International Journal of Ad Hoc and Ubiquitous Computing (20 papers in press)

Regular Issues

• Hybrid Cheetah and Artificial Rabbits Optimisation Algorithm for Micro Electro Mechanical System  
  by Sabitha Balasubramanian, Raffik Rasheed 
  Abstract: In this manuscript, a Hybrid Cheetah and Artificial Rabbits Optimization Algorithm fostered Digital Control Systems for Micro Electro Mechanical Systems Gyroscope(MEMS-GYS-HCO-ARO) is proposed. In this method the design Procedure starts by designing a closed loop control systems for Micro Electro Mechanical Systems Gyroscope and loop parameters are optimized using Hybrid algorithms. First the loop parameters are optimised by the help of HCO and ARO. Then Dwarf Mongoose Optimisation Algorithm and LMS demodulator utilized for demodulate noise signal. The performance of the proposed MEMS-GYS-HCO-ARO method attains amplitude analysis of 39.25%, 33.36% and 32.99% of lower amplitude compared with existing methods, like a design of micro electro mechanical systems gyroscope using Hybrid Optimisation Algorithm (MEMS-GYS-HOA), Digital Control and Readout of MEMS Gyroscope Using Second-Order Sliding Mode Control(MEMS-GYS-SOSMC) and A Digital Closed Loop Sense MEMS Disk Resonator Gyroscope Circuit Design Based on Integrated Analog Front-end (DCLS-MEMS-IAF) respectively.
  Keywords: MEMS gyroscope; closed-loop system; control systems; hybrid cheetah and Artificial Rabbits algorithm; Monte Carlo analysis; demodulator.
  DOI: 10.1504/IJAHUC.2024.10068845
   
  
• Visual Safety in the Meta-Universe using VR Technology  
  by Shan Wu, Hongjun Chen, Ningning Wei 
  Abstract: This work proposes the VR-based visual safety model (VR-VSM) with a meta-universe, which creates highly realistic virtual worlds that can be used within the construction ecosystem to enhance workers hazard recognition skills. The proposed VR-VSM incorporates a meta-universe eye-tracker to identify potential dangers on digital building sites, whilst a brainwave sensor captures the workers movement skills. Data collected from the eye tracker and brain sensor may be analysed and classified to reveal patterns in eye movement and brain processes, which may result in the proposed VR-VSM. The offered approach may elevate existing safety training programs to a whole new level by offering distinct comments to employees. The experimental results show the VR-VSM achieves a high recognition rate of 97.3%, a performance ratio of 98.5%, an accuracy ratio of 97.1% and a lower error rate of 27.2% compared with other methods.
  Keywords: Virtual Reality (VR); Meta-universe; Construction; Visual Safety.
  DOI: 10.1504/IJAHUC.2025.10070369
   
  
• ACIRO:Adaptive Clustering and Intelligent Routing Optimisation in Software Defined Vehicular Networks  
  by Sajithabegam A, T. Menakadevi 
  Abstract: To improve communication efficiency and reliability in Software-Defined Vehicular Networks (SDVN) utilizing density-based clustering and reinforcement learning approaches The proposed approach, Adaptive Clustering and Intelligent Routing Optimidation (ACIRO), employs a central controller to optimise cluster structure and routing decisions The optimization is based on density and Entropy-based Advantage Actor-Critic (EBA-AC) models A novel Density-based Fuzzy C-Means (DB-FCM) algorithm is introduced to optimise cluster formation, while cluster head selection criteria include distance, energy, and density parameters Furthermore, EBA-AC is utilised to determine the optimal path considering link availability, distance, and density parameters The proposed ACIRO approach is implemented and evaluated using the Mininet-WiFi emulator Comparative analysis with existing methods demonstrates improvements in cluster and network performance metrics, including cluster stability, throughput, packet delivery ratio, and end-to-end delay The ACIRO approach offers an effective solution for optimising cluster-based routing in dynamic vehicular environments
  Keywords: Density; Energy; Distance; Clustering; Routing; Optimization; VANET; SDVN.
  DOI: 10.1504/IJAHUC.2025.10070872
   
  
• Organ Transplantation using Federated Blockchain based Healthcare System  
  by Aman Alam, Abhishek Yadav, Anubhav Kumar, Aman Chaudhari, J. Sathish Kumar 
  Abstract: Organ transplantation is a critical aspect of modern healthcare, offering life-saving treatments to patients suffering from organ failure. However, the process of organ donation, allocation, and transplantation is complex, often fraught with challenges such as organ shortage, inefficient allocation mechanisms, and lack of transparency. Blockchain technology has emerged as a promising solution to address these challenges. By providing a decentralized and immutable ledger, blockchain can improve the efficiency, transparency, and security of organ transplantation processes. It can enable better tracking of organ donations, facilitate fair and transparent allocation of organs, and enhance trust among stakeholders in the healthcare ecosystem. In this paper, the aim is to develop a blockchain-based healthcare system for organ transplantation which leads to ultimately improving patient outcomes and saving more lives. Further, the proposed methods based on the blockchain has been implemented and validated with various parameters using hyperledger fabric network.
  Keywords: Organ Donation; Organ Transplantation System; Blockchain Technology; Hyperledger Fabric; Smart Contracts.
  DOI: 10.1504/IJAHUC.2025.10070875
   
  
• AI-Based Behaviour-Aware Path Risk Assessment for Wheelchair-Centric Smart Assistive Devices  
  by Chiao-Wen Kao, Shih-Tung Wang, Chi-Sheng Huang 
  Abstract: Mobility challenges in aging societies underscore the need for safer navigation solutions for wheelchair users, who often face hazards like uneven surfaces and indistinct curbs. Addressing these concerns, this study introduces an AI-based behavior-aware path risk assessment system designed for wheelchair-centric assistive devices. The system combines object detection and semantic segmentation to identify critical road features, such as sidewalks, curbs, and lanes. A Path Risk Assessment Module evaluates path safety using a scoring algorithm incorporating regional weighting and adjusting to varying environmental conditions. Integrating Internet of Behavior principles, the system adapts dynamically to user behaviors and conditions, offering personalized risk assessments. Tested on a custom dataset, the system demonstrated accurate real-time evaluations, with horizontal camera orientations excelling in dynamic scenarios and vertical setups suited for detailed analysis. These findings highlight its potential to enhance safety and mobility, with future work focusing on dynamic obstacles and multi-view integration for broader applications.
  Keywords: Behavior-Aware Systems; Smart Assistive Devices; Path Risk Assessment; Semantic Segmentation; Wheelchair-Centric.
  DOI: 10.1504/IJAHUC.2025.10070877
   
  
• Performance Analysis of Multi-RIS System with RIS Selection  
  by Caslav Stefanovic, Marko Smilic, Danijel Djosic 
  Abstract: In this research study, we delve into the performance analysis of systems that incorporate multi re-configurable intelligent surfaces (M-RIS), where RIS selection takes place at the destination. Specifically, our focus lies on deriving novel expressions for various end-to-end (e2e) metrics, including the probability density function (pSC (FSC s,L ), outage probability (PSC s,L ), cumulative distribution function s,L ), average bit error rate (PSC e,L ), and ergodic channel capacity (CSC s,L). These evaluations are conducted within the context of an M-RIS enabled system, where the RIS with the highest e2e signal-to-noise ratio (SNR) is selected, assuming independent Rayleigh fading channels. Furthermore, we present graphical representations and numerical evaluations of the analytically derived results for the aforementioned performance measures. This enables us to explore the impact of varying the number of RIS structures and the number of RIS elements on the systems overall performance
  Keywords: 6G; Multi-RIS enabled system; Performance analysis; Selection combining.
  DOI: 10.1504/IJAHUC.2025.10071084
   
  
• Tumour GAN-Based Augmentation and Hybrid Deep Learning Model for Classification of Brain Tumour using MRI Images  
  by Venkatesh Bhandage, Nijaguna G. S, Nagaraj Bhat, Manjunath G. Asuti, Praveen S. Challagidad 
  Abstract: Most existing methods for BT classification are not accurate enough due to a lack of labelled data. Therefore, effective BT classification is essential for improving the survival rates and enhancing the overall well-being of patients. The major objective of this research is to introduce a hybrid network Spinal_LeNet for BT classification. Initially, the input Magnetic Resonance Image (MRI) images undergo preprocessing. Subsequently, the image is segmented using squeeze M-SegNet. Thereafter, the image augmentation is done by the Tumor Generative Adversarial Network (TumorGAN). After the image augmentation, the extraction of features is employed to mine the significant morphological features like size and volume and histogram features such as magnitude, dispersion, asymmetry, flatness, and randomness. Finally, BT is classified using Spinal_LeNet, which is obtained by merging SpinalNet and LeNet. The devised model provides better values of Positive Predictive Value of 90.03%, and sensitivity of 92.92% compared to the existing methods.
  Keywords: MRI Images; Brain Tumour; TumorGAN; segmentation; Tumour classification.
  DOI: 10.1504/IJAHUC.2025.10071143
   
  
• Deep Convolutional Neuron Attention Forward Harmonic Network for Brain Tumour Detection and Classification using MRI Image  
  by Kalyani Ashok Bedekar, Anupama Sanjay Awati 
  Abstract: In recent years, brain tumours have been the deadliest brain disorder that occurs because of the collection or mass of aberrant brain tissues in the human brain. This research proposes a novel Deep Learning (DL) model, Deep Convolutional Neuron Attention Forward Harmonic Network (DCNasFH-Net), for accurate detection and classification of brain tumours. At first, the input Magnetic Resonance Imaging (MRI) image is pre-processed by utilising a high boost filter, and the Attention Gate ResU-Net (AGResU-Net) with a hybrid loss function is used to segment the interested brain tumour region. Following this, the features are extracted using the Spatial Grey-Level Dependence Matrix (SGLDM) and statistical features. Finally, the brain tumour is detected and classified effectively by utilising DCNasFH-Net. Moreover, the DCNasFH-Net attained effectual experimental outcomes with an accuracy of 93.01%, a True Negative Rate (TNR) of 91.89%, and a True Positive Rate (TPR) of 94%.
  Keywords: Attention Gate ResU-Net; Deep Convolutional Neuron Attention Forward Harmonic Network; Deep Convolutional Neural Network; Neural Architecture Search Network; High Boost Filter.
  DOI: 10.1504/IJAHUC.2025.10071365
   
  
• Health Insurance System Using Federated Blockchain Technology  
  by Pranil Tripathi, Prajwal Moon, Prakhar Mishra, Mohan Singh, J. Sathish Kumar 
  Abstract: This paper presents a healthcare insurance system using Hyperledger Fabric, a permissioned blockchain framework, to address key challenges such as data security, fraud, inefficiency, and lack of transparency. By utilising smart contracts, the system automates claims processing, reducing administrative burdens and minimising fraud. Hyperledger Fabric's modular architecture allows for customizable consensus mechanisms, identity management, and data privacy controls. The system improves collaboration among stakeholders, including healthcare providers, insurers, and patients/clients, by maintaining a single, shared ledger that streamlines verification processes and improves auditability with real-time updates. This blockchain-based solution significantly enhances operational efficiency, cost reduction, and trustworthiness in the healthcare insurance ecosystem, offering a promising framework for future advancements in the industry.
  Keywords: Federated Blockchain; Health Insurance System; Smart Contracts; Hyperledger Fabric.
  DOI: 10.1504/IJAHUC.2025.10071458
   
  
• Efficient SWIPT Transmission Optimisation in Decentralised FD-NOMA-V2X Systems  
  by Peiying Zhang, Shengpeng Chen, Yi Wang, Di Zhang, Lizhuang Tan, Jian Wang 
  Abstract: This paper investigates the effectiveness of Simultaneous Wireless Information and Power Transfer (SWIPT) in a decentralized Full-Duplex Non-Orthogonal Multiple Access Vehicle-to-Everything (FD-NOMA-V2X) system. The proposed model enables direct communication between devices, reducing the burden on cellular networks in vehicular communication. Additionally, the FD-NOMA technology further enhances the system's spectral efficiency (SE). Considering the limited battery capacity of devices, a SWIPT transmission framework is introduced based on the system model, formulated as an optimization problem. First, the optimal solution to this problem is derived using the CVX tool. To increase the model’s practical applicability, a suboptimal solution with lower computational complexity than the CVX-based method is also proposed, based on an analytical approach. Simulation results demonstrate that as the energy harvesting threshold increases, the power allocation coefficients in both methods converge to a common point. Furthermore, the analytical method significantly reduces computation time compared to the CVX tool.
  Keywords: Caching and sharing mechanism; 5G; Internet of things; Energy efficiency.
  DOI: 10.1504/IJAHUC.2025.10072018
   
  
• Intelligent Task Offloading in Vehicular Edge Computing using Federated Learning and Kolmogorov Arnold Networks (FL-KAN)  
  by Shabariram C. P, Shanthi N, Alisha Shinaz, Lakshana Ranganathan 
  Abstract: In recent times, Vehicular Edge Computing has risen to become a crucial paradigm to handle the increasing computational demands of Smart Transportation Systems. However, multiple challenges exist including the dynamic nature of vehicular environments, resource constraints of edge servers, and high privacy for data. To address these challenges, an intelligent task offloading framework integrates Kolmogorov Arnold Networks (KAN) and Federated Learning (FL) is proposed. The framework works within a multi-tier architecture, using two different modes Vehicle-to-Roadside Unit and Vehicle-to-Infrastructure to offload tasks. The KAN is trained with historical task data allows efficient task offloading, while FL ensures privacy and scalability across the local and global model. The experimental simulations were performed to optimize parameters like service latency, energy consumption and transfer delay. The results depict the proposed approach exceeds the existing algorithms such as Machine Learning, Selective Model Aggregation, Reinforcement Learning, Deterministic Policy Gradient by 18%, 25%, 13% and 21%.
  Keywords: Vehicular Edge Computing; Task offloading; Transport System; Kolmogorov Arnold Networks; Federated Learning.
  DOI: 10.1504/IJAHUC.2025.10072307
   
  
• Large Models for Fatigue Driving Detection in Future Vehicles: a Predictive Analytics with Ensemble Neural Networks  
  by Guangwu Hu, Tan Chen, Wei Liu, Yan Li, Dandan Hu 
  Abstract: In this paper, we introduce a system for fatigue driving detection via analyzing spatial-temporal electroencephalogram (EEG) features and employing ensemble neural networks. We extract time-domain EEG features and spatial-domain metric features related to the driving process from EEG data and brain functional network (BFN) data over time. To effectively utilize these features, we develop a feature contribution algorithm that assigns varying contribution coefficients to the time-domain EEG features and the spatial-domain BFN features based on their relationship with the target class. Subsequently, we utilize two sets of weighted features as inputs for two different neural networks: The long short-term memory (LSTM) network and the pseudo three-dimensional convolutional neural network, allowing to harness the complementary information of spatial-temporal EEG features and the data processing capabilities of these two neural network algorithms. Experimental results corroborate the superior performance of the proposed ensemble neural network model compared with the state-of-the-art methods.
  Keywords: Large models; Fatigue driving detection; Ensemble neural networks; LSTM; EEG.
  DOI: 10.1504/IJAHUC.2025.10072543
   
  
• A Novel Approach for Brain Thoughts to Text Conversion using Average Golden Search Optimisation and Deep Recurrent Neural Network (AGSO-DRNN)  
  by Adnan Ahmed, Waseemullah - 
  Abstract: This research proposes a brain signals-to-text converting framework using optimization-enabled deep learning. Here, conversion is carried out using the brain Electroencephalogram (EEG) signals based on the neural activity corresponding to attempted imaginary statements/questions. The EEG signal is subjected to various processes, like Signal pre-processing, Signal segmentation, Feature extraction, Character recognition, and Language modelling. Here, a Deep Recurrent Neural Network (DRNN) is employed to recognize the words in the EEG signals based on the extracted features, and the DRNN is trained using the Average Golden Search Optimization (AGSO) algorithm. Additionally, the successive words or characters are estimated by the usage of a language modelling with the Gaussian Mixture Model (GMM). The experimental validation of the proposed AGSO-DRNN is compared with other conventional techniques and the proposed model attained a maximum f-measure, MSE, precision, and sensitivity, recall, text conversion accuracy of 0.888, 0.005, 0.917, 0.891 and 0.873 respectively.
  Keywords: Brain-Computer Interface (BCI); Deep Recurrent Neural Network; Average Golden Search Optimization; Gaussian Mixture Model; Electroencephalogram.
  DOI: 10.1504/IJAHUC.2025.10072563
   
  
• Modelling of Reconfigurable Intelligent Surfaces (RIS) with Adaptive Power and Thermal Energy Harvesting  
  by Abdulrahman Alghamdi 
  Abstract: Reconfigurable Intelligent Surfaces (RIS) have emerged as transformative elements in the evolution of wireless communication systems, particularly within Cognitive Radio Networks (CRNs) These surfaces enable dynamic control over the wireless environment, allowing for enhanced signal propagation, interference mitigation, and spectral efficiency In this work, we explore the integration of RIS with adaptive transmit power control and thermal energy harvesting, where energy is harvested from ambient temperature gradients This approach enables the RIS to operate sustainably by converting environmental thermal differences such as those found in industrial or natural settings into usable electrical energy. We investigate the performance of RIS-enabled CRNs under various system configurations through comprehensive analytical modelling and simulation, we demonstrate that the proposed RIS-assisted CRN with thermal energy harvesting significantly outperforms conventional systems, offering notable improvements in throughput. These results underscore the potential of RIS, powered by ambient thermal gradients.
  Keywords: Cognitive Radio Networks (CRN); thermal energy harvesting; adaptive transmit power; Rayleigh channels.
  DOI: 10.1504/IJAHUC.2025.10072666
   
  
• Efficient Hardware Implementation of KLEIN Cipher with Power Analysis Attack Resistance  
  by Parthasarathy R, Saravanan Paramasivam 
  Abstract: Lightweight hardware implementation of block ciphers is critical to ensure the safety of data in resource-constrained environments. In this work, two iterative architectures are designed for Klein-64/80/96 and implemented in both the FPGA devices and the ASIC platform. The first architecture is of 64-bit datapath size, utilized a minimum number of slices in FPGAs and exhibited very good throughput compared to the existing works in the literature. The second architecture utilised a 32-bit datapath with a minimum number of S-boxes and mixcolumns equation which resulted in an area of 1451 GE in the ASIC platform, determined using the Cadence tool and 45 nm technology library gpdk045. A Correlation power analysis attack is mounted on the FPGA implementation of KLEIN-64 and successfully extracted the 64-bit key. Threshold implementation is defined as a countermeasure to mitigate the power analysis attack and secure the hardware implementation against the power analysis attack.
  Keywords: KLEIN Cipher; Lightweight Cryptography; Low cost implementation; Power analysis attack; Threshold implementation.
  DOI: 10.1504/IJAHUC.2025.10073177
   
  
• Energy Optimisation Model Based On Ant-Mating Optimisation for Collaborative Fog Computing In Internet Of Drones  
  by Dillon Leong Lon Zan, Muhammad Umair Munir, Rafidah Binti Md Noor, Ismail Ahmedy, Rami Sihwail, Husam Ahmed Al Hamad 
  Abstract: Rapid advancements in drone technology have facilitated their deployment in various applications, including aerial surveys through the Internet of Drones (IoD). Given that IoD operations are resource-intensive, efficient management is crucial to avoid overloading drones and reducing power consumption. This study introduces an energy-aware task scheduling model for IoD operations within fog networks, enhancing resource allocation by offloading tasks from drones to fog devices. This method optimises drone data handling and significantly reduces IoD energy usage. We implemented the model in a simulated environment using an augmented version of iFogSim, with a focus on minimising energy expenditure in fog devices. Our findings reveal that the proposed Ant-Mating Optimisation (AMO) algorithm markedly outperforms traditional genetic algorithms in efficiency, presenting a viable solution for energy optimisation in IoD systems.
  Keywords: Internet of Drones (IoD); Fog Computing; Energy-aware Task Scheduling; Resource Allocation; Power Consumption Optimisation.
  DOI: 10.1504/IJAHUC.2025.10073276
   
  
• An Efficient Lung Nodule Detection Model from 3D CT Images with Residual Bidirectional Long Short Term Memory and Adaptive Segmentation Schemes  
  by Maheswari S, Suresh S 
  Abstract: Presently, numerous advancements are achieved in biomedical imaging techniques to provide huge openings in the healthcare industries. Therefore, an effective lung nodule identification mechanism by a deep learning technique is implemented in this paper. Initially, the Three-Dimensional (3D) Computed Tomography (CT) images are acquired from standard database. Then, the gathered 3D CT images are offered as input to the segmentation region in which the Adaptive 3D Trans-DenseUNet (A-3D-TransDUNet) model is employed and the parameters are tuned with Fitness-based Cicada Swarm Optimization (FCSO) algorithm. Then, the acquired segmented lung nodule images are given to the deep learning based detection framework. Here, the Multiscale 3D DenseNet fused with “Residual Bidirectional Long Short Term Memory (M3D-DNet-RBi-LSTM)” is used as the detection framework. The final detected lung nodules are obtained from the implemented M3D-DNet-RBi-LSTM model. Various experimentations are executed to validate the efficacy rate provided by the suggested deep learning-aided lung nodule detection framework.
  Keywords: Lung Nodule Detection; Adaptive 3Dimensional Trans-DenseUNet; Multiscale 3D DenseNet fused with Residual Bidirectional Long Short Term Memory; Fitness-based Cicada Swarm Optimization.
  DOI: 10.1504/IJAHUC.2025.10073286
   
  
• Malware Analysis and Detection using Optimized Dynamic Path-Controllable Deep Unfolding Neural Network in PE Files using YARA Rules  
  by Vivek Kumar Anand, Sanjay Kumar Biswash 
  Abstract: The rapid evolution of malware necessitates an optimised approach for effective detection. This study proposes malware analysis and detection using an optimised dynamic path-controllable deep unfolding neural network in PE files with YARA rules (DPCDUNN-MA-PEF). Initially, PE file data undergoes pre-processing using the generalised multi-kernel maximum correntropy Kalman filter (GMKCKL) to remove redundancy. Relevant features are extracted using the multi-objective matched synchrosqueezing chirplet transform (MOMSSCT). The extracted features are analysed using the dynamic path-controllable deep unfolding network (DPCDUN) for malware classification. To enhance detection accuracy, the hunger games search optimisation algorithm (HGSOA) optimises DPCDUN parameters. The proposed method is implemented in Python and examined using performance metrics such as accuracy, precision, recall, F1-score, error rate, ROC, computational time. Experimental results show superior performance, with up to 29.28% higher F1-score compared to YARA-FH-FRMA, DGL-IDA-MD, and ERMD-CFT-DNN.
  Keywords: Dynamic Path-Controllable Deep Unfolding Network; Generalized Multi-kernel Maximum CorrentropyKalman Filter,Hunger Games Search Optimization and Multi-objective Matched SynchrosqueezingChirplet Transf.
  DOI: 10.1504/IJAHUC.2025.10073287
   
  
• Learning Boosting: Enhancing Predictive Modelling in Blended Learning Environments  
  by Zhihong Xu, Chin-Hwa Kuo, Chih-Yung Chang, Jinjun Liu, Chunyan Yu 
  Abstract: Accurate prediction of student learning outcomes is critical for early intervention and instructional decision-making in blended learning environments. This study proposes learning boosting, a structure-enhanced predictive framework integrating community-aware Louvain clustering with a gradient boosting classification. Student activity graphs are clustered to detect latent behavioural communities, and the resulting structural labels are embedded as features for final prediction. Experiments on real-world data from a blended learning course with 102 students evaluate the method under multiple classification granularities, data modalities, and clustering strategies. Results show that learning boosting consistently outperforms 11 baseline models, achieving an F1-score of 0.892, AUC of 0.883, and recall of 0.903 in the three-class task. Ablation studies confirm the complementary benefits of structural feature extraction and clustering. The findings demonstrate that combining graph-based structural modelling with boosting classifiers offers a robust and interpretable approach to learning analytics, especially in sparse and multimodal conditions.
  Keywords: Learning Analytics; Blended Learning; Learning Outcome Prediction; Graph-Based Modeling; Gradient Boosting.
  DOI: 10.1504/IJAHUC.2025.10073446
   
  
• SCS-DSSS: a Compressive Sensing and Deep Semantic Segmentation Framework for Cooperative Spectrum Sensing in 5G Cognitive Radio Networks  
  by Jebamalar Leavine E, Azhagu Subha M 
  Abstract: The rapid proliferation of wireless devices and the advent of advanced 5G New Radio (NR) standards have intensified the need for efficient spectrum utilization. Cognitive Radio (CR) with Dynamic Spectrum Access (DSA) offers a solution, yet existing wideband sensing methods face a trade-off between accuracy and data overhead. This work proposes SCS-DSSS, an end-to-end framework integrating Spectrogram Compressive Sensing (SCS) with DeepLabv3+ semantic segmentation using a ResNet-50 backbone. This framework system performs joint spectral occupancy detection with signal-type classification (5G-NR/LTE) directly from compressively sampled spectrograms, reducing sensing load without sacrificing accuracy. Unlike traditional approaches, SCS-DSSS eliminates the need for handcrafted features or prior signal knowledge. Experimental results show a sub-band detection accuracy of 96.85%, with notably low false negative rates, even under 25% compression. The framework demonstrates strong robustness in low-SNR settings and is well-suited for cooperative CRNs in edge-deployed IoT and future 6G systems, enabling scalable and intelligent DSA.
  Keywords: 5G New Radio (NR); Cognitive Radio (CR); Dynamic Spectrum Access (DSA); Spectrogram Compressive Sensing; Deep Spectrum Sensing Segmentation; Wideband Spectrum Sensing; Semantic Segmentation.
  DOI: 10.1504/IJAHUC.2025.10073449