https://ir.kdu.ac.lk/handle/345/7631 IJRC Wed, 08 Apr 2026 13:41:34 GMT 2026-04-08T13:41:34Z https://ir.kdu.ac.lk/handle/345/8907 Speech Emotion Recognition with Hybrid CNN- LSTM and Transformers Models: Evaluating the Hybrid Model Using Grad-CAM Kumari, HMLS; Kumari, HMNS; Nawarathne, UMMPK Emotional recognition and classification using artificial intelligence (AI) techniques play a crucial role in human-computer interaction (HCI). It enables the prediction of human emotions from audio signals with broad applications in psychology, medicine, education, entertainment, etc. This research focused on speech-emotion recognition (SER) by employing classification methods and transformer models using the Toronto Emotional Speech Set (TESS). Initially, acoustic features were extracted using different feature extraction techniques, including chroma, Mel-scaled spectrogram, contrast features, and Mel Frequency Cepstral Coefficients (MFCCs) from the audio dataset. Then, this study employed a Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and a hybrid CNN-LSTM model to classify emotions. To compare the performance of these models, classical image transformer models such as ViT (Visual Image Transformer) and BEiT (Bidirectional Encoder Representation of Images) were employed on the Mel-spectograms derived from the same dataset. Evaluation metrics such as accuracy, precision, recall, and F1-score were calculated for each of these models to ensure a comprehensive performance comparison. According to the results, the hybrid model performed better than other models by achieving an accuracy of 99.01%, while the CNN, LSTM, ViT, and BEiT models demonstrated accuracies of 95.37%, 98.57%, 98%, and 98.3%, respectively. To interpret the output of this hybrid model and to provide visual explanations of its predictions, the Grad-CAM (Gradient-weighted Class Activation Mappings) was obtained. This technique reduced the black-box character of deep models, making them more reliable to use in clinical and other delicate contexts. In conclusion, the hybrid CNN-LSTM model showed strong performance in audio-based emotion classification. Mon, 01 Jul 2024 00:00:00 GMT https://ir.kdu.ac.lk/handle/345/8907 2024-07-01T00:00:00Z https://ir.kdu.ac.lk/handle/345/8906 AI-Driven Disaster Prediction and Early Warning Systems: A Systematic Literature Review Luxshi, K Numerous advancements in artificial intelligence drive better accuracy and improved performance of disaster prediction and early warning systems for hazards. This review gathers and integrates current findings on AI management of disasters through machine learning, deep learning, and data analytics techniques that address natural disasters and human-made emergencies. The paper examines how artificial intelligence contributes to earthquake forecasting while also providing information on flood forecasting, wildfire detection systems, and other hazard assessment needs. This research explores how AI technology connects with the Internet of Things (IoT) and remote sensing systems for real-time disaster monitoring. The discussion includes detailed assessments of key barriers, such as data quality issues, system limitations, and ethical concerns. Future researchers can use this study to identify ways to enhance AI-based disaster resilience strategies. Mon, 01 Jul 2024 00:00:00 GMT https://ir.kdu.ac.lk/handle/345/8906 2024-07-01T00:00:00Z https://ir.kdu.ac.lk/handle/345/8905 Systematic Review on AI in Gender Bias Detection and Mitigation in Education and Workplaces Deckker, D; Sumanasekara, s Gender bias in artificial intelligence (AI) systems, particularly within education and workplace settings, poses serious ethical and operational concerns. These biases often stem from historically skewed datasets and flawed algorithmic logic, which can lead to the reinforcement of existing inequalities and the systematic exclusion of underrepresented groups, especially women. This systematic review analyses peer-reviewed literature from 2010 to 2024, sourced from IEEE Xplore, Google Scholar, PubMed, and SpringerLink. Using targeted keywords such as AI gender bias, algorithmic fairness, and bias mitigation, the review assesses empirical and theoretical studies that examine the causes of gender bias, its manifestations in AI-driven decision-making systems, and proposed strategies for detection and mitigation. Findings reveal that biased training data, algorithm design flaws, and unacknowledged developer assumptions are primary sources of gender discrimination in AI systems. In education, these systems affect grading accuracy and learning outcomes; in workplaces, they influence hiring, evaluations, and promotions. Mitigation approaches can be categorized into three main categories: data-centric (e.g., data augmentation and data balancing), algorithm-centric (e.g., fairness-aware learning and adversarial training), and post-processing techniques (e.g., output calibration). However, each approach faces implementation challenges, including trade-offs between fairness and accuracy, lack of transparency, and the absence of intersectional bias detection. The review concludes that gender fairness in AI requires integrated strategies that combine technical solutions with ethical governance. Ethical AI deployment must be grounded in inclusive data practices, transparent protocols, and interdisciplinary collaboration. Policymakers and organizations must strengthen accountability frameworks, such as the EU AI Act and the U.S. AI Bill of Rights, to ensure that AI technologies support equitable outcomes in education and employment. Mon, 01 Jul 2024 00:00:00 GMT https://ir.kdu.ac.lk/handle/345/8905 2024-07-01T00:00:00Z https://ir.kdu.ac.lk/handle/345/8904 Augmented Reality (AR) and Virtual Reality (VR) in Education: A Comprehensive Review Konara, KMHL; Dilani, GK; Peiris, TMHC; Dileka, RMR; Rathnayaka, TP; Kithulwatta, WMCJT; Jayathilake, RMD; Wijewardana, YNS; Somarathna, HMCC; Rathnayake, RMKT Now the world is fully moving towards a digitalized environment in all kinds of disciplines in education, agriculture, the banking industry, transportation, healthcare, etc., with the most prominent and trending topics. Augmented Reality (AR) and Virtual Reality (VR) technologies are contemporary tools that are gradually changing learning processes among plenty of newly arrived technologies. These tools enable educators, including lecturers, teachers, tutors, and instructors, to address their audiences creatively with ideas that are useful for learning purposes. This research study focuses on the effectiveness of AR and VR in classroom learning and discusses the tools' effects on access, retention, and collaborative learning. The study was nourished with thirty scholarly articles for the core review process and supplementary articles for designing the review process from reputed academic research databases. The research study observed on main educational aspects of the AR and VR concept, including virtual classrooms, AR labs, corporate training, facilities for special needs students, collaborative work, etc. Furthermore, the research study discusses cost-related issues, technical issues, ethical issues, and new directions, which entail combining Artificial Intelligence (AI) and increasing global availability. Therefore, while highlighting what has not yet been accomplished by AR and VR, this work focuses on the potential of true transformation of learning and education processes as tools for meaningful, effective, and accessible education. Finally, this research study obtained knowledge summarization and synthesis on modern AR and VR technologies in the education sector. Mon, 01 Jul 2024 00:00:00 GMT https://ir.kdu.ac.lk/handle/345/8904 2024-07-01T00:00:00Z