Waheed Ali Laghari
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Waheed Ali Laghari has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Aslina Baharum
Taylor’s University
1 shared publication - Audrey Huong
Universiti Tun Hussein Onn Malaysia
1 shared publication - Chang Choon Chew
Department of Electronic Engineering, Faculty of Electrical and Electronic Engineering, Universiti Tun Hussein Onn Malaysia, 86400 Parit Raja, Batu Pahat, Johor, Malaysia
1 shared publication - Ganesh Kumar
Department of Computing, Universiti Teknologi PETRONAS, Seri Iskandar 32610, Malaysia
1 shared publication - Javid Iqbal
UCSI University
1 shared publication - Kim Gaik Tay
Universiti Tun Hussein Onn Malaysia
1 shared publication
Latest Publications
- Open Access
- 23 January 2026
- Published by Association for Computing Machinery (ACM)
- p. 17-22
Interpretable MultiModal AI for Mental Health Detection via Facial and Vocal Emotions
Mental disorders, including depression and anxiety, constitute a major global health concern. Timely and precise identification is essential for effective treatment and suicide prevention. Human emotions, conveyed through facial expressions and vocalizations, provide an essential means for identification. However, current Artificial Intelligence (AI) frameworks in this domain face two major limitations: dataset imbalance and resultant low diagnostic accuracy. Moreover, previous work considered only one modality. This paper presents a comprehensive, multi-modal approach using facial and vocal emotions to improve mental health identification. A Generative Adversarial Network (GAN) is employed to mitigate data scarcity and imbalance by generating a variety of high-quality data samples, utilizing the public facial (FER2013) and vocal (CREMA-D) emotion datasets for training. Following data balancing, the YOLOv8 model is employed to generate a dataset categorized into specific mental disorder classifications (e.g., depression, anxiety, another disorder, and no disorder). A comparative analysis of various Convolutional Neural Network (CNN) architectures is performed to identify the most efficient classification model based on accuracy. The best model is subsequently analyzed using GRAD-CAM (Gradient-weighted Class Activation Mapping) to enhance the interpretability of its internal decision-making process. For real-world implementation and validation, a secure, full-stack web application is developed. This application uses streamlit for front-end, flask server for hosting CNN model (back-end), HTTP for communication and JSON web token for data security. The framework is tested in real-time using live facial and voice data collected from Sunway University students. This research establishes a high-accuracy framework for mental health recognition from multi-modal affective inputs by utilizing Generative Adversarial Networks (GANs) for effective data augmentation and integrating sophisticated Convolutional Neural Network (CNN) and YOLOv8 methodologies. The developed application facilitates accurate and expeditious detection, thereby enabling early intervention and the mitigation of adverse mental health outcomes.
- 23 January 2026
- Published by Association for Computing Machinery (ACM)
- p. 17-22
Open AccessInterpretable MultiModal AI for Mental Health Detection via Facial and Vocal Emotions
Mental disorders, including depression and anxiety, constitute a major global health concern. Timely and precise identification is essential for effective treatment and suicide prevention. Human emotions, conveyed through facial expressions and vocalizations, provide an essential means for identification. However, current Artificial Intelligence (AI) frameworks in this domain face two major limitations: dataset imbalance and resultant low diagnostic accuracy. Moreover, previous work considered only one modality. This paper presents a comprehensive, multi-modal approach using facial and vocal emotions to improve mental health identification. A Generative Adversarial Network (GAN) is employed to mitigate data scarcity and imbalance by generating a variety of high-quality data samples, utilizing the public facial (FER2013) and vocal (CREMA-D) emotion datasets for training. Following data balancing, the YOLOv8 model is employed to generate a dataset categorized into specific mental disorder classifications (e.g., depression, anxiety, another disorder, and no disorder). A comparative analysis of various Convolutional Neural Network (CNN) architectures is performed to identify the most efficient classification model based on accuracy. The best model is subsequently analyzed using GRAD-CAM (Gradient-weighted Class Activation Mapping) to enhance the interpretability of its internal decision-making process. For real-world implementation and validation, a secure, full-stack web application is developed. This application uses streamlit for front-end, flask server for hosting CNN model (back-end), HTTP for communication and JSON web token for data security. The framework is tested in real-time using live facial and voice data collected from Sunway University students. This research establishes a high-accuracy framework for mental health recognition from multi-modal affective inputs by utilizing Generative Adversarial Networks (GANs) for effective data augmentation and integrating sophisticated Convolutional Neural Network (CNN) and YOLOv8 methodologies. The developed application facilitates accurate and expeditious detection, thereby enabling early intervention and the mitigation of adverse mental health outcomes.
- Open Access
- 30 April 2023
- Published by The Korean Society of Medical Informatics in Healthcare Informatics Research
- Vol. 29 (2) , 152-160
Dorsal Hand Vein Pattern Recognition: A Comparison between Manual and Automatic Segmentation Methods
Objectives: Various techniques for dorsal hand vein (DHV) pattern extraction have been introduced using small datasets with poor and inconsistent segmentation. This work compared manual segmentation with our proposed hybrid automatic segmentation method (HHM) for this classification problem.Methods: Manual segmentation involved selecting a region-of-interest (ROI) in images from the Bosphorus dataset to generate ground truth data. The HHM combined histogram equalization and morphological and thresholding-based algorithms to localize veins from hand images. The data were divided into training, validation, and testing sets with an 8:1:1 ratio before training AlexNet. We considered three image augmentation strategies to enlarge our training sets. The best training hyperparameters were found using the manually segmented dataset.Results: We obtained a good test accuracy (91.5%) using the model trained with manually segmented images. The HHM method showed slightly inferior performance (76.5%). Considerable improvement was observed in the test accuracy of the model trained with the inclusion of automatically segmented and augmented images (84%), with low false acceptance and false rejection rates (0.00035% and 0.095%, respectively). A comparison with past studies further demonstrated the competitiveness of our technique.Conclusions: Our technique can be feasible for extracting the ROI in DHV images. This strategy provides higher consistency and greater efficiency than the manual approach.
- 30 April 2023
- Published by The Korean Society of Medical Informatics in Healthcare Informatics Research
- Vol. 29 (2) , 152-160
Open AccessDorsal Hand Vein Pattern Recognition: A Comparison between Manual and Automatic Segmentation Methods
Objectives: Various techniques for dorsal hand vein (DHV) pattern extraction have been introduced using small datasets with poor and inconsistent segmentation. This work compared manual segmentation with our proposed hybrid automatic segmentation method (HHM) for this classification problem.Methods: Manual segmentation involved selecting a region-of-interest (ROI) in images from the Bosphorus dataset to generate ground truth data. The HHM combined histogram equalization and morphological and thresholding-based algorithms to localize veins from hand images. The data were divided into training, validation, and testing sets with an 8:1:1 ratio before training AlexNet. We considered three image augmentation strategies to enlarge our training sets. The best training hyperparameters were found using the manually segmented dataset.Results: We obtained a good test accuracy (91.5%) using the model trained with manually segmented images. The HHM method showed slightly inferior performance (76.5%). Considerable improvement was observed in the test accuracy of the model trained with the inclusion of automatically segmented and augmented images (84%), with low false acceptance and false rejection rates (0.00035% and 0.095%, respectively). A comparison with past studies further demonstrated the competitiveness of our technique.Conclusions: Our technique can be feasible for extracting the ROI in DHV images. This strategy provides higher consistency and greater efficiency than the manual approach.