (FER) of online learners from real-time videos to determine which facial physical behaviors are associated with emotional states and then to determine how these emotional states are related to student understanding. Li [25], present a visual analytics approach to facilitate the proctoring of online exams by analyzing the exam videos of each student recorded by a webcam and mouse movement data of the student.…”

Section: Related Workmentioning

confidence: 99%

The Attention of Students in Online Education

Barba-Guamán

Valdiviezo-Díaz

2022

Int. J. Emerg. Technol. Learn.

View full text Add to dashboard Cite

The level of attention of students who receive classes through videoconferencing platforms is troubling. Different throughout an entire lecture, the instructor can perceive the participants' behavior, while in an online class, it is difficult to determine if they are attentive to the instructions given. An innovative method that helps solve this problem is the use of computer vision algorithms, with methods such as face detection, facial landmarks, face recognition, and head pose detection based on deep learning networks. In this paper, a neural network was trained using facial landmarks to estimate head position and thus attention levels. The model was applied in five classes using different videoconferencing platforms in the Intelligent Systems subject of a private university in Ecuador. Some limitations, such as lighting and video quality, affected the level of accuracy. The number of registered participants was 12, of which between 30% and 80% attended. The maximum level of attention detected was 91.9%, while the minimum level was 86.6%. This case study proves that the head position detection function employed by many videoconferencing platforms is a useful parameter that aids the instructor in this type of context.

show abstract

Facial expression recognition of online learners from real-time videos using a novel deep learning model

Cited by 20 publications

References 36 publications

Improved Quantum-inspired Hybrid Particle Swarm Optimization Based Resource Allocation for Quantum- Enabled IoT Devices in Ad Hoc Networks

Improved Quantum-inspired Hybrid Particle Swarm Optimization Based Resource Allocation for Quantum- Enabled IoT Devices in Ad Hoc Networks

An Optimized CNN Model for Engagement Recognition in an E-Learning Environment

The Attention of Students in Online Education

Contact Info

Product

Resources

About