Indonesian Journal of Electrical Engineering and Informatics (IJEEI)
Not a member yet
776 research outputs found
Sort by
Parallel Pipelined Hardware Acceleration of Fast Fourier Transforms on FPGA
Full text linkThe Fast Fourier Transform (FFT) is widely used in digital signal processing ap-plications and particularly for implementing convolution operation for real-time object detection using CNN. This paper proposes an efficient hardware architecture for Radix-2 FFT computation, implemented on an FPGA, employing multiple parallel and pipelined stages of butterfly units. The proposed architecture utilises Block RAM to store inputs and twiddle factor values to compute the transform. The hardware for the proposed architecture is synthesised on a Zync Ultrascale FPGA and its performance is evaluated using parameters such as critical path delay, throughput, device utilisation and power consumption.The performance of the proposed parallel pipelined architecture for 8 point FFT, measured in FFTOPS, is found to be 67% higher than the non-pipelined architecture. Performance comparison with the state-of-the-art parallel pipelined methods confirm the acceleration achieved by the proposed FFT architecture. A comprehensive comparison of the proposed hardware with the synthesised version of the FFT IP core bundled with the Vivado Design suite is also presented in the paper
Machine Learning-Driven Pre-Broadcast Video Codec Validation: Ensuring Seamless Television Transmission
Full text linkThis study addresses the critical challenge of ensuring uninterrupted television broadcasting by proactively detecting video codec errors, focusing on TV Laayoune, a prominent Moroccan channel. We developed a machine learningbased methodology that identifies incompatible codecs before they disrupt live broadcasts. The approach involves data collection from multiple sources, including TV Laayoune's archives, metadata extraction via FFmpeg, and a hybrid model that combines Convolutional Neural Networks (CNNs) with Long Short-Term Memory (LSTM) networks. Integrated into the broadcasting pipeline, this model achieved a 95% accuracy rate, significantly enhancing broadcast reliability and operational efficiency. Additionally, we propose a user-friendly interface for real-time error detection, comprehensive workflow integration, and automated alerts. This innovative solution addresses common broadcast challenges, reducing operational risks and improving the viewer experience
A Multiclass Support Vector Machine Based Direction-of-Arrival Estimation Technique using Spherical Antenna Array with Undefined Mutual Coupling
Full text linkIn antenna array signal processing, estimating the direction-of-arrival (DoA) remains a challenge and basic problem. In this paper, a DoA estimation technique using support vector machine (SVM) classification is developed using spherical antenna array (SAA). The source signal impinging on SAA is decomposed using spherical harmonics (SH). Both magnitude and phase features are computed from the decomposed SH signals. The magnitude and phase features are classified into DoA classes using multi-class SVM (MC-SVM) algorithm. Due to the deterministic and non-probabilistic nature of SVM algorithm, it exhibits high computational speed and less complex than the neural network-dependent learning algorithms. Numerical experiments and experimental measured data (generally accepted ground to test any method) are used to evaluate the performance of the proposed technique. The developed algorithm exhibit high level of robustness at different signal-to-noise ratios (SNR) in the estimation of DoA. Root mean square error (RMSE) performance metrics is employed in the analysis of the proposed method against the state-of-the-art. The results obtained are motivating enough for the deployment of the proposed algorithm in practical scenarios
Examining the ability of Advanced Systems of Wireless Communication Enhanced by IRS Technology
Full text linkIntelligent Reflecting Surfaces (IRSs) represents a pivotal component of technology, facilitating the enhancement of wireless communication performance and the manipulation of electromagnetic propagation environment. IRS technology has the remarkable capability to transform wireless channels from highly probabilistic to notably deterministic, effectively mitigating the substantial losses encountered in the millimeterwave (mmWave) band. Our analysis emphasizes how this innovative technology has ushered in a new era in wireless communications. Within the scope of this study, we delved into investigating the effectiveness of IRSassisted wireless transmissions across various scenarios, encompassing both line-of-sight (LOS) and non-line-of-sight (NLOS) conditions. Our investigation involved the simulation of a 32×32 IRS array with a wavelength of 1 meter and an incident angle of 45 degrees. By manipulating the phase shifts of individual IRS elements, we examined their impact on achievable data rates concerning the number of elements. We also explored the relationship between throughput and separation distances, highlighting the significance of IRS placement in achieving optimal data rates. Channel capacity analysis was conducted for single IRS configurations with 50 and 100 elements, as well as dual IRS setups, shedding light on the capacity improvements achievable in different arrangements. Additionally, our study delved into Bit Error Rate (BER) performance in cooperative doubled IRS-aided wireless communication, employing a range of digital modulation techniques across various Signal-to-Noise Ratio (SNR) levels. This insight offers a valuable perspective on the reliability of IRS-aided systems across diverse modulation schemes. We also undertook a comprehensive Spectral Efficiency (SE) analysis, investigating IRS-assisted Multiple-Input, Single-Output (MISO) and Multiple-Input, Multiple-Output (MIMO) communications using various modulation schemes. Finally, we examined path loss characteristics across indoor encompassing different environments, especially at 20 GHz and 28 GHz using vertical to vertical (V-V) polarization. The culmination of this thorough simulation study underscores the tremendous potential of IRS technology in revolutionizing wireless communication across diverse scenarios, offering invaluable insights for future design and development endeavors
Flexible Potentiostat Readout Circuit for Electrochemical Sensors
Full text linkPersonalised health wearables reach their full potential when sensors are integrated with its interfacing system. Recent approaches have primarily focused on the development of readout circuits limited to the electrochemical chip and basic signal conditioning components. However, integrating a readout circuit with a microcontroller offers significant advantages such as enhanced data processing capabilities. Other than incorporating a microcontroller within the readout circuit, we also designed the entire potentiostat system on a flexible polyimide substrate, making it suitable for wearable applications. In this work, we describe the design, fabrication and testing of a flexible potentiostat readout circuit for electrochemical sensors. The core of the interface circuit is two chips, a microcontroller ATSAMD21G18A-MUT (Microchip Technology) and a programmable analog front-end integrated circuit from Texas Instruments. These chips along with a voltage regulator, resistors and capacitors were integrated onto a single, flexible, printed circuit board. To verify the functionality of the flexible readout circuit, it was connected to an electrochemical sensor and Cyclic Voltammetry (CV) was performed. The separation between peaks (ΔEp), were measured using the flexible board and compared with a commercial potentiostat (Emstat Pico). EmStat Pico has ΔEp = 0.133V, while our potentiostat produced ΔEp of 0.132V, indicating minimal variations with the same PCB layout, despite using different substrates. The standard rate constant (Ks) of electron transfer can also be obtained from CV and was measured to be 0.0037 for the rigid PCB and 0.0035 for the flexible PCB
DR-CNN+ Approach for Standardized Diabetic Retinopathy Severity Assessment
Full text linkDiabetic retinopathy (DR) is a serious eye disorder that damages the retina and can lead to vision impairment and blindness, especially in individuals with diabetes. Early identification is crucial for a positive outcome, however, diabetic retinopathy can only be diagnosed with color fundus photographs, which is a technique that is difficult and time-consuming. To address this issue, this paper presents a Deep Learning-based algorithm that utilizes DR - convolutional neural network+ (DR-CNN+) to classify retinal pictures into different stages of diabetic retinopathy. The proposed algorithm is trained on a dataset of 11000 colored retinal pictures from the training set and 2200 photos from the testing set. The simulation results demonstrate that the DRCNN+-based algorithm can achieve high levels of accuracy, sensitivity, and specificity. Our proposed DR-CNN+ model not only improves diagnostic performance for diabetic retinopathy severity evaluation, but it also saves training time by 95% when compared to current models." Overall, this paper highlights the potential of using deep learning and CNNs to improve the detection and grading of diabetic retinopathy, which could have a significant impact on the prevention of blindness caused by this disease
Advanced Techniques for Improved Bangladeshi Number Plate Detection and Character Recognition in Automated Parking Systems
Full text linkThis paper presents a novel technique for efficient extraction of vehicle number plates from camera-captured images and accurate recognition of Bangla characters embedded within them. With the exponential growth of vehicular traffic in densely populated regions like Bangladesh, automation becomes crucial, making vehicle plate recognition pivotal for tracking stolen vehicles and enhancing traffic control measures. Leveraging conventional computer vision and image processing techniques, our proposed system incorporates specific features inherent to Bangladeshi number plates, thus enhancing recognition accuracy. Our application makes use of the OpenCV library to underscore the strength of the algorithm, which has been confirmed through real-time testing across different weather conditions and varying image qualities. The results show a remarkable accuracy rate of 92.3%, affirming our technique's reliability in vehicle number plate detection and character recognition. Moreover, the integration with MySQL database and Arduino UNO enables real-time application in automated parking systems, offering seamless entry procedures and accurate billing, thus addressing critical concerns in modern transportation management systems. Our algorithm not only enhances security measures but also streamlines parking facility management, contributing to safer and more efficient urban mobility solutions
Wireless Need Sharing and Home Appliance Control for Quadriplegic Patients Using Head Motion Detection Via 3-Axis Accelerometer
Full text linkPatients who are quadriplegic are immobile in all four limbs. Quadriplegic patients with low voices struggle to communicate their needs to family members or caregivers, requiring assistance to use household items like fans and lights. This paper presents an electronic system designed to enhance the quality of life of quadriplegic patients by enabling them to share needs, manage household items, and monitor their health. The quadriplegic patient can move their head. In the proposed system, an accelerometer sensor placed on the patient’s forehead to record head movement, which is processed to detect and share needs or operate home appliances. The system consists of two units: one in the patient’s bed and another in a common place at home. Both communicate through Bluetooth. By moving head in the right direction, patients can share needs like water, rice, snacks, sickness or washroom. The common unit notifies caregivers through a matrix display and makes sounds with a buzzer. Patients can also control specific household appliances through left-head movements. The system also features a pulse oximeter sensor for monitoring heart rate and oxygen saturation. A prototype of the system has been developed and tested, and it is functioning smoothly. This system will free the quadriplegic patients from dependence on others and make their lives easier
The Efficiency of HEVC/H.265, AV1, and VVC/H.266 in Terms of Performance Compression and Video Content
Full text linkIn recent times, there has been a significant focus on digital compression. The purpose of this study is to undertake a comparative evaluation and examination of the efficacy of the latest standards, namely HEVC, AVI, and its successor VVC. The determination of which standard to utilize relies heavily on factors such as the inherent characteristics of the video, its functionalities, quantization parameters, image quality, as well as the size and video content, this latter, is often classified by spatio-temporal complexity using spatial and temporal information (SI/TI). In reality, they are mostly used for original video sources. The efficiency of encoding original video sources is unknown. The results show that each standard has characteristics that sometimes make it superior to others. In addition, We observe that By understanding how SI and TI affect encoding efficiency, we will be able to better optimize the encoding process and reduce the amount of data that needs to be stored, transmitted, and processed. This could help to reduce the amount of time and energy required to encode video content, as well as reduce the amount of storage space needed to store it. Compared to H.265/HEVC, AV1 is more efficient at compressing HD and FHD video, and more efficient for SD video. In addition, experiments show that VVC/H.266 has higher compression efficiency
Malware Classification Using Machine Learning and Dimension Reduction Techniques on PE File Data
Full text linkThe digital transformation has enhanced efficiency, transparency, and accessibility but has also led to a notable increase in cyber incidents, including malware attacks. According to the 2022 annual report from the Honeynet Project by the National Cyber and Encryption Agency, Indonesia experienced over 370 million cyber attacks, with 800,000 of these being malware attacks. The increasing complexity of Portable Executable files further complicates accurate classification in machine learning models. This research aims to develop an effective malware detection approach using machine learning classifiers—Random Forest, XGBoost, and AdaBoost—on raw feature dataset and integrated feature dataset. Dimension reduction techniques such as Principal Component Analysis and Linear Discriminant Analysis were utilized to enhance classification efficiency. The results demonstrated that Random Forest and XGBoost consistently outperformed AdaBoost, particularly in classifying ransomware, achieving recall values ranging from 0.72 to 0.85 and F1-scores from 0.74 to 0.81 For the trojan class, both Random Forest and XGBoost achieved recall values ranging from 0.96 to 0.97, with corresponding F1-scores between 0.95 and 0.97. Both classifiers maintained high precision, recall, and F1-scores across all malware classes, even with reduced feature sets