Sunusi Bala Abdullahi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Sunusi Bala Abdullahi has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- Kosin Chamnongthai
Department of Electronic and Telecommunication Engineering, King Mongkuts University of Technology Thonburi, Bangkok, Thailand
6 shared publications Zakariyya Abdullahi Bature
4 shared publications- Amjad Rehman
CCIS Prince Sultan University, Riyadh, Saudi Arabia
3 shared publications Kosin Chamnongthai
3 shared publications- Abdulkarim Hassan Ibrahim
Emirates Aviation University
2 shared publications - Auwal Bala Abubakar
Department of Art and Science, George Mason University, Songdomunhwa-ro 119-4, Yeonsu-gu, Incheon 21985, Republic of Korea
2 shared publications
Latest Publications
- 1 September 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 153-158
End-to-End Class-Balanced Multilayer Perceptron for Network Intrusion Detection
- 1 September 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 153-158
End-to-End Class-Balanced Multilayer Perceptron for Network Intrusion Detection
- Open Access
- 20 August 2026
- Published by Elsevier in SSRN Electronic Journal
Multi-Resolution Temporal Keyframe Modeling with a Non-Causal TCN for Eye-Writing Recognition
Eye-writing recognition can support communication for individuals with severe motor impairments; however, variations in gaze trajectories across users, trials, sensing modalities, and writing styles remain challenging for reliable recognition. This study proposes a subject-invariant eye-writing recognition framework based on a multi-resolution temporal keyframe representation and a multi-kernel non-causal temporal convolutional network (TCN). The framework first constructs multi-resolution representations of two-dimensional gaze trajectories to retain coarse trajectory structure and fine temporal eye-movement dynamics. Multi-kernel non-causal temporal convolutions are then used to model complementary temporal dependencies, while parallel global average and maximum pooling aggregate sequence-level information for classification. The proposed framework was evaluated using a Tobii eye tracker, a standard webcam, and electrooculography (EOG). Leave-one-subject-out (LOSO) evaluation achieved mean Accuracy values of 96.79% on the Tobii dataset and 98.63% on the camera dataset. The results demonstrate that the proposed framework provides robust recognition across optical eye-tracker, webcam-based, and EOG-based eye-writing signals.
- 20 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMulti-Resolution Temporal Keyframe Modeling with a Non-Causal TCN for Eye-Writing Recognition
Eye-writing recognition can support communication for individuals with severe motor impairments; however, variations in gaze trajectories across users, trials, sensing modalities, and writing styles remain challenging for reliable recognition. This study proposes a subject-invariant eye-writing recognition framework based on a multi-resolution temporal keyframe representation and a multi-kernel non-causal temporal convolutional network (TCN). The framework first constructs multi-resolution representations of two-dimensional gaze trajectories to retain coarse trajectory structure and fine temporal eye-movement dynamics. Multi-kernel non-causal temporal convolutions are then used to model complementary temporal dependencies, while parallel global average and maximum pooling aggregate sequence-level information for classification. The proposed framework was evaluated using a Tobii eye tracker, a standard webcam, and electrooculography (EOG). Leave-one-subject-out (LOSO) evaluation achieved mean Accuracy values of 96.79% on the Tobii dataset and 98.63% on the camera dataset. The results demonstrate that the proposed framework provides robust recognition across optical eye-tracker, webcam-based, and EOG-based eye-writing signals.
- Open Access
- 19 August 2026
- Published by Elsevier in SSRN Electronic Journal
Multi-Resolution Temporal Keyframe Modeling with a Non-Causal TCN for Eye-Writing Recognition
Eye-writing recognition can support communication for individuals with severe motor impairments; however, variations in gaze trajectories across users, trials, sensing modalities, and writing styles remain challenging for reliable recognition. This study proposes a subject-invariant eye-writing recognition framework based on a multi-resolution temporal keyframe representation and a multi-kernel non-causal temporal convolutional network (TCN). The framework first constructs multi-resolution representations of two-dimensional gaze trajectories to retain coarse trajectory structure and fine temporal eye-movement dynamics. Multi-kernel non-causal temporal convolutions are then used to model complementary temporal dependencies, while parallel global average and maximum pooling aggregate sequence-level information for classification. The proposed framework was evaluated using a Tobii eye tracker, a standard webcam, and electrooculography (EOG). Leave-one-subject-out (LOSO) evaluation achieved mean Accuracy values of 96.79% on the Tobii dataset and 98.63% on the camera dataset. The results demonstrate that the proposed framework provides robust recognition across optical eye-tracker, webcam-based, and EOG-based eye-writing signals.
- 19 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMulti-Resolution Temporal Keyframe Modeling with a Non-Causal TCN for Eye-Writing Recognition
Eye-writing recognition can support communication for individuals with severe motor impairments; however, variations in gaze trajectories across users, trials, sensing modalities, and writing styles remain challenging for reliable recognition. This study proposes a subject-invariant eye-writing recognition framework based on a multi-resolution temporal keyframe representation and a multi-kernel non-causal temporal convolutional network (TCN). The framework first constructs multi-resolution representations of two-dimensional gaze trajectories to retain coarse trajectory structure and fine temporal eye-movement dynamics. Multi-kernel non-causal temporal convolutions are then used to model complementary temporal dependencies, while parallel global average and maximum pooling aggregate sequence-level information for classification. The proposed framework was evaluated using a Tobii eye tracker, a standard webcam, and electrooculography (EOG). Leave-one-subject-out (LOSO) evaluation achieved mean Accuracy values of 96.79% on the Tobii dataset and 98.63% on the camera dataset. The results demonstrate that the proposed framework provides robust recognition across optical eye-tracker, webcam-based, and EOG-based eye-writing signals.
- 17 August 2026
- Published by Elsevier in Computers & Industrial Engineering
- Vol. 221
A blockchain-verified explainable AI pipeline for multi-scale agri-food carbon accounting
- 17 August 2026
- Published by Elsevier in Computers & Industrial Engineering
- Vol. 221
A blockchain-verified explainable AI pipeline for multi-scale agri-food carbon accounting
- 22 July 2026
- Published by Elsevier in International Journal of Medical Informatics
- Vol. 220, 106632
Age-stratified machine learning using de-identified clinical and transcriptomic data for pediatric appendicitis
- 22 July 2026
- Published by Elsevier in International Journal of Medical Informatics
- Vol. 220, 106632
Age-stratified machine learning using de-identified clinical and transcriptomic data for pediatric appendicitis
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 129255
Corrections to “Enhanced Weak Spatial Modeling Through CNN-Based Deep Sign Language Skeletal Feature Transformation”
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 129255
Open AccessCorrections to “Enhanced Weak Spatial Modeling Through CNN-Based Deep Sign Language Skeletal Feature Transformation”
- Open Access
- 28 December 2025
- Published by Research Direct Publisher in International Journal of Theoretical & Applied Computational Intelligence
- Vol. 2025, 305-322
Human Action Recognition: A Comprehensive Survey of Multimodal Advances, Challenges, and Emerging Directions
Human Action Recognition (HAR) has emerged as a pivotal domain within computer vision and machine learning, driven by its transformative potential across surveillance, healthcare, human-computer interaction, and sports analytics. Despite notable advances, a persistent gap remains between benchmark-driven performance and real-world applicability, particularly in scenarios demanding cross-subject generalization, fine-grained understanding, computational efficiency, and scalability. This survey presents a systematic and critical review of HAR research published between 2022 and 2025, encompassing 30 peer-reviewed articles published in IEEE venues and retrieved primarily through the IEEE Xplore digital library. We trace the progression from unimodal frameworks to multimodal fusion architectures, highlighting innovations across skeleton-based, sensor-based, and vision-based modalities. Key architectural trends include transformer-based models, graph neural networks, and self-supervised learning, alongside domain-specific adaptations in healthcare and sports. We further examine methodological shifts toward lightweight and generalizable systems. By synthesizing these developments, this work offers a structured roadmap for future research, emphasizing the need for robust evaluation protocols, ethical considerations, and deployment-ready HAR solutions.
- 28 December 2025
- Published by Research Direct Publisher in International Journal of Theoretical & Applied Computational Intelligence
- Vol. 2025, 305-322
Open AccessHuman Action Recognition: A Comprehensive Survey of Multimodal Advances, Challenges, and Emerging Directions
Human Action Recognition (HAR) has emerged as a pivotal domain within computer vision and machine learning, driven by its transformative potential across surveillance, healthcare, human-computer interaction, and sports analytics. Despite notable advances, a persistent gap remains between benchmark-driven performance and real-world applicability, particularly in scenarios demanding cross-subject generalization, fine-grained understanding, computational efficiency, and scalability. This survey presents a systematic and critical review of HAR research published between 2022 and 2025, encompassing 30 peer-reviewed articles published in IEEE venues and retrieved primarily through the IEEE Xplore digital library. We trace the progression from unimodal frameworks to multimodal fusion architectures, highlighting innovations across skeleton-based, sensor-based, and vision-based modalities. Key architectural trends include transformer-based models, graph neural networks, and self-supervised learning, alongside domain-specific adaptations in healthcare and sports. We further examine methodological shifts toward lightweight and generalizable systems. By synthesizing these developments, this work offers a structured roadmap for future research, emphasizing the need for robust evaluation protocols, ethical considerations, and deployment-ready HAR solutions.
- 25 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
MetaFSelect: No-Code Metaheuristic Feature Selection for Machine Learning
Feature selection (FS) enhances model performance and interpretability but remains challenging due to algorithmic and hyperparameter sensitivity. We introduce MetaFSelect, a stochastic, no-code framework integrating MEALPY, MAFESE, and machine learning models for reproducible metaheuristicdriven FS. Evaluated on six benchmark datasets, MetaFSelect achieved 93.5% feature reduction with only $1-21 {\%}$ accuracy deviation, statistically insignificant in four cases ($p>0.05$), while reducing setup time to under five minutes. The framework unifies automation and statistical rigor for transparent, scalable FS.- 25 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
MetaFSelect: No-Code Metaheuristic Feature Selection for Machine Learning
Feature selection (FS) enhances model performance and interpretability but remains challenging due to algorithmic and hyperparameter sensitivity. We introduce MetaFSelect, a stochastic, no-code framework integrating MEALPY, MAFESE, and machine learning models for reproducible metaheuristicdriven FS. Evaluated on six benchmark datasets, MetaFSelect achieved 93.5% feature reduction with only $1-21 {\%}$ accuracy deviation, statistically insignificant in four cases ($p>0.05$), while reducing setup time to under five minutes. The framework unifies automation and statistical rigor for transparent, scalable FS.- 5 November 2025
- Published by SAGE Publications in Intelligent Data Analysis: An International Journal
- Vol. 30 (3) , 745-767
Minimizing redundancy in hand dynamic features for enhanced sign language recognition
This paper presents a wavelet-based framework for 3-D sign language recognition that outperforms conventional ZCA approaches by preserving critical spatiotemporal relationships in hand kinematics. We formulate hand dynamics as a wavelet exponent energy problem using Daubechies transforms, where joint coordinates, mother coefficients, and window sizes are jointly optimized, avoiding PCA’s disruptive axis rotation while maintaining interpretable feature orientations. The proposed method combines this wavelet energy representation with an ensemble deep learning feature selection layer using binary masking ( λ 1 = 0.10 , λ 2 = 0.01 ) and regularization to eliminate redundancy, achieving 94.09% accuracy (vs. 92.75% for PCC+ZCA (SVD)) with 4.3 × faster inference on Turkish SL datasets. Statistical validation (ANOVA, p < 0.001 ) confirms significant improvements over ZCA-derived features ( t = 24.15 ), with weight analysis revealing proximal arm kinematics as dominant recognition factors. The framework’s computational efficiency (0.12ms inference) and preservation of kinematic relationships demonstrate its superiority for real-life SLR applications compared to both PCC and ZCA approaches.
- 5 November 2025
- Published by SAGE Publications in Intelligent Data Analysis: An International Journal
- Vol. 30 (3) , 745-767
Minimizing redundancy in hand dynamic features for enhanced sign language recognition
This paper presents a wavelet-based framework for 3-D sign language recognition that outperforms conventional ZCA approaches by preserving critical spatiotemporal relationships in hand kinematics. We formulate hand dynamics as a wavelet exponent energy problem using Daubechies transforms, where joint coordinates, mother coefficients, and window sizes are jointly optimized, avoiding PCA’s disruptive axis rotation while maintaining interpretable feature orientations. The proposed method combines this wavelet energy representation with an ensemble deep learning feature selection layer using binary masking ( λ 1 = 0.10 , λ 2 = 0.01 ) and regularization to eliminate redundancy, achieving 94.09% accuracy (vs. 92.75% for PCC+ZCA (SVD)) with 4.3 × faster inference on Turkish SL datasets. Statistical validation (ANOVA, p < 0.001 ) confirms significant improvements over ZCA-derived features ( t = 24.15 ), with weight analysis revealing proximal arm kinematics as dominant recognition factors. The framework’s computational efficiency (0.12ms inference) and preservation of kinematic relationships demonstrate its superiority for real-life SLR applications compared to both PCC and ZCA approaches.
- 12 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (22) , 37630-37645
ℱsign-Net: Depth Sensor Aggregated Frame-Based Fourier Network for Sign Word Recognition
Hand-tracking is a challenging problem during hand gesture recognition due to abnormal hand patterns across depth signs and errors between normal pixels and backgrounds. In this article, we propose F sign-Net: Fourier pixel-wise approach based on the Fourier Convolution Neural Network (FCNN) and Time-distributed-based Bi-directional Long-short-term Memory (BiLSTM). FCNNs have been widely researched, reaching state-of-the-art performance on spatial recognition tasks. However, it is still difficult for the Fourier model to learn the temporal patterns due to the chaotic nature of the hand motion data. F sign-Net aggregates time-based information from spatial and temporal modules to a given Fourier convolution in three stages: (1) Each depth frame is regarded as a window, and it is selected so that the aggregated sums of the pixels across the hand joints of the selected window are aligned, (2) a truncated pooling is applied that summarizes the generated featured map of the Fourier convolution to avoid over-fitting, and (3) the long-term temporal dependencies among pixels for the Fourier convolution are captured using the shared Time-based BiLSTM layers. This allows the proposed model to learn hand patterns that are temporally oriented. Finally, the proposed F net-Sign is evaluated on depth sign language public data sets and demonstrates state-of-the-art performance. Simulation results proved that improved Fourier features are good features for the proposed hand-tracking approach.
- 12 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (22) , 37630-37645
ℱsign-Net: Depth Sensor Aggregated Frame-Based Fourier Network for Sign Word Recognition
Hand-tracking is a challenging problem during hand gesture recognition due to abnormal hand patterns across depth signs and errors between normal pixels and backgrounds. In this article, we propose F sign-Net: Fourier pixel-wise approach based on the Fourier Convolution Neural Network (FCNN) and Time-distributed-based Bi-directional Long-short-term Memory (BiLSTM). FCNNs have been widely researched, reaching state-of-the-art performance on spatial recognition tasks. However, it is still difficult for the Fourier model to learn the temporal patterns due to the chaotic nature of the hand motion data. F sign-Net aggregates time-based information from spatial and temporal modules to a given Fourier convolution in three stages: (1) Each depth frame is regarded as a window, and it is selected so that the aggregated sums of the pixels across the hand joints of the selected window are aligned, (2) a truncated pooling is applied that summarizes the generated featured map of the Fourier convolution to avoid over-fitting, and (3) the long-term temporal dependencies among pixels for the Fourier convolution are captured using the shared Time-based BiLSTM layers. This allows the proposed model to learn hand patterns that are temporally oriented. Finally, the proposed F net-Sign is evaluated on depth sign language public data sets and demonstrates state-of-the-art performance. Simulation results proved that improved Fourier features are good features for the proposed hand-tracking approach.