Velagapudi Ramakrishna Siddhartha Engineering College, Vijayawada

The “INDIAai” National AI Portal of India offers weekly articles that showcase the AI research contributions of universities across the country. Each article provides an in-depth report on the work of a specific university, allowing researchers and students to present concise explanations of their research. 

Siddhartha Academy of Higher Education (SAHE), formerly Velagapudi Ramakrishna Siddhartha Engineering College (VRSEC), is a Deemed to be University located in Vijayawada, Andhra Pradesh, India. It offers undergraduate programs in engineering and postgraduate programs in engineering, business administration, and computer applications. 

As the first private institution to provide engineering education in the United Andhra Pradesh, SAHE also pioneered offering postgraduate engineering programs in 1977. The University Grants Commission (UGC) approved it as an autonomous institution in 1977 and as a Deemed to be University in 2024.

This week, the portal highlights the top AI research contributions from Velagapudi Ramakrishna Siddhartha Engineering College, Deemed to be University, Kanuru, Vijayawada.

Authors: 

• Pothuri Surendra Varma, Department of Computer Science and Engineering, Velagapudi Ramakrishna Siddhartha Engineering College, Vijayawada, India
• Veena Anand, Department of Information Technology, Atal Bihari Vajpayee Indian Institute of Information Technology and Management (ABV-IIITM), Gwalior, India
• Praveen Kumar Donta, Distributed Systems Group, TU Wien, Wien, Austria

Journal: IEEE Transactions on Consumer Electronics ( Volume: 70, Issue: 1, February 2024)

Indoor positioning (IP) is crucial for various smart indoor applications, such as automated energy management, patient tracking, and indoor navigation. Traditional systems use Wi-Fi access points and a centralized server, which can compromise user privacy by disclosing the location of target nodes. To address this, a federated KNN-based technique is proposed, which encrypts the location coordinates of each target node, ensuring privacy. While this approach can increase localization error, the use of Hausdorff distance to deploy dynamic access points based on target node movements within convex hull regions improves accuracy. The proposed system enhances localization accuracy without additional hardware, making it cost-efficient.

The AI aspect in this work involves using a federated KNN algorithm to enhance privacy in indoor positioning by encrypting location data. It also utilizes AI techniques to optimize access point deployment using the Hausdorff distance, improving accuracy while maintaining cost efficiency and not compromising user privacy.

Authors: 

• Naga Venkata Rishika Guggilam, Department of Computer Science and Engineering, VR Siddhartha Engineering College(A), Siddhartha Academy of HIgher Education (Deemed to be University), Vijayawada, India, 
• Rupa Chiramdasu, Department of Computer Science and Engineering, VR Siddhartha Engineering College(A), Siddhartha Academy of HIgher Education (Deemed to be University), Vijayawada, India, 
• Akhil Babu Nambur, Department of Computer Science and Engineering, VR Siddhartha Engineering College(A), Siddhartha Academy of HIgher Education (Deemed to be University), Vijayawada, India, 
• Naveena Mikkineni, Department of Computer Science and Engineering, VR Siddhartha Engineering College(A), Siddhartha Academy of HIgher Education (Deemed to be University), Vijayawada, India, 
• Yaodong Zhu, Jiaxing University School of Information Science and Engineering, Jiaxing Zhejiang 314001, China, 
• Thippa Reddy Gadekallu, Division of Research and Development, Lovely Professional University, Phagwara, India.

Journal: Computers & Security, Elsevier

The integration of deep learning with security is advancing rapidly, particularly in maritime surveillance. Despite challenges like low accuracy, high computational complexity, and limited GPU utilization, deep learning applications require large datasets for accurate results. This paper introduces a privacy-preserving deep learning-based vessel monitoring system that emphasizes data integrity and authenticity. The system integrates the YOLOv8 model with SHA-256 to track and classify vessels while ensuring data security. It uses a balanced dataset of 693 photo-realistic video sequences and includes CSPDarkNet53, Spatial Pyramid Pooling, and a head layer for detection. The proposed model improves precision by 9.3% over YOLOv7, offering superior performance compared to other state-of-the-art methods.

The AI perspective in this work involves integrating the YOLOv8 deep learning model with SHA-256 for vessel monitoring, ensuring data integrity and authenticity. The system optimizes GPU utilization and improves accuracy in maritime surveillance, addressing challenges like low accuracy and high computational complexity. The model uses a class-balanced dataset and advanced architecture for better feature extraction and bounding box prediction, achieving a 9.3% increase in precision over YOLOv7.

Authors: 

• Radhesyam Vaddi, Velagapudi Ramakrishna Siddhartha Engineering College, Department of Information Technology, Vijayawada 520007, India, 
• Phaneendra Kumar B.L.N, Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation Vaddeswaram, Andhra Pradesh 522302, India, 
• Prabukumar Manoharan, Vellore Institute of Technology, School of Computer Science Engineering & Information Systems, Vellore 632014, India, 
• L. Agilandeeswari, Vellore Institute of Technology, School of Computer Science Engineering & Information Systems, Vellore 632014, India, 
• V. Sangeetha, Vellore Institute of Technology, School of Computer Science Engineering & Information Systems, Vellore 632014, India. 

Journal: The Egyptian Journal of Remote Sensing and Space Sciences

Technological advancements in spectroscopy enable the collection of extensive data about materials on Earth’s surface, with numerous applications. However, hyperspectral images (HSIs) often contain hundreds of spectral bands, many of which are highly correlated, and limited training samples, reducing classification accuracy. Dimensionality Reduction (DR) is essential before HSI classification to address this issue. This work reviews and compares various DR techniques for HSI classification, providing guidance for researchers in selecting appropriate methods for real-time applications. Key challenges addressed include the large volume of hyperspectral data, nonlinearity, redundancy among bands, the “curse of dimensionality,” and maintaining computational efficiency.

The AI perspective in this work focuses on using dimensionality reduction (DR) techniques to improve the classification accuracy of hyperspectral images (HSI). The study reviews and compares various DR methods, which are essential due to the highly correlated information in spectral bands and limited training samples. This comparative analysis aids researchers in selecting the appropriate DR technique for real-time applications in hyperspectral remote sensing.

Authors:

• Yalamanchili Salini, Department of Information Technology, Velagapudi Ramakrishna Siddhartha Engineering College, Kanuru, Vijayawada, India
• Sachi Nandan Mohanty, School of Computer Science & Engineering (SCOPE), VIT-AP University, Amaravati, Andhra Pradesh, India
• Janjhyam Venkata Naga Ramesh, Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, Guntur, Andhra Pradesh, India
• Ming Yang, College of Computing and Software Engineering, Kennesaw State University, Kennesaw, GA, USA
• Mukkoti Maruthi Venkata Chalapathi, School of Computer Science & Engineering (SCOPE), VIT-AP University, Amaravati, Andhra Pradesh, India

Journal: IEEE Access

Pregnancy complications can significantly impact maternal and child health, making early detection vital. Traditional manual analysis of cardiotocography (CTG) tests, used to monitor fetal health, is labor-intensive and often unreliable. This study explores the use of machine learning (ML) techniques to improve fetal health classification based on CTG data. Using a small yet rich CTG dataset, various ML models, including Random Forests, Logistic Regression, Decision Trees, Support Vector Classifiers, Voting Classifiers, and K-Nearest Neighbors, were tested. The ML models achieved a high accuracy of 93%, demonstrating their effectiveness in improving diagnostic precision and facilitating timely medical interventions. The findings advocate for integrating ML models into clinical practice to streamline fetal health assessments, optimize resource allocation, and enhance prenatal care. Further research is encouraged to refine and expand ML applications in this field.

This study applies machine learning (ML) techniques to improve the classification of fetal health using cardiotocography (CTG) data. It compares various ML models, including Random Forests and Support Vector Classifiers, to enhance diagnostic accuracy and enable timely interventions. The models achieved a notable accuracy of 93%, highlighting the potential of ML in streamlining prenatal care and optimizing medical resources. The research underscores the importance of early detection of pregnancy complications and advocates for integrating ML models into clinical practices.

Authors: 

• Shaik Sajiha, Department of Electronics & Communication Engineering, Velagapudi Ramakrishna Siddhartha Engineering College, Kanuru, Vijayawada, 520007, Andhra Pradesh, India
• Kodali Radha, Department of Electronics & Communication Engineering, Velagapudi Ramakrishna Siddhartha Engineering College, Kanuru, Vijayawada, 520007, Andhra Pradesh, India
• Dhulipalla Venkata Rao, Department of Electronics & Communication Engineering, Velagapudi Ramakrishna Siddhartha Engineering College, Kanuru, Vijayawada, 520007, Andhra Pradesh, India
• Nammi Sneha, Department of Electronics & Communication Engineering, Velagapudi Ramakrishna Siddhartha Engineering College, Kanuru, Vijayawada, 520007, Andhra Pradesh, India
• Suryanarayana Gunnam, Department of Electronics & Communication Engineering, Velagapudi Ramakrishna Siddhartha Engineering College, Kanuru, Vijayawada, 520007, Andhra Pradesh, India
• Durga Prasad Bavirisetti, Department of Computer Science, Norwegian University of Science and Technology, 7034, Trondheim, Norway

Journal: EURASIP Journal on Audio, Speech, and Music Processing

Dysarthria is a speech disorder characterized by articulation difficulties, impacting communication. This research introduces a novel method for automatic dysarthria detection (ADD) and severity assessment (ADSLA) using a variable continuous wavelet transform (CWT) layered convolutional neural network (CNN) model. The model’s effectiveness was evaluated using two datasets, TORGO and UA-Speech, containing speech signals from dysarthria patients and healthy individuals. The study examined different wavelets, including Amor, Morse, and Bump, for their ability to model raw waveforms without feature extraction. Results showed that the Amor wavelet excelled in signal representation, noise suppression, and feature extraction, making it the most accurate choice. The proposed CWT-layered CNN model highlights the importance of selecting the right wavelet for signal processing tasks, with potential applications in speech recognition and early diagnosis of dysarthria, streamlining intervention measures.

This research develops a novel method using a variable continuous wavelet transform (CWT) layered convolutional neural network (CNN) for automatic dysarthria detection and severity assessment. By utilizing wavelets like Amor, Morse, and Bump, the study finds the Amor wavelet most effective in accurately representing speech signals, surpassing the others in signal reconstruction, noise suppression, and feature extraction. The approach preserves the original signal’s nuances, crucial for applications in speech and signal processing. The model, tested on the TORGO and UA-Speech datasets, aims to streamline early diagnosis and intervention for dysarthria.