ICSEMT May 2025 Proceeding

Emotion Recognition In Text And Image Using LSTM And CNN Architecture

Authors: Dr. Gaurav Aggarwal, Narinder Yadav

Abstract: Emotion recognition in texts and images is an area under rapid development at the moment, as advancements in deep learning, natural language processing, and computer vision seem to facilitate human-computer interaction, mental health monitoring, and many more aspects of sentiment analysis. The study seeks to propose a hybrid model integrating LSTM networks for text and a CNN-based architecture for image data in order to handle imbalanced datasets, sarcasm detection, and feature extraction from graphical content. The multimodal fusion will help the proposed framework capture the nuanced emotional signals of both modalities, providing a more holistic understanding of human emotions. This is evaluated on publicly available datasets that show improvements in terms of accuracy, precision, and F1-score compared to traditional approaches. But this work goes well beyond the technical boundaries of emotion recognition and raises ethical concerns and demands privacy and fairness in applications. More importantly, emotion-aware systems have transformative potential from customer sentiment analysis to adaptive learning environments and support for mental health.

An Analysis Of Requirements Of Deep Learning Based Classification Of Attacks Over Big Data Security

Authors: Assistant Professor Dr. Banita, Ms. Jyoti Ahlawat

Abstract: The rapid growth of big data across diverse digital ecosystems has made cyber security a critical concern, especially as traditional intrusion detection systems struggle to scale and adapt. This study explores the necessity and advantages of deploying deep learning-based techniques for the classification of cyber-attacks in big data environments. We will analyze the limitations of conventional machine learning models in handling high-volume, high-velocity, and high-variety data streams, emphasizing the unique challenges posed by modern attack vectors. The paper evaluates various deep learning architectures—including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and hybrid models—on their ability to detect complex and evolving threats. Additionally, we will address the infrastructural and computational requirements, such as distributed processing frameworks like Apache Spark and the role of GPU acceleration, to support deep learning at scale.

Experimental Investigation And Wear Characterization Of Detonation-Sprayed TiMo (CN) And NiCrAlY+CeO₂ Coatings On EN45 Steel For Automotive Applications

Authors: Assistant Professor Navdeep Singh Grewal, Veerpaul Soi

Abstract: This study presents a comprehensive examination of surface coating strategies to improve the wear resistance of EN45 steel using detonation gun (D-Gun) sprayed TiMo (CN) and NiCrAlY+0.4 wt% CeO₂ coatings. EN45 is commonly deployed in automotive suspension and axle systems but suffers from poor surface durability under frictional stress. Through X-ray diffraction (XRD), scanning electron microscopy (SEM), and Energy Dispersive X-ray Analysis (EDAX), the coatings were characterized for phase composition, morphology, and structural integrity. Pin-on-disc tests were performed under dry sliding conditions with varying loads to assess wear rate and frictional behavior. The results affirm that both coatings significantly enhance performance, with NiCrAlY+CeO₂ exhibiting superior wear resistance. These findings position detonation-sprayed coatings as viable upgrades for extending the service life of EN45-based automotive components.

DOI: http://doi.org/10.5281/zenodo.15729137

Enhancing Wear Resistance Of EN45 Steel Axles For Lightweight Trucks Using Detonation Sprayed Coatings

Authors: Assistant Professor Navdeep Singh Grewal, Veerpaul Soi

Abstract: This study investigates the enhancement of wear resistance in EN45 steel axles, widely used in lightweight trucks but limited by poor surface durability under abrasive and fatigue conditions. The application of advanced detonation-sprayed coatings—specifically TiMo (CN) and NiCrAIY+0.4 wt% CeO₂—is explored to improve mechanical performance. Systematic wear testing and microstructural characterization aim to establish a foundation for integrating these coatings in automotive axle applications.

DOI: https://zenodo.org/records/15728927

Deep Learning-Based Real-Time Fraud Detection In Digital Payments

Authors: Mr. Amit Punia, Dr. Neha Bhat

Abstract: The rapid growth of digital payment systems has significantly increased the risk of fraudulent transactions. Traditional fraud detection methods often fail to identify sophisticated and evolving fraud patterns in real time. This paper proposes a deep learning-based approach for detecting fraudulent transactions in digital payment systems. The model utilizes advanced neural network architectures to analyze transaction patterns and identify anomalies with high accuracy. By leveraging real-time data processing and adaptive learning techniques, the proposed system improves detection efficiency while minimizing false positives. Experimental results demonstrate that deep learning models outperform traditional machine learning approaches in terms of accuracy, precision, and recall. The study highlights the importance of intelligent systems in securing digital financial ecosystems.

DOI: https://doi.org/10.5281/zenodo.19234052