In this work, we analyse pointing error effects on the performance of amplify-and-forward (AF) relaying multiple-input multiple-output (MIMO) free-space optical (FSO) communication system employing sub-carrier quadrature amplitude modulation (SC-QAM) signal over log-normal and gamma-gamma distributed atmospheric turbulence channels. We study the pointing error effects by taking into account the influence of beam-width, aperture size and jitter variance on the average symbol error rate (ASER), which is derived in closed-form expressions of MIMO/FSO and SISO/FSO systems. In addition, the number of relaying stations is taken into account in the statistical model of the combined channel including atmospheric loss, atmospheric turbulence and pointing error impairments. The numerical results show that by combining AF relaying stations and MIMO/FSO configurations, the link length can be extended because the transmitted power is reduced according to the amplifier gain. Moreover, the performance of AF relaying MIMO/FSO systems is better than that of AF relaying SISO/FSO systems at the same link length.
This article presents a comprehensive scenario for utilizing drones to search for victims during high-rise fires. Given the complexity and difficulty of accessing such environments, employing drones for rapid victim searches may significantly reduce rescue operation time. Specifically, we have devised a two-phase drone control scenario. In phase 1, the drone autonomously navigates to the target location, benefiting from Deep Learning technology to automatically avoid obstacles during its movement. In the second phase, we have introduced a scanning algorithm to assist the drone in systematically surveying the entire rescue and recovery area to locate victims. We have constructed a test scenario and assessed the results using AirSim simulation software.
Setting analog cellular computers based on cellular neural networks systems (CNNs) to change the way analog signals are processed is a revolutionary idea and a proof as well of the high importance devoted to the analog simulation methods. We provide an in-depth description of the concept exploiting analog computing based on the CNN paradigm to solve nonlinear and highly stiff ordinary differential equations (ODEs) and partial differential equations (PDEs). We appply our method to the analysis of the dynamics of two systems modeled by complex and stiff equations. The first system consists of three coupled Roumlssler oscillators in a Master-Slave-Auxiliary configuration. The capabilities of this coupled system to exhibit regular and chaotic dynamics have been demonstrated so far. The synchronization modes of the coupled system can be exploited in chaotic secure communications. The second system is the Burgers' equation which is a well-known classical model for analyzing macroscopic traffic flow motions/ scenarios. As a proof of concept of the proposed approach, the results obtained in this paper are compared with the results available in the relevant literature (benchmarking) and, the proposed concept is validated by a very good agreement obtained. The computation based CNNs paradigm is advantageous as it provides accurate and ultra-fast solutions of very complex ODEs and PDEs and performs real-time computing.
1 April 2022 in Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization
Tumours and inflammatory lesions in the brain are common and very dangerous brain diseases. Processing MRI images to detect and distinguish inflammatory lesions from tumours early and quickly will help doctors have timely treatment methods, preventing the disease from getting worse. In this paper, with the MRI image datasets of inflammatory lesions and brain tumours collected from Bach Mai hospital, Vietnam, we developed a complex network with low computational complexity to extract the features of inflammatory lesions and tumours and then use machine learning classification algorithms including SVM, kNN, Random Forest, Logistic Regression, Naive Bayes to distinguish. The performance of the classification algorithms is evaluated through the parameters of accuracy, precision, F1-score, recall and ROC curve. Accuracy achieved by all algorithms is very high in which the highest result is up to 90.26%. This research has successfully applied a complex network with low computational complexity to extract features of MRI images, thereby improving the accuracy in distinguishing inflammatory lesions and tumours in the brain using machine learning algorithms. The rapid and accurate classification of inflammatory lesions and brain tumours is of great significance in screening patients. With such relatively accurate results, in the future, the system can be directly applied in practice.