Название: Next-Generation Smart Systems: Bridging Sustainability and Computational Intelligence
Автор: V. Sekhar, P. Ajay Kumar Reddy, K. Logesh, T. Raghunadha Reddy, Dara Raju
Издательство: CRC Press
Серия: Taylor and Francis Proceedings in Computer Science and Engineering
Год: 2027
Страниц: 480
Язык: английский
Формат: pdf (true)
Размер: 16.9 MB

This book provides a comprehensive exploration of how next-generation computing drives ecological and social resilience and highlights the pivotal role of intelligent automation and data-driven strategies in solving complex global challenges.

It delves into the integration of Artificial Intelligence (AI), Internet of Things (IoT), and high-performance computing to create "green" technological ecosystems. This book covers a broad spectrum of research, including smart grid management, precision agriculture, carbon footprint optimization through Machine Learning, and the development of circular economy models via blockchain. By synthesizing theoretical frameworks with real-world case studies, the contributions illustrate how advanced algorithms can minimize resource waste while maximizing industrial efficiency. Together, these papers offer a roadmap for sustainable digital transformation across energy, healthcare, and urban planning sectors by emphasizing the shift toward "Cognitive Sustainability", where systems do not just automate tasks but actively learn to balance human needs with environmental constraints.

Transfer Learning-Based Autonomous Landing Site Detection for Unmanned Aerial Vehicles (UAVs): Drones require safe landing and it travels in any different environmental conditions even in congested places. Autonomous landing scene recognition plays a vital role for down navigation, which ensures safe landing in unpredictable environments with varying conditions related to the terrain and obstacles. There are many methods where traditional methods use handcrafted future extraction, which struggle to adapt and which struggles to adapt dynamic settings. As everything needs advancements, deep learning advancement improvised the capabilities of scene recognition. This paper dive deep into a transfer learning-based approach using deep convolutional neural networks (CNNs). Identification of the landing site for a drone is a major base for safe operations. The data, named landing scenes-7 which includes the ability of images captured at different altitudes, lighting conditions and geographical locations, was evolved to train and evaluate the model. The architecture that is introduced in the system is ResNeXt-50 optimized with the ADAM optimizer, known for its group conversion structure that enhances future extraction and model Generalization. The analysis between the comparison of ResNeXt-50 outperforms ResNet-50 and stochastic gradient descent (SGD) processing speed, classification, accuracy, and robustness in detecting safe landing zones.

Emotion Detection Using Voice Modulation: In order to enhance human-computer interaction, particularly in customer-agent service scenarios, this research presents a sophisticated voice modulation-based emotion recognition system. The proposed approach integrates Wav2Vec2.0 characteristics and Mel-frequency cepstral coefficients (MFCC) to effectively capture time-frequency and contextual voice representations. These features are fed into a Convolutional Neural Network (CNN) that has been trained to classify emotions such as surprise, rage, sadness, and happiness. The model performed better than standalone MFCC-based techniques and conventional models like SVM, with a test accuracy of 85.75%. A comprehensive evaluation is conducted to validate the system's robustness across different emotion classes using precision, recall, and F1-score measures. An ablation study and comparing baselines are left for future research, although it has been empirically shown that the combined feature strategy enhances model performance. Figures illustrating feature distributions and model performance are cited in the study; nevertheless, they need to be properly embedded and described. Issues with diversity, dataset bias, and licensing persist despite the system's potential for real-time affective computing applications. Future developments will include expanding the dataset, optimizing the architecture, and looking into more emotion categories. This work increases the integration of emotional intelligence in automated systems and paves the way for more compassionate AI-driven communication tools.

Unsupervised Video Summarization Through Temporal Attention Mechanisms: The rapid growth of multimedia data, there is a pressing need for systems that can automatically generate concise summaries and meaningful captions for long videos without relying on manual annotations. In this paper, we propose an unsupervised video summarization framework that integrates temporal segmentation with temporal attention mechanisms to identify and focus on semantically important content. Our method segments videos into coherent events and applies attention to learn which segments are most relevant, thus improving the coherence and informativeness of the generated summaries. Leveraging advances in transformer-based architectures and multimodal large language models (LLMs), our system effectively combines visual, audio, and textual cues to produce richer, context-aware summaries. Experiments conducted on lecture videos demonstrate that our model achieves competitive F1-scores and BLEU-4 scores, confirming its adaptability to diverse video content. This framework addresses the limitations of supervised methods by eliminating the need for large labeled datasets and shows potential for real-world applications such as content retrieval, accessibility, and highlight detection.This paper presents an unsupervised framework for video summarization that combines temporal segmentation with temporal attention mechanisms.

This book is an essential resource for academic researchers, post-graduate students, and industry practitioners specializing in Computer Science and environmental engineering. It also serves policymakers and urban planners seeking data-backed strategies to implement sustainable technologies within smart city infrastructures and corporate ESG frameworks.

Contents: