Linga Reddy Cenkeramaddi

• Arigela, Sanjana; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2026). mmWave Radar-Based Wall Classification Using Progressive Feature Selection. IEEE Sensors Journal. ISSN 1530-437X. 26(18), s. 28013–28021. doi: 10.1109/JSEN.2026.3716595. Fulltekst i vitenarkiv Vis sammendrag

  This article proposes a progressive feature selection framework for millimeter-wave (mmWave) radar-based wall material classification. Raw complex baseband I–Q data are collected using a TI AWR1843 radar over multiple days from eight indoor surface types: transparent glass wall, smooth glass board, cement wall, metal door, vinyl floor, olefin carpet floor, ceramic tile, and concrete wall. We introduce an incremental-data decremental-feature learning (IDDFL) strategy that incrementally aggregates daywise training data while progressively pruning low-contribution features using a posthoc explainer. A preprocessing module extracts 100 handcrafted descriptors covering time-domain, frequency-domain, time–frequency, and radar-specific characteristics from the acquired I–Q measurements. Feature relevance is quantified via SHapley additive exPlanations (SHAP) analysis on a LightGBM (LGBM) classifier, and the least discriminative features are removed across days. The proposed IDDFL process reduces the feature set from 100 to 30 while maintaining high classification performance. The proposed model achieves a final test accuracy of 79.80% with an F1 -score of 0.79 , demonstrating improved computational efficiency with minimal performance degradation. The final 30-feature LGBM model outperforms nine state-of-the-art benchmark methods reported in the literature and is deployed on edge platforms (Raspberry Pi 4 and Rock Pi 4 C+) to demonstrate real-time applicability. The proposed feature selection and edge-ready pipeline is well-suited for building inspection, autonomous navigation, and security applications.

• Reddy, Madhava Krishna Kadiyala Muni Venkata; Syversen, Andreas Ekeberg; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2026). A Multinode Thermal Sensor Network With Edge AI Inference for Indoor Moisture and Humidity Classification. IEEE Sensors Journal. ISSN 1530-437X. 26(18), s. 27081–27089. doi: 10.1109/JSEN.2026.3719511. Fulltekst i vitenarkiv
• Thangaraj, Priya; Bethi, Pardhasaradhi; Granmo, Ole-Christoffer & Cenkeramaddi, Linga Reddy (2026). Clustering-based Significance-Aware Automated Data Binarization Framework for Tsetlin Machines. IEEE Transactions on Artificial Intelligence (TAI). doi: 10.1109/TAI.2026.3737096. Fulltekst i vitenarkiv Vis sammendrag

  The Tsetlin machine (TM) is an interpretable machine learning (ML) algorithm that requires relatively modest computational resources. While most ML models use continuous data representations, TM operates on binary inputs. Binarization of continuous data via brute-force thresholding has shown promising accuracy but can be computationally expensive and hinder generalization. Moreover, traditional binarization approaches, such as mean, median, and fixed thresholds, yield single-bit representations. Multi-bit representations, such as quantile, histogram, and standard binarization, are more expressive, but typically require manual definition of the number of threshold levels or the number of bits per feature. However, not all features require the same number of bits, nor do they contribute equally to prediction. To address this, we propose a simple and adaptive binarization framework that automatically determines how each feature should be represented based on its data characteristics. Instead of relying on fixed thresholds. This method identifies meaningful groups within each feature and encodes them accordingly, resulting in a more informative and balanced representation. In this paper, we propose a significance-aware adaptive binarization framework for TM learning based on k-means clustering. Each feature is individually analyzed using the Silhouette score to determine its optimal clustering structure. Features are then one-hot encoded according to their cluster partitions, resulting in a significance-aware binarized representation. All the binarized features are concatenated to form the final input for the TM. To ensure scalability, the framework includes an efficient mechanism to handle large datasets while maintaining consistency in the encoding process. To reduce the complexity of clustering, we propose a chunk-based centroid-to-centroid association mechanism, efficiently solved using the Hungarian algorithm. The proposed method is evaluated on multiple datasets and benchmarked against traditional single-bit and multi-bit binarization techniques. Moreover, we extend this analogy to continual feature binarization.

• Abbu, Soumya; Kummari, Naveen Kumar; Babu, Sobhan; Cenkeramaddi, Linga Reddy & Chalavadi, Krishna Mohan (2026). Ego-centric congestion understanding from multi-camera traffic scenes: A benchmark dataset and graph reasoning approach for autonomous vehicles. Applied Soft Computing. ISSN 1568-4946. 201. doi: 10.1016/j.asoc.2026.115701. Fulltekst i vitenarkiv Vis sammendrag

  The rapid advancements in autonomous vehicles (AVs) highlight the need for robust perception and planning systems to navigate complex traffic scenarios. AVs rely on congestion state estimation to assess real-time traffic, optimize routes, and navigate safely in dynamic environments. Accurate congestion estimation faces challenges like dynamic conditions and scaling to process large urban traffic data. Existing methods use spatio-temporal graph models with LiDAR or cameras. While LiDAR offers high accuracy, its cost makes camera-based systems a viable alternative. Single-camera setups are limited by narrow fields of view and incomplete coverage, highlighting the need for multi-camera approaches that enhance coverage and scalability. In this work, an ego-centric perspective on traffic scene understanding is adopted, where congestion is defined relative to the autonomous vehicle’s immediate surroundings rather than global traffic flow. Grounded in this perspective, an end-to-end Autonomous Vehicle Congestion State Classification Network (AVCSCNet) is proposed using the past-and-future reasoning for tracking (PF-Track) backbone to generate bird’s-eye view (BEV) visualizations of multi-camera inputs. BEV data is transformed into 2D spatio-temporal graphs, with the ego vehicle and surrounding agents as nodes and edges representing relationships. Further, GNNs are employed for spatial feature extraction, followed by LSTMs for traffic state classification into: Congested, Neutral, and Free-flow. In addition, we introduce the Autonomous Vehicles Congestion 3 (AVCong3) dataset, featuring manually annotated spatio-temporal graphs generated using multi-camera data and three traffic classes. Extensive experiments on AVCong3 and PREVENTION datasets demonstrate the effectiveness and scalability of the AVCSCNet framework. In particular, AVCSCNet achieves 86.6% congestion classification accuracy using six cameras, significantly improving from 68.3% with a single camera setup, while also demonstrating promising generalization on the PREVENTION dataset with 78.05% accuracy.

• Bhardwaj, Jyotirmoy & Cenkeramaddi, Linga Reddy (2026). Unsupervised Detection of Multivariate Water Quality Anomalies via Residual Feature Embeddings. IEEE Access. 14, s. 97948–97961. doi: 10.1109/ACCESS.2026.3704534. Fulltekst i vitenarkiv Vis sammendrag

  Monitoring water quality in water utilities via multiple distributed sensors is critical for detecting diverse pollution events. Multivariate water-quality time series, generated by such water quality sensors, often exhibit overlapping seasonal patterns, long-term trends, and irregular disturbances, making anomaly detection challenging without extensive labeled data. In this work, we present an anomaly detection approach based on unsupervised machine learning (ML) using residual feature embeddings and multivariate clustering. To extract anomalies, for each sensor, we compute candidate seasonal periods and fit a polynomial trend, yielding residual vectors that serve as feature embeddings for Isolation Forest (IF), an unsupervised ML approach. IF was used to identify anomalous observations through residual embeddings, and K-Means clustering was used to group residual vectors into similar water-quality events. To aid interpretability, the clusters were further analyzed statistically to identify the potential signatures of anomalous events. Validation is performed on data annotated by domain experts. This approach enables robust unsupervised anomaly detection by effectively handling multiple overlapping seasonality and trends that might otherwise obscure anomalies. An experimental evaluation of multivariate water-quality datasets demonstrates that the proposed method successfully identifies unusual events and provides valuable insights for water utilities.

• Achala, G.; Acharya, U. Shripathi; Srihari, Pathipati & Cenkeramaddi, Linga Reddy (2026). On the design and implementation and evaluation of Euclidean Geometry Quasi-cyclic LDPC codes for NAND flash memories. IEEE Open Journal of the Communications Society. doi: 10.1109/OJCOMS.2026.3696826. Fulltekst i vitenarkiv
• Lade, Modha Sreya; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2026). Thermal Imaging-based Perception for Automotive Edge AI Applications using a Lightweight Custom YOLO Model. IEEE Sensors Journal. ISSN 1530-437X. doi: 10.1109/JSEN.2026.3691670. Fulltekst i vitenarkiv
• Manda, Advik; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2026). TMCR Edge-YOLO: Thermal Sensing-based Mobility-Context Recognition on the Edge. IEEE Sensors Journal. ISSN 1530-437X. doi: 10.1109/JSEN.2026.3686674. Fulltekst i vitenarkiv
• Patel, Sumit; Sharma, Sahil; Gandi, Suman; Parne, Saidi Reddy & Cenkeramaddi, Linga Reddy (2026). Biomimetic and bio‑inspired multispectral stealth materials: Design strategies, fabrication, and radar infrared performance. Next Materials. 12. doi: 10.1016/j.nxmate.2026.102117. Fulltekst i vitenarkiv Vis sammendrag

  Biomimetic and bio-inspired approaches offer a promising route to next-generation multispectral stealth by combining structure-driven electromagnetic (EM) wave control with materials-level thermal regulation. This review synthesizes recent advances in biomimetic radar and infrared-stealth materials, focusing on biological design principles, fabrication strategies, EM- and IR-loss mechanisms, and multifunctional integration. We first summarize the theoretical basis of radar cross-section reduction and IR signature suppression, then categorize key bio-inspired motifs-moth-eye, lotus-leaf, spider-web, honeycomb, plant-derived and other hierarchical structures and discuss how each modulates scattering, impedance matching, multiple internal reflection, interfacial/dipolar polarization and conductive/magnetic loss to achieve broadband microwave absorption. Next, we review IR-stealth strategies inspired by polar bear fur, Saharan silver ants, cephalopods and other organisms for thermal insulation, emissivity control and adaptive camouflage. Representative fabrication methods-hydrothermal, electrospinning, templating, 3D printing, freeze-drying, CVD and metamaterial patterning and common compositional platforms including carbonaceous materials, MXenes, conductive polymers, ferrites and nanocomposites are compared with emphasis on process–structure–function links. We highlight examples of multifunctional stealth materials that simultaneously deliver low radar cross-section, reduced IR emissivity, mechanical robustness, superhydrophobicity and lightweight design. Finally, we identify current challenges-multispectral compatibility, environmental durability, thickness/weight constraints, and scalable manufacturing and propose future directions for integrated material structural optimization and active/adaptive stealth systems. This comprehensive review provides design guidelines and a roadmap for developing high-performance biomimetic multispectral stealth materials for aerospace, naval and wearable applications.

• Singh, Nitin Pratap; Lata, Kusum & Cenkeramaddi, Linga Reddy (2026). Lattice thermal conductivity of MgB<inf>2</inf>: Scattering physics and design strategies for quantum-enabling superconducting systems. Materials Today Advances. 30. doi: 10.1016/j.mtadv.2026.100795. Fulltekst i vitenarkiv
• Shivakumar, Thrishank Hesaraghatta; Bethi, Pardhasaradhi; Boppu, Srinivas; Granmo, Ole-Christoffer & Cenkeramaddi, Linga Reddy (2026). Multi-Sensor Fusion using Composite Tsetlin Machine with Adaptive Feature Bank Selection. IEEE Sensors Journal. ISSN 1530-437X. 26(11), s. 17210–17223. doi: 10.1109/JSEN.2026.3680022. Fulltekst i vitenarkiv Vis sammendrag

  A wide range of applications relies on multisensor setups to enhance their perception and decision-making capabilities. These sensors generate raw time-series data that can be transformed into the frequency domain using feature banks such as mel frequency cepstral coefficients (MFCC), Chroma-STFT (C-STFT), and spectral descriptors. Most deep learning (DL) and machine learning (ML) models utilize these frequency-domain representations, typically in the form of numeric data or spectrogram images, to enhance their decision-making capabilities. However, interpretability, resource constraints, and inference time remain challenging in DL and ML algorithms. The Tsetlin machine (TM) is an energy-efficient alternative based on propositional logic. TM applications to multisensor classification and fusion remain largely unexplored. In this work, a novel multisensor fusion framework using composite-TM (Co-TM) is proposed, where each sensor’s time-series signals are converted into frequency-domain representations using several feature banks. The framework adaptively selects the best frequency-domain representation during the feature bank selection phase and performs sensor-level fusion during training. The feature bank selection is guided by the TM’s learning and delearning behaviors, while sensor-level fusion is achieved through a Co-TM framework. To monitor TM learning dynamics and adaptively select the desired feature bank for each sensor, we propose a set of metrics, including average accuracy, top-m accuracy, stability coefficient, and the slope of the accuracy curve. The proposed framework is validated using an industrial machine fault detection dataset (multisensor heterogeneous data) and a rotating machinery fault dataset (multisensor homogeneous data). The experimental results demonstrate that the proposed method achieves state-ofthe- art accuracy compared with ML, convolutional neural network (CNN) models, and TM models.

• Puduru, Vamsi Krishna; Yakkati, Rakesh Reddy; Yeduri, Sreenivasa Reddy; Koorapati, Sagar & Cenkeramaddi, Linga Reddy (2026). WiAR: Wi-Fi-based human activity recognition using time-frequency analysis and lightweight deep learning for smart environments. Internet of Things: Engineering Cyber Physical Human Systems. 36. doi: 10.1016/j.iot.2026.101881. Fulltekst i vitenarkiv Vis sammendrag

  Recognizing human activities in smart environments has significant usage in home automation, security, healthcare monitoring, etc. This paper proposes WiAR, which is a Wi-Fi-based human activity recognition method using Continuous Wavelet Transform (CWT) and lightweight Convolutional Neural Networks (CNNs). The proposed approach is evaluated on the IEEE 802.11ax Channel State Information (CSI) dataset. First, WiAR utilizes the CWT to generate the spectrogram images from the CSI extracted from Wi-Fi signals for different activities: walking, running, staying in place, and empty space. Then, these spectrogram images are processed with a CNN to classify these activities efficiently. Experimental results show that the proposed WiAR achieves an accuracy of approximately 91.1% when compared to various pre-trained models such as DenseNet, EfficientNet, MobileNet, ResNet, and VGGNet. Finally, the proposed CNN model is deployed on various edge computing devices, including Raspberry Pi 5, to validate its real-time implementation in terms of inference time.

• Sonny, Amala & Cenkeramaddi, Linga Reddy (2026). Q-Learning-Based Multi-UAV Path Planning with Dynamic Obstacle Avoidance. Communication Systems and Networks and Workshops. Proceedings. ISSN 2155-2487. s. 627–632. doi: 10.1109/COMSNETS67989.2026.11418253. Fulltekst i vitenarkiv
• Gadari, Charitha; Yeduri, Sreenivasa Reddy; Granmo, Ole-Christoffer & Cenkeramaddi, Linga Reddy (2026). Human Activity Recognition Using 5G RF Sensing and Interpretable Tsetlin Machines. Communication Systems and Networks and Workshops. Proceedings. ISSN 2155-2487. s. 953–958. doi: 10.1109/COMSNETS67989.2026.11418257. Fulltekst i vitenarkiv
• Salvi, Bhumika; Yeduri, Sreenivasa Reddy; Granmo, Ole-Christoffer & Cenkeramaddi, Linga Reddy (2026). Tsetlin Machine-based Network Anomaly Detection System. Communication Systems and Networks and Workshops. Proceedings. ISSN 2155-2487. s. 67–74. doi: 10.1109/COMSNETS67989.2026.11418184. Fulltekst i vitenarkiv
• Soumya, A.; Dubey, Abhishek Kumar; Mohan, C. Krishna & Cenkeramaddi, Linga Reddy (2026). Point cloud-based multi-target 3D object detection using LiDAR sensor and deep learning. Journal of Ambient Intelligence and Humanized Computing. ISSN 1868-5137. 17(2), s. 413–428. doi: 10.1007/s12652-026-05037-y. Fulltekst i vitenarkiv Vis sammendrag

  Point cloud-based three-dimensional (3D) object detection is a critical task in autonomous driving, robotics, and augmented reality, where accurate localization and classification of objects are essential under diverse and challenging scenarios. This work introduces a point cloud-based multi-target 3D object detection framework using Light Detection and Ranging (LiDAR) sensors. The key contribution lies in improving the encoding technique to preserve spatial information and adapt to varying point densities, enabling efficient processing of raw LiDAR data for robust 3D detection. Specifically, we design an efficient network architecture that incorporates a single unified classification head to jointly handle positive and negative samples, simplifying the design and improving training stability. Our model detects all classes (car, pedestrian, cyclist) simultaneously within a single network, enhancing computational efficiency while avoiding separate networks for each class. Furthermore, the proposed feature extraction pipeline combines a voxel-based encoder with a sparsely embedded convolutional backbone and a feature pyramid network, facilitating multi-scale feature representation and effective detection of objects at different scales. This encoding method facilitates the efficient processing of raw LiDAR data, enabling accurate object detection across diverse scenarios. As a result, our model achieves state-of-the-art performance on the KITTI dataset, surpassing baseline methods such as PointPillars and VoxelNet by delivering superior Average Precision (AP) across all difficulty levels for the car class in both bird’s-eye view (BEV) and 3D object detection tasks. Specifically, it achieves AP40 scores of 92.84%, 88.63%, and 85.68% in BEV detection and 87.84%, 76.65%, and 73.60% in 3D detection for Easy, Moderate, and Hard levels, respectively. These results highlight the effectiveness of our encoding technique in enhancing model efficiency and detection accuracy across diverse scenarios.

• Saha, Ramesh; Bharadwaj, Sanjeev Kumar; Saif, Sohail; Biswas, Suparna; Bansal, Mohan & Jain, Prateek [Vis alle 7 forfattere av denne artikkelen] (2025). SWAST KHOJ: An IoT-driven real-time health monitoring system prototype. Internet of Things: Engineering Cyber Physical Human Systems. 34. doi: 10.1016/j.iot.2025.101796. Fulltekst i vitenarkiv Vis sammendrag

  This study focuses on developing a health monitoring system prototype using the Internet of Medical Things (IoMT) to provide continuous and real-time tracking of vital signs. The advent of IoT-enabled healthcare equipment is drastically changing the landscape of healthcare infrastructure, facilitating seamless communication among diverse devices and platforms. In this context, the significance of smart health monitoring applications in IoT has led to the exploration of various IoT frameworks by researchers. In this work, we have deployed a Wireless Body Area Network (WBAN) architecture in real-time healthcare prototype using discrete heterogeneous sensor nodes. The purpose of this prototype is to collect vital signs, which include temperature, heart rate, peripheral oxygen saturation (SpO2), and electrocardiogram (ECG). A Graphical User Interface (GUI) is used to display the gathered data in accordance with the Modified Early Warning Score (MEWS) medical criteria also risk level classification using k-Nearest Neighbor algorithm (kNN) based trained model. Periodically, our prototype, called SWAST KHOJ, uploads the data together with the timestamp to a local server. Additionally, we evaluate the prototype’s Quality of Service (QoS) parameter in an interior setting using a variety of short-range communication techniques, such as Bluetooth, ZigBee, and Local Area Network (LAN) or wired connection. Our evaluation reveals end-to-end delays of 0.514 ms for wired communication, 0.62 ms for Wi-Fi, 0.417 ms for ZigBee, and 1.92 ms for Bluetooth. Notably, our method demonstrates lower end-to-end delays compared to previous studies. In addition, we test the prototype’s throughput with several connection protocols.

• Akella, Sriram Mahateja; Thangaraj, Priya; Yeduri, Sreenivasa Reddy & Cenkeramaddi, Linga Reddy (2025). 5GDC-estimating drone count using 5G CSI measurements and multi-channel CNN. Journal of Ambient Intelligence and Humanized Computing. ISSN 1868-5137. doi: 10.1007/s12652-025-05026-7. Fulltekst i vitenarkiv
• Raj, Goutham; Narayanan, Sivaranjini Perikamana; Manikandan, M. Sabarimalai; Mathew, Raji Susan & Cenkeramaddi, Linga Reddy (2025). Thermal Imaging-based Intrusion Detection using Deep Learning. IEEE Sensors Journal. ISSN 1530-437X. 26(3), s. 3658–3665. doi: 10.1109/JSEN.2025.3641595. Fulltekst i vitenarkiv Vis sammendrag

  Intrusion warning systems serve as the first line of defense against unauthorized access, theft, vandalism, and other security breaches in both residential and commercial settings. Automated intrusion warning systems enhance security by continuously monitoring sensitive areas and generating immediate alerts upon detecting suspicious activity. Thermal imaging is preferred over RGB camera-based human presence detection due to the increased privacy and usability in outdoor environments, irrespective of the weather conditions and illumination. Accurately classifying individuals in thermal images is vital for identifying potential threats and implementing effective security measures. In this article, we present a thermal imaging-based human pose detection model using the you only look once (YOLO) architecture for intrusion warning systems. The model can detect five intruder poses, including creeping, crawling, stooping, climbing, and “other,” which can be used to detect the presence of an intruder. We also study the effectiveness of various YOLO architectures for intruder detection. Experimental results demonstrate that the YOLOv12n-based human pose classification model achieved superior performance compared to many different state-of-the-art models, with a precision of 94.1%, recall of 95.3%, mean average precision of 97.7%, and class-wise accuracies above 96% across the five classes. The model showcased robust performance across various image degradations and geometrical transformations. The YOLO architectures were also implemented on various edge computing platforms, showcasing promising real-time performance.

• Beiermann, Sindre; Tande, Andrea Emilie Johnsen; Yeduri, Sreenivasa Reddy & Cenkeramaddi, Linga Reddy (2025). Multi-target classification and localization using thermal camera and deep learning. Journal of Ambient Intelligence and Humanized Computing. ISSN 1868-5137. 17, s. 277–289. doi: 10.1007/s12652-025-05033-8. Fulltekst i vitenarkiv Vis sammendrag

  Reliable detection and localization of small objects such as unmanned aerial vehicles, birds, and other aerial vehicles is highly challenging. Further, recent studies have mostly focused on utilizing RGB cameras for the detection and classification of these objects. Nevertheless, extreme weather conditions like fog, snow, and dust complicate RGB-based localization and identification. To overcome these challenges, in this work, a thermal imaging-based target detection and localization method is proposed that performs well even in poor lighting conditions. Initially, we created a dataset of thermal images corresponding to humans, cars, and drones. Thereafter, the YOLOv5 model is utilized by the proposed method for the classification and localization of these targets. In contrast to the state-of-the-art, the proposed method localizes the target based on the angle at which it is situated within the field of view of the camera. Through numerical results, the performance of the proposed method is evaluated for target detection. Further, the k-fold cross-validation results are presented for obtaining the angle at which the target is located. Finally, root mean square error is used to find the error between the true angle and the predicted angle of the target using the proposed method.

• Utture, Rohil P.; Biju, Alan K.; Yarlagadda, Vishnu Srinivasa Murthy; Cenkeramaddi, Linga Reddy & Kumari, Nancy (2025). Weather Prediction based on the Cloud Images using Machine and Deep Learning Algorithms, 2025 International Conference on Next Generation Communication & Information Processing (INCIP). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331528140. s. 937–942. doi: 10.1109/INCIP64058.2025.11020379. Fulltekst i vitenarkiv Vis sammendrag

  The process of weather forecasting based on the climatic images captured at various geographical locations and times can guide humans to plan their daily routines. In this paper, we propose forecasting the weather based on the images captured from the sky. The image dataset collected from various online search engines and mobile cameras pertains to timings such as sunrise, sunset, daytime, and night. Care has been taken to consider various weather conditions such as clear, foggy, rainy, and glare. Machine learning (ML) is widely adapted for various real-time applications. The inception of deep learning (DL) architectures has already been proven to outperform the traditional feature engineering approaches. This paper proposes two approaches. One is using the traditional ML models and the other is using trending convolutional neural networks (CNNs). In ML-based approach, decision tree (DT) and random forest (RF) are considered. Manually extracted features such as brightness, sharpness, contrast, white pixel intensity, haze factor, and color histogram are fed to them. Meanwhile, in the CNN-based approach, VGG16 and RESNET architectures are used. The performance of CNNs is found to be better when compared to the traditional ML algorithms. The basic CNN, VGG16, and RESNET models provide an accuracy of 62%, 72%, and 73.35%, respectively.

• S, Vandana G; Pathipati, Srihari; Kumar, Uday; Nandagiri, Ashwin; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2025). Real-Time UAV Altitude Estimation and Data Transmission via mmWave Radar and Edge Computing, 2024 IEEE 21st India Council International Conference (INDICON). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798350391282. doi: 10.1109/INDICON63790.2024.10958360. Fulltekst i vitenarkiv Vis sammendrag

  This paper presents a novel approach for UAV altitude estimation and data transmission using a 60 GHz IWR6843 mmWave radar mounted on a micro-drone, coupled with a Raspberry Pi edge device. The radar, configured in a long-range mode, leverages its high accuracy in altitude measurement, surpassing the performance of traditional UAV altimeters. The radar altimeter data is processed on the Raspberry Pi and wirelessly transmitted to the cloud, from which it can be accessed by a ground station for real-time monitoring and analysis. To validate the accuracy of the radar-based altitude measurements, GPS data is simultaneously recorded on the UAV, serving as a ground truth reference. Experimental results demonstrate that the radar-based measurements closely match the GPS-derived altitudes, showcasing the effectiveness of the proposed system. This approach not only improves altitude estimation accuracy but also enhances the reliability of UAV operations in various environments. Potential applications of this system include precision agriculture, disaster management, and search and rescue operations, where accurate altitude data is critical for mission success. The integration of mmWave radar with edge computing and cloud-based data management opens new avenues for real-time UAV monitoring and autonomous navigation.

• Bethi, Pardhasaradhi; Pathipati, Srihari & Cenkeramaddi, Linga Reddy (2025). Improved GNSS Anti Spoofing by Integrating Attribute Information into Measurement to Measurement Association Framework, 2024 IEEE 21st India Council International Conference (INDICON). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798350391282. s. 1–6. doi: 10.1109/INDICON63790.2024.10958488. Fulltekst i vitenarkiv Vis sammendrag

  Global navigation satellite system (GNSS) is famous for providing position, navigation, and time (PNT) information at low cost anywhere globally. However, this GNSS is highly misled in the name of spoofing because of its readily available blueprints. This paper proposes to integrate the attribute information within the measurement to measurement association (M2MA) mathematical framework to improve the capabilities of GNSS anti-spoofing. The satellite location, received power, clock offset, and correlation are the attributes associated with the acquired pseudo ranges. This formulation assumes that authentic and spoofed GNSS signals are locked into the multi-correlator GNSS receiver and capable of estimating each signal's attributes. This process of using the attribute information facilitates screening authentic pseudoranges from spoofed pseudoranges. The results are evaluated at an epoch in a Monte Carlo sense with hit rate, dilution of precision (DOP), and mean square error (MSE) as metrics. The results reveal that the attribute information in M2MA drastically improves the anti-spoofing capabilities.

• Narayanan, Sivaranjini Perikamana; Jha, Ajit; Cenkeramaddi, Linga Reddy & Manikandan, M. Sabarimalai (2025). Deep Learning-Based Road Activity Image Classification for Intelligent Traffic Management, 2025 17th International Conference on Electronics, Computers and Artificial Intelligence (ECAI). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331533526. doi: 10.1109/ECAI65401.2025.11095446. Fulltekst i vitenarkiv Vis sammendrag

  Automatic vehicle classification is a crucial aspect of intelligent transportation systems for monitoring traffic flow, security enforcement, automated parking, driver assistance, and autonomous vehicle control. These systems face challenges such as limited computational resources, limited sensor range, and communication channels, which demand fast and lightweight algorithms for accurate and reliable performance. This paper presents an accurate, fast, and lightweight convolutional neural network (CNN)-based vehicle image classification model for intelligent traffic management applications using the urban scene database. The dataset includes multiple scenes captured during morning, evening, and night, presenting diverse lighting and environmental conditions. The performance of the proposed model is evaluated in terms of class-wise accuracy, overall accuracy, and computational resources. The proposed CNN model achieves an overall accuracy of 99.33% with an inference speed of 188 fps and a model size of 16.6 MB, and outperforms most of the benchmark pretrained models. The classification accuracy highlights the model's reliability and adaptability to real-world challenges such as variable lighting conditions.

• Pathipati, Srihari; S, Vandana G; Kumar, Uday; Nandagiri, Ashwin; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2025). FMCW Radar-Based Detection and Tracking of Drones Using DBSCAN Clustering and Extended Kalman Filter for Anti-Drone Defense Systems, 2024 IEEE 21st India Council International Conference (INDICON). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798350391282. doi: 10.1109/INDICON63790.2024.10958522. Fulltekst i vitenarkiv Vis sammendrag

  This paper aims to develop a radar-based detection and tracking system to mitigate the threats posed by drones, particularly those carrying malicious payloads. Due to the limitations of cameras in adverse weather and the high costs of LiDAR systems, radar technology is employed as a cost-effective alternative. The system utilizes 3D FFT followed by CA-CFAR for drone range-azimuth detections. The range-azimuth detections are clustered using DBSCAN. We simplified the extended target tracking problem into point target tracking based on the drone's size, with the dBSCAN cluster center acting as the measurement for the tracker. The tracking algorithm combines an Extended Kalman Filter (EKF) with Global Nearest Neighbor (GNN) data association. Experiments were conducted using a 77 GHz AWR1642 radar sensor to track a micro drone of hexacopter type within a range of 10m to 100m. The results demonstrated effective tracking capabilities with radar sensors successfully generating tracks. This study highlights the viability of radar-based systems for anti-drone applications, offering a practical solution for enhancing infrastructure security against potential drone threats.

• Shetty, Ashwitha; Goud, Madhusudhan; Kandasamy, Krishnamoorthy; Pathipati, Srihari; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2025). A High-Gain Half-Bow-Tie Antenna with Tapered Slots for Foliage Penetration Radar Application, 2024 IEEE 21st India Council International Conference (INDICON). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798350391282. s. 1–6. doi: 10.1109/INDICON63790.2024.10958326. Fulltekst i vitenarkiv Vis sammendrag

  In this study, a compact, low-profile, high-gain half-bow-tie antenna with tapered slots is designed specifically for foliage penetration radar (FOPEN) applications. The proposed antenna design begins by vertically bisecting a conventional bow-tie antenna, followed by the addition of two vertical metal strips to enhance the operational bandwidth. To achieve effective impedance matching across the desired frequency range, two horizontal conductive strips and a central balun circuit are integrated into the design. To further optimize the antenna for integration into aerial platforms, the gain and radiation pattern are refined by incorporating two additional metal strips strategically placed on the right side of the bow-tie structure. These design modifications result in a compact antenna with dimensions of 0.43λx 0.4λ x 0.005λ where λ is the wavelength corresponding to the lower frequency of operation. The optimized antenna achieves a realized gain of 6.5 dBi and an impedance bandwidth of 110 MHz, making it highly suitable for FOPEN applications. The enhanced gain, achieved with a single dielectric layer, demonstrates the potential of this design for efficient foliage penetration radar systems.

• Gouda, Bhabani Sankar; Panigrahi, Trilochan; Das, Sudhakar; Panda, Meenakshi & Cenkeramaddi, Linga Reddy (2025). Distributed fault detection in sparse wireless sensor networks utilizing simultaneous likelihood ratio statistics. Pervasive and Mobile Computing. ISSN 1574-1192. 110. doi: 10.1016/j.pmcj.2025.102043. Fulltekst i vitenarkiv Vis sammendrag

  Sensor nodes in wireless sensor networks (WSNs) for several remote applications are deployed in harsh environments and are coupled with low-cost components. Because of these factors, sensor nodes are becoming faulty, resulting in serious data inaccuracy in the network if not diagnosed in a timely manner. The current approaches to centralized or distributed fault detection algorithms are based on statistical methods or machine learning algorithms. Statistical methods require more data to achieve the desired detection accuracy and may be impractical for sparse networks. Machine learning approaches are computationally demanding. We know that the mean and variance of data from a faulty node differ from those from a healthy node. As a result, simultaneous likelihood ratio statistics are proposed here to determine the sensor node’s fault status in WSNs. The proposed hybrid method, in which the faulty status of the node connected to the anchor node is diagnosed by the anchor node, assumes that the anchor node has statistics for all connected nodes. During the diagnosis time, the simultaneous likelihood ratio statistics (SLRS) are computed using the data received by the anchor node over a specific time period. The fault status is determined by comparing the likelihood ratio to a predetermined threshold based on the tolerance limit. The algorithm’s performance is determined and compared to state-of-the-art algorithms using real-time measured data from the literature.

• Lata, Kusum; Gupta, Chitraksh & Cenkeramaddi, Linga Reddy (2025). A Cryptographic Framework for Secure Medical Imaging in Smart Healthcare Environments. Results in Engineering (RINENG). 27. doi: 10.1016/j.rineng.2025.106780. Fulltekst i vitenarkiv Vis sammendrag

  In the dynamic environment of innovative healthcare, securing medical images is indispensable because they are commonly sent and stored over digital media. This paper introduces a strong hybrid encryption scheme combining AES-128 (Advanced Encryption Standard) and RSA-1024 (Rivest-Shamir-Adleman) to improve security and the performance of medical image encryption. The new model employs AES-128 to encrypt images speedily and efficiently and RSA-1024 to securely share keys. Experimental verification was performed on a brain MRI dataset of 7023 images of 512 × 512 pixels. Encryption time ranged from 0.5 to 2.9 seconds depending on the AES mode, whereas decryption was significantly faster, from 0.1 to 0.9 seconds. The throughput analysis indicated encryption rates of up to 21.8 Mbps and decryption rates of 22.5 Mbps, with the best performance in CTR and OCB modes. Security was measured with MSE, PSNR, NPCR, UACI, histogram uniformity, correlation analysis, and entropy. PSNR was more than 50 dB in CFB and CBC modes, NPCR was more than 99.5 %, and entropy was close to the ideal value of 8 in encrypted images. The findings validate that the hybrid AES-RSA scheme attains an optimal level of confidentiality without loss of computational efficiency, and it is thus fit for real-time and secure medical image processing applications in contemporary healthcare systems.

• Dayal, Aveen; S., Shrusti; Cenkeramaddi, Linga Reddy; C., Krishna Mohan & Kumar, Abhinav (2025). Leveraging Mixture Alignment for Multi-Source Domain Adaptation. IEEE Transactions on Image Processing. ISSN 1057-7149. 34, s. 885–898. doi: 10.1109/TIP.2025.3532094. Fulltekst i vitenarkiv Vis sammendrag

  In a conventional Domain Adaptation (DA) setting, we only have one source and target domain, whereas, in many real-world applications, data is often collected from several related sources in different conditions. This has led to a more practical and challenging knowledge transfer problem called Multi-source Domain Adaptation (MDA). Several methodologies, such as prototype matching, explicit distance discrepancy, adversarial learning, etc., have been considered to tackle the MDA problem in recent years. Among them, the adversarial-based learning framework is a popular methodology for transferring knowledge from multiple sources to target domains using a min-max optimization strategy. Despite the advances in adversarial-based methods, several limitations exist, such as the need for a classifier-aware discrepancy metric to align the domains and the need to consider target samples’ consistency and semantic information while aligning the domains. To mitigate these issues, in this work, we propose a novel adversarial learning MDA algorithm, MDAMA, which aligns the target domain with a mixture distribution that consists of source domains. MDAMA uses margin-based discrepancy and augmented intermediate distributions to align the domains effectively. We also propose consistency of target samples by confidence thresholding and transfer of semantic information from multiple source domains to the augmented target domain to further improve the performance of the target domain. We extensively experiment with the MDAMA algorithm on popular real-world MDA datasets such as OfficeHome, Office31, PACS, Office-Caltech, and DomainNet. We evaluate the MDAMA model on these benchmark datasets and demonstrate top performance in all of them.

• Anirudhan, Jayarani Medakal; Manikandan, Manikandan Sabarimalai; Mula, Subrahmanyam & Cenkeramaddi, Linga Reddy (2025). Resource-Efficient Compressive PPG Sensing Encoders with Best Sensing Matrix and Sparse Recovery Algorithm for IoT Wearable and Edge Devices. IEEE Access. 13, s. 78571–78591. doi: 10.1109/ACCESS.2025.3562291. Fulltekst i vitenarkiv Vis sammendrag

  Resource-constrained wearable photoplethysmography (PPG) monitoring devices highly demand energy consumption reduction strategies to maximize the battery lifetime under continuous health monitoring applications. In this paper, we present a digital compressed PPG sensing (DCS-PPG) framework with a resource-efficient sensing matrix, sparse recovery algorithm, and best sparsifying matrix for wearable and edge PPG analysis devices in achieving a higher compression ratio (CR) with low waveform distortion and enabling data encryption or parameter extraction without reconstruction. We consider four sensing matrices (deterministic block binary diagonal (DBBD), random sparse fixed binary (RSFB), random Bernoulli (RBe), and random Gaussian (RG)), four sparsifying matrices (discrete cosine Ψ=[C] , a combination of discrete cosine and discrete sine Ψ=[C S] , a combination of impulse, discrete cosine and discrete sine Ψ=[I C S] and Haar wavelet), and six sparse recovery algorithms (orthogonal matching pursuit (OMP), approximate message passing (AMP), L1-minimization (L1-min), compressive sampling matching pursuit (CoSaMP), iterative hard thresholding (IHT) and iterative soft thresholding (IST)). On a wide variety of normal and abnormal PPG signals and five distortion measures, results show that the DCS-PPG framework based on the DBBD sensing matrix and OMP-based recovery algorithm with the sparsifying matrix Ψ=[I C S] achieved a lower percentage root mean square difference (PRD) of 1.86 ± 1.65 % at a higher CR of 3 with a lesser recovery time as compared to the results of other algorithms. Results show that the DBBD-based DCS-PPG sensing encoder can enable simultaneous data reduction and parameter extraction in the CS domain. The RSFB-based DCS-PPG sensing encoder can enable simultaneous data reduction and encryption with a lower recovery time and PRD of 2.55 ± 1.19 % at a CR of 2.4 using AMP-based reconstruction with the sparsifying matrix Ψ=[I C S] . We demonstrate the real-time feasibility of the DBBD and RSFB-based DCS-PPG encoder on an Arduino Due platform, while the Raspberry Pi platform is used to implement sparse recovery algorithms for reconstructing the compressed and encrypted data. The real-time implementation of the RSFB and AMP-based DCS-PPG encoding and decoding framework achieves a lower PRD of 2.93±1.10% and energy consumption (EC) of 430.26±30.3 mJ at a CR of 2.4 over a duration of two seconds, compared to other reconstruction algorithms such as OMP (PRD=5.84%) and L1-min (PRD=7.05%), while also achieving energy savings of ≈96.3% with respect to L1-min. The real-time evaluation results on the Arduino platform show that the digital compressed sensing (DCS) using sensing matrix DBBD achieved energy savings of ≈86.8% and a processing time of 295±30.5 μ s compared to the DCS using RSFB. The results show that selecting the optimal sensing matrix, recovery algorithm, and sparsifying matrix plays a vital role in improving energy efficiency with reduced latency.INDEX TERMS Compressed sensing, data compression and encryption, energy efficiency, Internet of Things, PPG signal analysis, sparse recovery algorithms, wearable health monitoring.

• Kumar, Ankit; Vellalacheruvu, Venkata Siva Sai Anil Kumar; Bethi, Pardhasaradhi; Koorapati, Sagar & Cenkeramaddi, Linga Reddy (2025). Marine Mammals Sound Classification Using a Lightweight CNN for Edge Computing Devices, 2025 IEEE 14th International Conference on Communication Systems and Network Technologies (CSNT). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331531935. s. 1182–1187. doi: 10.1109/CSNT64827.2025.10968906. Fulltekst i vitenarkiv
• Amarnath, Siddharth; Bethi, Pardhasaradhi; Vellalacheruvu, Venkata Siva Sai Anil Kumar; Koorapati, Sagar & Cenkeramaddi, Linga Reddy (2025). Indoor Sound Classification Using Lightweight CNN With Real-Time IoT Integration, 2025 IEEE 14th International Conference on Communication Systems and Network Technologies (CSNT). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331531935. s. 1149–1153. doi: 10.1109/CSNT64827.2025.10968499. Fulltekst i vitenarkiv Vis sammendrag

  This paper presents a lightweight Convolutional Neural Network (CNN) for indoor sound classification, specifically designed to identify ten distinct classes: knock on the door, mouse click, keyboard typing, wooden door squeaks, opening the can, washing machine, vacuum cleaner, alarm clock, clock ticking, and broken glass from [1]. The sound signals were captured using a microphone and transformed from the time domain to time-frequency images through Continuous Wavelet Transform (CWT). The proposed CNN, with a compact model size of 2.1 MB, was benchmarked against several well-known pre-trained models, including DenseNet, EfficientNet, InceptionNet, MobileNet, NASNet, ResNet, VGGNet, and Xception. Experimental results demonstrate that our lightweight CNN significantly outperformed these models, achieving a validation accuracy of 80% in a ten-fold cross-validation framework. The best performance among the pre-trained models was observed with ResNet101V2, which achieved only 62% validation accuracy in a ten-fold cross-validation framework with a considerably larger model size of 163 MB. Furthermore, the proposed model was successfully deployed in real-time on various IoT and processor modules, including the Raspberry Pi, NVIDIA A100, Tesla V100, and Intel Xeon Platinum, where it demonstrated superior efficiency, running at just 35 milliseconds per inference on the Raspberry Pi edge device. The lightweight CNN’s high accuracy and low latency make it an ideal candidate for applications in smart home systems, security monitoring, and real-time audio event detection on resource-constrained devices.

• K, Nair Akarsh; Coleri, Sinem; Sahoo, Jayakrushna; Cenkeramaddi, Linga Reddy & Deni, Raj Ebin (2025). Incentivized Federated Learning: A Survey. IEEE Transactions on Emerging Topics in Computational Intelligence (TETCI). 9(5), s. 3190–3209. doi: 10.1109/TETCI.2025.3547609. Fulltekst i vitenarkiv Vis sammendrag

  Federated Learning (FL) is an emerging learning paradigm facilitating privacy-preserving machine learning at a large scale without the need for training data aggregation. The existing literature on FL mainly focuses on optimizing learning to enhance convergence time and accuracy. However, the need for developing incentivisation strategies to motivate clients to actively participate in training is a highly relevant area of research within FL.Considering the relevance of the problem,the need for a comprehensive review of incentive mechanism development and working in FL is really high. Initially, this survey provides a basic introduction into FL and incentive mechanisms. Secondly, the fundamental aspects of incentivisation are covered, including formally defining incentivisation, the rationale behind incentivisation, benefits of incentivising clients, implementation methods, and aspects of bias. Next, different types of incentive distribution mechanisms, auction theory, contract theory, and game theory are introduced and discussed in detail, citing methodologies and limitations. Following this, the survey presents contribution evaluation mechanisms, discussing the detailed workings of Shapley value and reputation-based systems. Motivated by the inferences generated, studies on fairness during incentivisation are also presented. Lastly, some of the major considerations in incentivized FL related to cross-silo applications, free-riders, straggler mitigation, and blockchain systems are presented, followed by a discussion on open research problems and prospective research directions. Collectively, the study provides readers with a state-of-the-art and comprehensive perspective on incentivisation in FL, hoping to be a benchmark for future researchers studying fundamental aspects and applications across various domains.

• K, Naveen Kumar; C., Krishna Mohan; Cenkeramaddi, Linga Reddy & Awasthi, Navchetan (2025). Minimal data poisoning attack in federated learning for medical image classification: An attacker perspective. Artificial Intelligence in Medicine. ISSN 0933-3657. 159. doi: 10.1016/j.artmed.2024.103024. Fulltekst i vitenarkiv Vis sammendrag

  The privacy-sensitive nature of medical image data is often bounded by strict data sharing regulations that necessitate the need for novel modeling and analysis techniques. Federated learning (FL) enables multiple medical institutions to collectively train a deep neural network without sharing sensitive patient information. In addition, FL uses its collaborative approach to address challenges related to the scarcity and non-uniform distribution of heterogeneous medical domain data. Nevertheless, the data-opaque nature and distributed setup make FL susceptible to data poisoning attacks. There are diverse FL data poisoning attacks for classification models on natural image data in the literature. But their primary focus is on the impact of the attack and they do not consider the attack budget and attack visibility. The attack budget is essential for adversaries to optimize resource utilization in real-world scenarios, which determines the number of manipulations or perturbations they can apply. Simultaneously, attack visibility is crucial to ensure covert execution, allowing attackers to achieve their objectives without triggering detection mechanisms. Generally, an attacker’s aim is to create maximum attack impact with minimal resources and low visibility. So, considering these three entities can effectively comprehend the adversary’s perspective in designing an attack for real-world scenarios. Further, data poisoning attacks on medical images are challenging compared to natural images due to the subjective nature of medical data. Hence, we develop an attack with a low budget, low visibility, and high impact for medical image classification in FL. We propose a federated learning attention guided minimal attack (FL-AGMA), that uses class attention maps to identify specific medical image regions for perturbation. We introduce image distortion degree (IDD) as a metric to assess the attack budget. Also, we develop a feedback mechanism to regulate the attack coefficient for low attack visibility. Later, we optimize the attack budget by adaptively changing the IDD based on attack visibility. We extensively evaluate three large-scale datasets, namely, Covid-chestxray, Camelyon17, and HAM10000, covering three different data modalities. We observe that our FL-AGMA method has resulted in 44.49% less test accuracy with only 24% of IDD attack budget and lower attack visibility compared to the other attacks.

• Narayanan, Sivaranjini Perikamana; Manikandan, M. Sabarimalai & Cenkeramaddi, Linga Reddy (2025). YOLOv9-Based Human Face Detection and Counting Under Human-Animal Faces, Complex Imaging Environments, and Image Qualities. IEEE Access. 13, s. 129600–129637. doi: 10.1109/ACCESS.2025.3591247. Fulltekst i vitenarkiv Vis sammendrag

  Automatic human face detection and counting can play a vital role in the recognition and tracking of infant and adult faces in both outdoor and indoor human surveillance applications and facial-vital sign measurement. Despite the advancements in deep learning networks, accurate and reliable detection is still a challenging task in the presence of different kinds of objects, animal faces, and image characteristics. In this paper, we study the effectiveness of the YOLOv9-based face detection and counting method under skin color variations, different face sizes and number of faces, mixture of human and animal faces, image properties (brightness, contrast, illumination, glare), image qualities (image blurring, low-light facial images, lens flared images, grainy images taken at night, different types of noises), and pareidolia effects. The proposed method is trained and validated using the Wider Face database. In addition, we created image databases with different kinds of image qualities and image characteristics with the above-mentioned challenges. The YOLOv9-based face detection model achieves a precision of 86%, a recall of 62.8%, and a mean average precision of 70.8% at an inference time of 15.2 ms on the Wider Face database. Evaluation results demonstrate that the YOLOv9-based face counting outperforms most of the state-of-the-art face detection and people counting methods with a mean absolute error (MAE) of 3.36 and root mean square error (RMSE) of 22.38. The model was also deployed on the Raspberry Pi edge computing platform to study the real-time performance. The YOLOv9-based method achieves an MAE of 0.53-5.76 on the untrained infant database with Gaussian, salt and pepper, and speckle noises and an MAE of 0.43-2.87 on faces inside vehicles. The study further highlights the effectiveness of the YOLOv9 model in achieving promising face detection and counting results under a range of illumination and skin color variations. Evaluation results on a wide variety of animal faces and pareidolia-induced faces demonstrate more false positives due to the lack of contextual intelligence in the generation of deep-face models. Further, results show that the deep-face model detects artificial faces (statues, art, paintings, posters) if the model is deployed in uncontrolled face-based application environments. The performance of the model is degraded under different kinds of noise and blurred images. The results of this study highlight that performance can be improved by using noise-specific filtering techniques with optimal filtering parameters, but this requires the automatic identification of noise types and their corresponding parameters.

• Lata, Kusum; Singh, Prashant; Saini, Sandeep & Cenkeramaddi, Linga Reddy (2025). Privacy-preserving brain tumor detection using FPGA-accelerated deep learning on Kria KV260 for smart healthcare. Computer Methods and Programs in Biomedicine Update. 8. doi: 10.1016/j.cmpbup.2025.100205. Fulltekst i vitenarkiv Vis sammendrag

  Technological advancements in high-performance electronics have fueled the development of cutting-edge medical applications, leading to exponential growth in effective treatment and diagnostic solutions for various medical problems. Incorporating deep learning-based systems with medical imaging technologies has revolutionized the field of disease detection. Ensuring the security and privacy of patient’s health records is crucial to developing sophisticated medical imaging diagnostic applications. This paper presents a privacy-focused, vision-based approach for effective brain tumor detection using deep learning algorithms such as ResNet-18, ResNet-50, and InceptionV3, deployed on the KV260 board, which is based on Xilinx® Kria™ K26 System on Module (SOM) platform, a Zynq® UltraScale+ MPSoC. We have integrated the AES-128 cryptographic algorithm with the Password-Based Key Derivation Function 2 (PBKDF2) hashing algorithm to maintain patients' privacy in MRI scans. This ensures the protection of patient data on the server and data movement to and from external servers. The designed system is evaluated for performance by examining its technical metric parameters- accuracy, precision, F1 score, and Recall. Security parameters such as entropy, energy, contrast, and correlation are used to evaluate the security strength of the proposed system. Microsoft operating systems compatible web application is also developed while integrating the above-proposed system on the KV 260 FPGA board. This application can be used remotely to upload the MRI scans and get the prediction results quickly and accurately. Performance assessment shows that ResNet18 outperforms testing-related metric parameters and execution time on the KV260 FPGA board while keeping patient data confidential, making it an ideal edge-device implementation for real-time clinical use.

• Puduru, Vamsi Krishna; Bethi, Pardhasaradhi; Koorapati, Sagar & Cenkeramaddi, Linga Reddy (2025). Lightweight Deep Learning Model for Airborne Object Classification, 2025 IEEE 14th International Conference on Communication Systems and Network Technologies (CSNT). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331531935. s. 230–234. doi: 10.1109/CSNT64827.2025.10968603. Fulltekst i vitenarkiv Vis sammendrag

  Airborne object detection and classification play an important role in applications like military surveillance, border security, and public safety. The ability to accurately distinguish between airplanes, helicopters, drones, and birds is essential for monitoring and responding to potential aerial threats in real-time. This study leverages the AOD 4 dataset, which contains a comprehensive collection of 22,516 images spanning diverse environmental conditions. After organizing the dataset into four distinct classes, a custom lightweight Convolutional Neural Network (CNN) was developed to classify airborne objects. The proposed lightweight CNN achieves a remarkable accuracy of 97.12% and is able to outperform state-of-the-art pre-trained models in terms of accuracy and model size. On top of it, the model is also targeted at the various processors to unleash its potential to run in a real-time environment. The proposed model infers in 2.9 milliseconds on intel core i7 processor, which is a fourfold improvement compared to the MobileNetv3Small model. The findings from this study contribute to advancing the field of airborne object classification and pave the way for more robust real-world implementations.

• Viswanathuni, Venkata Raghava Shashank; Puranapanda, Sarada; Bethi, Pardhasaradhi; Koorapati, Sagar & Cenkeramaddi, Linga Reddy (2025). Lightweight CNN for Everyday Human Sound Classification on Edge Devices, 2025 IEEE 14th International Conference on Communication Systems and Network Technologies (CSNT). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331531935. s. 1171–1175. doi: 10.1109/CSNT64827.2025.10967858. Fulltekst i vitenarkiv Vis sammendrag

  This study presents a lightweight Convolutional Neural Network (CNN) that recognizes everyday sounds on IoT edge devices. Audio signals encompassing daily activities are captured via a microphone and transformed into time-frequency images using Continuous Wavelet Transform (CWT). The proposed CNN is optimized for low computational requirements and performs better than established pre-trained models such as DenseNet, EfficientNet, InceptionNet, MobileNet, NASNet, ResNet, VGGNet, and Xception. The model achieves an overall validation accuracy of 84 % in a ten-fold cross-validation framework, with a compact model size of only 854KB and an inference time of 7.3 milliseconds (ms) on a Raspberry Pi. In comparison, EfficientNetV2B1, the second-best performer, achieves comparable accuracy but with a significantly larger size of 27.1 MB and an inference time of 119.99 ms. The model’s inference time was evaluated on various devices, including Raspberry Pi, NVIDIA A100, Tesla V100, and Intel Xeon Platinum, demonstrating its efficiency across different platforms. The dataset includes everyday human sounds, such as crying, applause, breathing, coughing, footsteps, laughter, brushing teeth, snoring, and drinking water. The efficiency and accuracy of the proposed CNN make it ideal for real-time sound recognition in smart home automation, health monitoring, and ambient assisted living, where low latency and resource efficiency are critical.

• Singh, Nitin Pratap; Lata, Kusum & Cenkeramaddi, Linga Reddy (2025). Electron-phonon interactions and Helmholtz free energy in confined systems: Advancing low-dimensional superconductors for quantum technologies. Results in Physics. 73. doi: 10.1016/j.rinp.2025.108270. Fulltekst i vitenarkiv Vis sammendrag

  The thermodynamic and superconducting properties of low-dimensional superconductors are profoundly influenced by electron–phonon interactions, which play a critical role in determining the Helmholtz free energy. Therefore, the quantitative study of Helmholtz free energy using electron–phonon interactions in low-dimensional superconductors is essential to comprehending and improving superconducting materials, which are also the building blocks of developing quantum technologies. This article presents a quantitative analysis of our theoretically developed model with the impact of electron–phonon coupling on the Helmholtz free energy in low-dimensional superconductors. Using Green’s function methods and advanced many-body theory, we thoroughly examine the changes in the free energy landscape caused by changes in electron–phonon coupling strengths, temperature variation, effects of defects, and system dimensionality with the confinement phenomenon. The analysis also encompasses the behavior of electron–phonon interactions in quantum wells, offering insights into energy confinement and the effects of temperatures. The findings indicate notable changes in thermodynamic properties, especially in systems characterized by reduced dimensionality with the confinement of electrons in one dimension. The findings also underscore the significant impact of electron–phonon interactions on enhancing superconducting performance and establishing a framework for customizing superconducting material properties. The proposed quantitative model based on Helmholtz free energy and electron–phonon interactions provides a powerful framework for exploring the properties of low-dimensional superconductors, especially as they relate to practical applications in quantum technology.

• Yalagala, Sivanjaneyulu; Manikandan, M. Sabarimalai; Boppu, Srinivas & Cenkeramaddi, Linga Reddy (2025). Quality-Aware PPG-Based Blood Pressure Classification for Energy-Efficient Trustworthy BP Monitoring Devices With Reduced False Alarms. IEEE Access. 13, s. 132514–132536. doi: 10.1109/ACCESS.2025.3591977. Fulltekst i vitenarkiv Vis sammendrag

  Early detection of hypertension (HT) is crucial for significantly reducing the risk of serious health complications, which demands continuous or long-term blood pressure (BP) monitoring to ensure the timely management of hypertension. This paper proposes a new quality-aware photoplethysmogram (PPG) feature-based hypertension recognition framework to reduce false alarms and overall energy consumption by considering the affordable resource-constrained continuous BP monitoring devices. The proposed framework includes a high-pass filter (HPF), PPG signal quality assessment (PPG-SQA), PPG waveform feature extraction (FE), and BP classification (hypertension and non-hypertension (NHT)). In this paper, we present four SQA methods and nine machine learning (ML) based BP classification models, including logistic regression, decision tree, random forest, multilayer perceptron, k-nearest neighbours, XGBoost, AdaBoost, Bagged Tree, and one-dimensional convolutional neural network (1D-CNN). Four SQA methods are based on the average magnitude difference function (AMDF/(SQA-M1)) features and the AMDF features with the total number of zero-crossings present in raw/original (SQA-M2), derivative (SQA-M3), and smoothed derivative (SQA-M4) PPG segment features. On three standard databases, results showed that the SQA-M4 method outperforms the other three SQA methods with an overall false alarm reduction rate (FARR) of 50.8% in the case of motion artifact (MA) corrupted PPG signals and FARR of 91.0% in the case of additive white Gaussian noise (AWGN)-corrupted PPG signals. The quality-aware PPG feature extraction is presented to extract twelve features, which are combined with nine ML classifiers to explore the best model with higher recognition accuracy. Among the nine BP-classify models, the random forest (RF)-based BP model achieved the highest accuracy of 92.1% with a specificity of 93.8% for the BP model with SQA approach as compared to the BP model without SQA approach. On the implementation of the best model using the Raspberry Pi (R-Pi) platform, the quality-aware BP (QA-BP) model had FARR ranging from 32.4% to 62.2% for the NHT events and FARR ranging from 30.8% to 99.7% for the HT events in the case of random noise corrupted PPG signals, and the QA-BP model can reduce false alarms from 11.9% to 52.5% for NHT events and from 7.1% to 55.5% for HT events in the case of motion corrupted PPG signals. For the case of processing a 60s noisy PPG signal, the QA-BP method had an energy saving of 41.3% with display mode and an energy saving of 95.3% with the data transfer mode as compared to the BP classification method without the SQA approach. The proposed QA-BP method had energy saving of 8.1% - 85.6% for the 10% - 90% noisy portions in 24-hour recordings and has great potential in the development of affordable, energy-efficient cuffless BP monitoring devices with reduced false alarms.

• Viswanathuni, Venkata Raghava Shashank; Yakkati, Rakesh Reddy; Yeduri, Sreenivasa Reddy & Cenkeramaddi, Linga Reddy (2025). WiCNNAct: Wi-Fi-based Human Activity Recognition utilizing Deep Learning on the Edge Computing Devices. IEEE Access. 13, s. 104173–104183. doi: 10.1109/ACCESS.2025.3579649. Fulltekst i vitenarkiv Vis sammendrag

  Nowadays, human activity recognition plays an essential role in the application of human-computer interaction. Comprehensive systems, however, mostly rely on wearables, video cameras, and ambient sensors, which might be expensive and difficult to deploy or cause privacy issues. Wi-Fi-based wireless human sensing utilizes channel state information to detect human presence and movements, leveraging changes in signal properties caused by humans. This cost-effective approach utilizes existing Wi-Fi infrastructure and can operate in obscured or obstructed environments. However, signal interference and environmental conditions can affect accuracy and reliability. To address these challenges, a WiCNNAct for human activity recognition is proposed in this paper. The proposed approach utilizes the channel state information (CSI) measurements (complex values) from Wi-Fi and processes the different combinations of the real, imaginary, and absolute values using multi-channel 1D convolutional neural networks (1D-CNN). After conducting preliminary investigations, we validated various combinations of multi-channel 1D-CNNs and identified three methods for accurate activity recognition: a three-channel (real, imaginary, and absolute) setup, two channels (real and imaginary/real and absolute/imaginary and absolute), single channel (real/imaginary/absolute). The proposed three-channel Wi-Fi-Net underwent verification with 10-fold testing validation and achieved an overall accuracy of 98.29% with a standard deviation of 0.33%. The model is deployed on various edge computing devices, including Raspberry Pi, to assess real-time deployment feasibility. In this work, 1D-CNN is chosen based on its ability to automatically extract spatial and temporal features, reducing manual feature engineering. Experimental validation identifies the optimal setup for robust activity recognition.

• Sreedev, Sreekumar; SK, Shashank; Pathipati, Srihari; Nandagiri, Ashwin; S, Vandana G & Bethi, Pardhasaradhi [Vis alle 7 forfattere av denne artikkelen] (2025). FMCW Radar-based UAV Detection and Tracking using Transfer Learning, 2024 IEEE 21st India Council International Conference (INDICON). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798350391282. doi: 10.1109/INDICON63790.2024.10958556. Fulltekst i vitenarkiv Vis sammendrag

  This research investigation offers a novel method for monitoring and detecting unmanned aerial vehicles (UAVs) by combining transfer learning neural networks with Frequency Modulated Continuous Wave (FMCW) radar. The system utilizes a 60 GHz Texas Instruments IWR6843ISK radar with a DCA1000 board to capture raw radar signals, which are subsequently processed to generate range-angle heat maps. Ground truth data for UAV positioning is meticulously obtained using a dual GPS setup, where one GPS is stationed at the radar and the other is mounted on the UAV. The processed range-angle heat maps serve as the input for various transfer learning models, including DenseNet, InceptionV3, MobileNet, ResNet, and VGG, which are employed to compute the range data and angle data of the UAV. The results emphasize the potential of transfer learning in improving radar signal processing by demonstrating the effectiveness of these models in attaining accurate UAV detection and tracking. This approach is pivotal for applications requiring precise UAV monitoring, offering a robust solution for scenarios where traditional radar systems may fall short. The study underscores the advantages of leveraging transfer learning for improved radar-based UAV detection and sets the stage for future advancements in autonomous aerial monitoring and surveillance systems.

• SK, Shashank; Sreedev, Sreekumar; Murugan, Manjith; Pathipati, Srihari; S, Vandana G & Bethi, Pardhasaradhi [Vis alle 7 forfattere av denne artikkelen] (2025). Enhanced Intruder Detection Using mmWave Radar: A Spatiotemporal Clustering Approach for Robust Human Detection, 2024 IEEE 21st India Council International Conference (INDICON). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798350391282. doi: 10.1109/INDICON63790.2024.10958230. Fulltekst i vitenarkiv
• Sonny, Amala; Kumar, Abhinav & Cenkeramaddi, Linga Reddy (2024). Dynamic Targets Occupancy Status Detection utilizing mmWave Radar Sensor and Ensemble Machine Learning. IEEE Open Journal of the Industrial Electronics Society (OJ-IES). 5, s. 251–263. doi: 10.1109/OJIES.2024.3377012. Fulltekst i vitenarkiv
• Chilamkurthy, Naga Srinivasarao; Hakeem, Shaik Abdul; Tiwari, Satish K.; Ghosh, Anirban; Cenkeramaddi, Linga Reddy & Pandey, Om Jee (2024). Optimal Routing Protocol in LPWAN Using SWC: A Novel Reinforcement Learning Framework. IEEE Sensors Journal. ISSN 1530-437X. 24(9), s. 15607–15619. doi: 10.1109/JSEN.2024.3378463. Fulltekst i vitenarkiv
• Dayal, Aveen; Lalla, Rishabh; Cenkeramaddi, Linga Reddy; C., Krishna Mohan; Kumar, Abhinav & Balasubramanian, Vineeth N (2024). Improving Unsupervised Domain Adaptation: A Pseudo-Candidate Set Approach, European Conference on Computer Vision (ECCV 2024). Springer Nature. ISSN 9783031734113. s. 127–144. doi: 10.1007/978-3-031-73411-3_8. Fulltekst i vitenarkiv
• Yalagala, Sivanjaneyulu; Manikandan, M. Sabarimalai; Boppu, Srinivas & Cenkeramaddi, Linga Reddy (2024). Cuffless Blood Pressure Stratification and Hypertension Recognition Using Wearable PPG Sensor for Continuous Soldier Health Monitoring, 2024 IEEE International Workshop on Technologies for Defense and Security (TechDefense). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331505585. s. 98–103. doi: 10.1109/TechDefense63521.2024.10863395. Fulltekst i vitenarkiv
• Narayanan, Sivaranjini Perikamana; Manikandan, M. Sabarimalai & Cenkeramaddi, Linga Reddy (2024). Deep Learning Based Soldier Face Detection and Counting Method for Military Tactical Operations and Artificial Intelligence Powered Weapons, 2024 IEEE International Workshop on Technologies for Defense and Security (TechDefense). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331505585. s. 255–260. doi: 10.1109/TechDefense63521.2024.10863110. Fulltekst i vitenarkiv
• Vadakkan, Jomole Varghese; Manikandan, M. Sabarimalai & Cenkeramaddi, Linga Reddy (2024). Quality-Aware Rhythm-Aware Heart Rate Variability Analysis for Soldier Health and Wellness Monitoring Using Wearable ECG Sensing Devices, 2024 IEEE International Workshop on Technologies for Defense and Security (TechDefense). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331505585. s. 261–266. doi: 10.1109/TechDefense63521.2024.10863793. Fulltekst i vitenarkiv
• K, Nair Akarsh; Jayakrushna, Sahoo; Cenkeramaddi, Linga Reddy; Gaurav, Jaswal & Deni, Raj Ebin (2024). A streamlined FL framework for process optimisation via client clustering using rough c-means algorithm, CPSS '24: Proceedings of the 10th ACM Cyber-Physical System Security Workshop. Association for Computing Machinery (ACM). ISSN 9798400704208. s. 88–97. doi: 10.1145/3626205.3659151. Fulltekst i vitenarkiv
• Mondikathi, Chiranjeevi; Dhuli, V Sateeshkrishna; Enduri, Murali Krishna & Cenkeramaddi, Linga Reddy (2024). Identifying and Ranking of Best Influential Spreaders with Extended Clustering Coefficient Local Global Centrality Method. IEEE Access. 12, s. 52539–52554. doi: 10.1109/ACCESS.2024.3387745. Fulltekst i vitenarkiv
• Mohammad, Buran Basha; Dhuli, Venkata Sateeshkrishna; Enduri, Murali Krishna & Cenkeramaddi, Linga Reddy (2024). A Novel Convex Combination-Based Mixed Centrality Measure for Identification of Influential Nodes in Complex Networks. IEEE Access. 12, s. 123897–123920. doi: 10.1109/ACCESS.2024.3450296. Fulltekst i vitenarkiv
• Andrade, José Amendola Netto; Cenkeramaddi, Linga Reddy & Jha, Ajit (2024). Drone Landing and Reinforcement Learning: State-of-Art, Challenges and Opportunities. IEEE Open Journal of Intelligent Transportation Systems. 5, s. 520–539. doi: 10.1109/OJITS.2024.3444487. Fulltekst i vitenarkiv Vis sammendrag

  Unmanned aerial vehicles, and special multirotor drones, have shown great relevance in a plethora of missions that require high affordance, field of view, and precision. Their limited payload capacity and autonomy make its landing a crucial task. Despite many attempts in the literature to address drone landing, challenges and open gaps still exist. Reinforcement Learning has gained notoriety in a variety of control problems, with recent proposals for drone landing applications. This work aims to present a systematic literature review on works employing Deep Reinforcement Learning for multirotor drone landing in both static and dynamic platforms. It also revisits Reinforcement Learning Algorithms, the main frameworks and simulators adopted for specific landing operations. The comprehensive analysis performed on reviewed works revealed that there are important untackled challenges when it comes to wind disturbances, unpredictability of moving landing targets, sensor latency, and sim-to-real gap. Finally, we present our critical analysis of how recent state-of-the-art deep learning concepts can be combined with reinforcement learning to leverage the latter in addressing the open gaps in future works.

• Havås, Robin; Noori, Nadia Saad; Auby, Christian & Cenkeramaddi, Linga Reddy (2024). Optimized Unmanned Aerial Vehicle (UAV) Localization and Autonomous Navigation Stack for Tightly Closed Industrial Spaces, 2024 IEEE International Systems Conference (SysCon). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798350358803. doi: 10.1109/SysCon61195.2024.10553582. Fulltekst i vitenarkiv
• Mohammad, Buran Basha; Dhuli, V Sateeshkrishna; Enduri, Murali Krishna & Cenkeramaddi, Linga Reddy (2024). Convex Isolating Clustering Centrality to Discover the Influential Nodes in Large Scale Networks. IEEE Access. 12, s. 48404–48419. doi: 10.1109/ACCESS.2024.3384242. Fulltekst i vitenarkiv
• Gupta, Khushi Anil; Yeduri, Sreenivasa Reddy & Cenkeramaddi, Linga Reddy (2024). Breathing Rate Classification using Piezoresistive Sensor utilizing Continuous Wavelet Transform and Lightweight CNN. IEEE Access. 12, s. 50362–50375. doi: 10.1109/ACCESS.2024.3384983. Fulltekst i vitenarkiv
• A., Soumya; C., Krishna Mohan & Cenkeramaddi, Linga Reddy (2024). High Precision Single Shot Object Detection in Automotive Scenarios, Proceedings of the 19th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - (Volume 3). SciTePress. ISSN 9789897586798. s. 604–611. doi: 10.5220/0012383100003660. Fulltekst i vitenarkiv
• Dayal, Aveen; B, Vimal K; Cenkeramaddi, Linga Reddy; C., Krishna Mohan; Kumar, Abhinav & Balasubramanian, Vineeth N (2024). MADG: Margin-based Adversarial Learning for Domain Generalization, NIPS '23: Proceedings of the 37th International Conference on Neural Information Processing Systems. ACM Digital Library. ISSN 9781713899921. s. 58938–58952. Fulltekst i vitenarkiv
• Joshi, Anushka; Raman, Balasubramanian; C., Krishna Mohan & Cenkeramaddi, Linga Reddy (2024). A new machine learning approach for estimating shear wave velocity profile using borelog data. Soil Dynamics and Earthquake Engineering. ISSN 0267-7261. 177. doi: 10.1016/j.soildyn.2023.108424. Fulltekst i vitenarkiv
• Vadakkan, Jomole Varghese; Manikandan, M. Sabarimalai & Cenkeramaddi, Linga Reddy (2024). CNN Based Heart Rate Classification Using ECG Signal Without R-peak Detection for Rhythm-Aware Health and Emotion Monitoring, 2024 16th International Conference on Electronics, Computers and Artificial Intelligence (ECAI). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798350371154. s. 1–5. doi: 10.1109/ECAI61503.2024.10607442. Fulltekst i vitenarkiv

Se alle arbeider i NVA

• Kumar, Ankit; Vellalacheruvu, Venkata Siva Sai Anil Kumar; Bethi, Pardhasaradhi; Koorapati, Sagar & Cenkeramaddi, Linga Reddy (2025). Marine Mammals Sound Classification Using a Lightweight CNN for Edge Computing Devices. Fulltekst i vitenarkiv
• Manikandan, M. Sabarimalai; Karumudi, Harika & Cenkeramaddi, Linga Reddy (2025). Effectiveness of YOLOv8n Model Based Pothole Detection Under Noisy Conditions for Adverse Road Maintenance Environments. Fulltekst i vitenarkiv
• Bhardwaj, Jyotirmoy & Cenkeramaddi, Linga Reddy (2025). Scalable Online Anomaly Detection in Multi-Parameter Sensor Data Stream for Water Networks. Fulltekst i vitenarkiv
• Raj, Goutham; Mathew, Raji & Cenkeramaddi, Linga Reddy (2025). Aerial Vehicle Classification Using a Custom Lightweight CNN: A Deep Learning Approach. Fulltekst i vitenarkiv
• Narayanan, Sivaranjini Perikamana; Jha, Ajit; Cenkeramaddi, Linga Reddy & Manikandan, M. Sabarimalai (2025). Deep Learning-Based Road Activity Image Classification for Intelligent Traffic Management. Fulltekst i vitenarkiv
• Miryala, Neha; Yakkati, Rakesh Reddy; Puranapanda, Sarada; Yeduri, Sreenivasa Reddy & Cenkeramaddi, Linga Reddy (2025). Lightweight CNN for Real-Time Nature Sound Classification on Edge Devices. Fulltekst i vitenarkiv
• Ellore, Ram Chandra Kamavaram; Yakkati, Rakesh Reddy; Bethi, Pardhasaradhi; Zhou, Jing & Cenkeramaddi, Linga Reddy (2025). A Lightweight CNN for Real-Time Domestic Animal Sound Classification on IoT Devices Using Time-Frequency Imaging. Fulltekst i vitenarkiv
• Kattepogu, Navya; Vellalacheruvu, Venkata Siva Sai Anil Kumar; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2025). Agricultural Gas Classification Using Gas Sensor Arrays and PCA-Based Dimensionality Reduction. Fulltekst i vitenarkiv
• Utture, Rohil P.; Biju, Alan K.; Yarlagadda, Vishnu Srinivasa Murthy; Cenkeramaddi, Linga Reddy & Kumari, Nancy (2025). Weather Prediction based on the Cloud Images using Machine and Deep Learning Algorithms. Fulltekst i vitenarkiv
• Vellalacheruvu, Venkata Siva Sai Anil Kumar; Yarlagadda, Vishnu Srinivasa Murthy; V, Manoj Kumar; Vemulapalli, Sravani; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2025). An Ensemble Learning Approach to Covariance Intersection Fusion. Fulltekst i vitenarkiv
• Sirisilla, Sacharitha; Bethi, Pardhasaradhi; Zhou, Jing; Jha, Ajit & Cenkeramaddi, Linga Reddy (2025). Lightweight 2D CNN Fusion for Human Activity Recognition Using Multi-Transceiver CSI Data on Edge Devices. Fulltekst i vitenarkiv
• Amarnath, Siddharth; Bethi, Pardhasaradhi; Vellalacheruvu, Venkata Siva Sai Anil Kumar; Koorapati, Sagar & Cenkeramaddi, Linga Reddy (2025). Indoor Sound Classification Using Lightweight CNN With Real-Time IoT Integration. Fulltekst i vitenarkiv
• Viswanathuni, Venkata Raghava Shashank; Puranapanda, Sarada; Bethi, Pardhasaradhi; Koorapati, Sagar & Cenkeramaddi, Linga Reddy (2025). Lightweight CNN for Everyday Human Sound Classification on Edge Devices. Fulltekst i vitenarkiv
• Puduru, Vamsi Krishna; Bethi, Pardhasaradhi; Koorapati, Sagar & Cenkeramaddi, Linga Reddy (2025). Lightweight Deep Learning Model for Airborne Object Classification. Fulltekst i vitenarkiv
• Havås, Robin; Noori, Nadia Saad; Auby, Christian & Cenkeramaddi, Linga Reddy (2024). Optimized Unmanned Aerial Vehicle (UAV) Localization and Autonomous Navigation Stack for Tightly Closed Industrial Spaces. Fulltekst i vitenarkiv
• Yakkati, Rakesh Reddy; Boddu, Anudeep Bhaskar; Bethi, Pardhasaradhi; Yeduri, Sreenivasa Reddy & Cenkeramaddi, Linga Reddy (2024). Low Probability of Intercept Radar Signal Classification using Multi-Channel 1D-CNN. Fulltekst i vitenarkiv
• Narayanan, Sivaranjini Perikamana; Manikandan, M. Sabarimalai & Cenkeramaddi, Linga Reddy (2024). Deep Learning Based Soldier Face Detection and Counting Method for Military Tactical Operations and Artificial Intelligence Powered Weapons. Fulltekst i vitenarkiv
• Yalagala, Sivanjaneyulu; Manikandan, M. Sabarimalai; Boppu, Srinivas & Cenkeramaddi, Linga Reddy (2024). Cuffless Blood Pressure Stratification and Hypertension Recognition Using Wearable PPG Sensor for Continuous Soldier Health Monitoring. Fulltekst i vitenarkiv
• Vadakkan, Jomole Varghese; Manikandan, M. Sabarimalai & Cenkeramaddi, Linga Reddy (2024). Quality-Aware Rhythm-Aware Heart Rate Variability Analysis for Soldier Health and Wellness Monitoring Using Wearable ECG Sensing Devices. Fulltekst i vitenarkiv
• N, Dayananda B; Srivastava, Neetesh; G, Achala; Nandagiri, Ashwin; Pathipati, Srihari & Bethi, Pardhasaradhi [Vis alle 7 forfattere av denne artikkelen] (2024). Depth Information Fusion Using Radar-LiDAR-Camera Experimental Setup for ADAS Applications. Fulltekst i vitenarkiv
• P, Vishnu Prasad; S, Vandana G; Nandagiri, Ashwin; Pathipati, Srihari; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2024). Automotive Radar Signal Authentication via Correlation and Power Spectral Density. Fulltekst i vitenarkiv
• A, AnoopKumar K; N, Dayananda B; G, Achala; Pathipati, Srihari; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2024). Design of Barker-7 Radar Pulse Compressor Using DSP Architecture Minimization Techniques. Fulltekst i vitenarkiv
• Jena, Paramananda; Singh, Ashutosh; S, Vandana G; Nandagiri, Ashwin; Pathipati, Srihari & Bethi, Pardhasaradhi [Vis alle 7 forfattere av denne artikkelen] (2024). MIMO-SAR Image Reconstruction Experiment Using Back-Projection Algorithm With Automotive Radar for ADAS Applications. Fulltekst i vitenarkiv
• Alli, Sai Prakash; S, Vandana G; Nandagiri, Ashwin; Pathipati, Srihari; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2024). Cyber Attacking Active FMCW Radar Signal AoA Estimation Using Passive FMCW Radar for ADAS Applications. Fulltekst i vitenarkiv
• Saji, Sidharth; Manikandan, M. Sabarimalai; Zhou, Jing & Cenkeramaddi, Linga Reddy (2024). Underwater Debris Detection using Visual Images and YOLOv8n for Marine Pollution Monitoring. Fulltekst i vitenarkiv
• Nandam, Sreevidya; Nathala, Sai Sree; Dayal, Aveen; Manikandan, M. Sabarimalai; Zhou, Jing & Cenkeramaddi, Linga Reddy (2024). Classification of Anomalies in Industrial Machines utilizing Machine Sounds and Deep Learning. Fulltekst i vitenarkiv
• Vadakkan, Jomole Varghese; Manikandan, M. Sabarimalai & Cenkeramaddi, Linga Reddy (2024). Lightweight Time-Domain Statistical Parameter Based Muscle Artifacts Detection for Trustworthy Wearable ECG Analysis Devices. Fulltekst i vitenarkiv
• Pudi, Mohan; N, Dayananda B; G, Achala; Pathipati, Srihari; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2024). High-Speed Strassen Matrix Multiplication Accelerators for 2D Kalman Filter. Fulltekst i vitenarkiv
• Sreedev, Sreekumar; Shashank, S K; Vandana, G S; Pathipati, Srihari; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2024). Adult and Child Classification Using Automotive Radar for In-Cabin Monitoring. Fulltekst i vitenarkiv
• N., Sivaranjini P.; Manikandan, M. Sabarimalai & Cenkeramaddi, Linga Reddy (2024). “Spectrogram and LSTM Based Infant Cry Detection Method for Infant Wellness Monitoring Systems. Fulltekst i vitenarkiv
• N, Sivaranjini P; Manikandan, M. Sabarimalai & Cenkeramaddi, Linga Reddy (2024). Fast Autocorrelation Feature-Based Infant Cry Detector for Resource-Efficient Affordable Edge Cry Sound Analysis Systems. Fulltekst i vitenarkiv
• Soma, Anil Kumar; G, Achala; S, Vandana G; Pathipati, Srihari; Bethi, Pardhasaradhi & Cenkeramaddi, Linga Reddy (2024). Real-Time Radar Imaging With Time Domain Correlation and Doppler Beam Sharpening. Fulltekst i vitenarkiv
• Dayal, Aveen; Lalla, Rishabh; Cenkeramaddi, Linga Reddy; C., Krishna Mohan; Kumar, Abhinav & Balasubramanian, Vineeth N (2024). Improving Unsupervised Domain Adaptation: A Pseudo-Candidate Set Approach. Fulltekst i vitenarkiv
• V., Srinivasa Murthy Y.; Agrawal, Smiti; Malu, Reya; Cenkeramaddi, Linga Reddy & Madhusudhan, K. S. (2024). A Comparative Approach for Weapon Detection from Images using Deep Learning Algorithms. Fulltekst i vitenarkiv
• Salvi, Bhumika; Cenkeramaddi, Linga Reddy & Bethi, Pardhasaradhi (2025). Centralized FMCW Radar Point Cloud Clustering and Tracking Framework for Human Intruder Application. Universitetet i Agder. Fulltekst i vitenarkiv
• Muripala, Nehaa; Yeduri, Sreenivasa Reddy & Cenkeramaddi, Linga Reddy (2025). AUTOENCODER-DRIVEN ANOMALY DETECTION AND RSSI-BASED ANOMALOUS DEVICE LOCALIZATION IN LORAWAN. Universitetet i Agder. Fulltekst i vitenarkiv
• K, Koushik; Cenkeramaddi, Linga Reddy & Bethi, Pardhasaradhi (2025). Complementary Sensing-Based Adaptive Radar Target Tracking. Universitetet i Agder. Fulltekst i vitenarkiv
• Cenkeramaddi, Linga Reddy; Bethi, Pardhasaradhi & Viswanathuni, Venkata Raghava Shashank (2025). Multimodal Drone Classification: Integrating RGB, Thermal, and Radar Sensor Data Using Deep Learning. Universitetet i Agder. Fulltekst i vitenarkiv
• Cenkeramaddi, Linga Reddy; Jha, Ajit & Ellore, Ram Chandra Kamavaram (2025). Photonic laser sensor based multipoint motor bearing fault classification. Universitetet i Agder. Fulltekst i vitenarkiv
• Nelson, Wilson Ayyanthole; Cenkeramaddi, Linga Reddy; Jha, Ajit & Kumar, Abhinav (2024). Smart Sensing and Communications for UAVs/Drones. Universitetet i Agder. ISSN 9788284272023. Fulltekst i vitenarkiv Vis sammendrag

  Reliable sensing and energy-efficient communication are an integral part of unmanned aerial vehicle (UAV)-based applications. However, it is challenging to ensure reliable sensing under all weather conditions. Furthermore, enabling robust and energy-efficient communication of this sensed data for seamless autonomous operations is equally demanding. Low-complexity and lightweight algorithms play a vital role in this regard. As a result, the goal of this Ph.D. research project is to design, develop, and implement reliable sensing and energy-efficient communication techniques to provide seamless autonomous operations for UAVs and ground control stations. To achieve the intended goal, one of the objectives is directed toward developing novel lightweight low-complexity smart sensing algorithms that attain improved detection performance in adverse weather and lighting conditions too. In this regard, this research focuses on utilizing thermal, acoustic, and mmWave frequency-modulated continuous wave (FMCW) radar sensors with lightweight machine learning algorithms to obtain improved detection and localization performance. The sensed information along with telemetry data is required to be transmitted to other UAVs or ground control stations to facilitate efficient and secure coordination of autonomous UAVs. Hence, the secondary objective of this research addresses the challenge of achieving energy-efficient data transmission for a UAV-assisted wireless network. The research for attaining energy-efficient data transmission focuses on the design of a dynamic and adaptively optimized switching algorithm that leverages the distinct features of the various onboard communication modules. This thesis is a compendium of seven publications that are organized into five chapters. Overall the primary objective is to develop reliable smart sensing algorithms along with energy-efficient communication schemes to enable seamless autonomous UAV operations. The first chapter briefly introduces the various challenges dealt in the thesis and provides an overview of the thesis structure. Chapter 2 provides the necessary literary material to navigate the subsequent chapters. Chapter 3 discusses three sensing approaches using thermal, acoustic, and mmWave FMCW radar sensors to achieve improved detection and localization performance. In chapter 4, the notion of an adaptively optimized switching algorithm for energyefficient communication is conceived. Two approaches to form the hybrid communication switching network are discussed. Finally, chapter 5 concludes the Ph.D. dissertation by summarizing the proposed approaches and providing avenues for future research.

Se alle arbeider i NVA