Bulletin of Electrical Engineering and Informatics
Not a member yet
2885 research outputs found
Sort by
Facial micro-expression classification through an optimized convolutional neural network using genetic algorithm
Get PDFComputer vision facilitates machines to interpret the visual world using various computer aided detection (CAD)-based techniques. It plays a crucial role in micro-expression auto classification. A micro-expression is a brief facial movement which reveals a genuine emotion that a person tries to conceal, it usually lasts for a short duration and is imperceptible with normal vision. To reveal people’s genuine emotions, an automatic micro-expression screening using convolutional neural network (CNN) is in great need. Traditional methods for micro-expression recognition (MER) suffer from low classification accuracy due to inadequate CNN hyperparameters selection. The proposed approach addresses these challenges by using an optimized CNN with adequate learning rate, batch size, epochs, and dropout rate. Real-coded genetic algorithm (RCGA) has been employed for the hyperparameter optimization. In this experimentation, features are extracted from the onset and apex frames of microexpression video clips of CASME II dataset. The proposed model's performance is measured using various metrics, including accuracy, precision, and recall. The proposed approach’s performance is then compared with an optimized CNN using random search algorithm. The empirical investigation of existing CNN-based methods has proven efficacy of our proposed model
Expert judgment, limitation inference, and threshold values to optimize diagnosis in eye diseases expert system
Get PDFThis research aimed to develop an optimal expert system by adopting a simplified approach. The methodology integrates an expert judgment approach, limitation inference, and establishing a threshold value. Expert judgment is pivotal in assigning a percentage weight to each rule, facilitating a nuanced evaluation of diagnostic criteria to augment the system's precision. Moreover, incorporating limitation inference strategically constrains the number of user inquiries, streamlining the diagnostic process and enhancing overall efficiency. Additionally, the imposition of a threshold value ensures a more precise early diagnosis by delineating specific criteria for condition identification. This comprehensive approach underscores the paramount importance of user experience and aims to alleviate the burden on individuals seeking a diagnosis. Ultimately, the anticipated outcome of this study is the development of an expert system poised to deliver early diagnoses with heightened efficiency and accuracy. By integrating expert judgment, limitation inference, and threshold values, this research embodies a refined and user-centric paradigm for eye disease diagnosis, promising significant advancements in global eye health
A stereo-vision system for real-time person detection in ADAS applications using a fine-tuned version of YOLOv5
Get PDFPedestrian detection holds significant importance in advanced driver assistance systems (ADAS) applications, and presents a challenging task in this field. While the advent of deep learning has facilitated the introduction of various pedestrian detectors characterized by accuracy and low inference speed, there persists a need for further improvements. Notably, ADAS requires accurate detection of pedestrians in various environmental conditions that can adversely impact the model’s performance, such as poor lighting, and bad weather. Furthermore, an imperative requirement involves the incorporation of distance estimation in conjunction with pedestrian detection, with an extension of detection capabilities to encompass cyclists and riders, who are equally crucial for ensuring road safety. Therefore, this paper introduces a stereovision system designed for the detection of pedestrians, cyclists, and riders. The initial phase, involves improving the performance of you only look once version 5 (YOLOv5s) through a fine-tuning process with a custom dataset integrating augmentation techniques to common objects in context (COCO) dataset. The detector is trained using Google Colab, and tested in real-time with a Raspberry Pi 4 model B, 8 G RAM. A comparative analysis is conducted between the YOLOv5s and the fine-tuned model to prove the accuracy of our approach. The results showcase a high performance of the detector reaching an accuracy exceeding 79%
Predicting graduation in Moroccan open-access bachelors: early indicators and re-enrollment data
Get PDFThe primary aim of higher education institutions is the successful graduation of their students. This study explores open-access higher education in Morocco, introducing a predictive model for assessing the probability of students achieving a science bachelor's degree. We analyzed data from 2012 to 2022, initially encompassing 45,573 student entries, and narrowed it down to 14,054 records after data cleaning. Focusing on early academic indicators from enrollment onwards-excluding current program performance—we used popular machine learning classifiers to examine the predictive capacity for student graduation and early dropout. Our comparison included analyses with and without re-enrollment data. Upon analyzing various machine learning algorithms, we attained accuracies between 79% and 86%, identifying random forest (RF) as the superior model for predicting outcomes both with and without incorporating re-enrollment data. This analysis was grounded on initial indicators observed during enrollment and throughout subsequent years, deliberately excluding current academic performance metrics from consideration
An efficient snow flake schema with hash map using SHA-256 based on data masking for securing employee data
Get PDFIn various organizations and enterprises, data masking is used to store sensitive data efficiently and securely. The data encryption and secret-sharing-based data deploying strategies secure privacy of subtle attributes but not secrecy. To solve this problem, the novel snowflake schema with the hash map using secure hash algorithm-256 (SHA-256) is proposed for the data masking. SHA-256 approach combines data masking by secret sharing for relational databases to secure both privacy as well as the confidentiality of secret employee data. The data masking approach supports preserving and protecting the privacy of sensitive and complex employee data. The data masking is developed on selected database fields to cover the sensitive data in the set of query outcomes. The proposed method embeds one or multiple secret attributes about multiple cover attributes in a similar relational database. The proposed method is validated through different performance metrics such as peak signal-to-noise ratio (PSNR) and error rate (ER) and it achieves the values of 50.084dB and 0.0281 when compared to the existing methods like Huffman-based lossless image coding and quad-tree partitioning and integer wavelet transform (IWT)
Hyperparameter tuning for deep learning model used in multimodal emotion recognition data
Get PDFThis study attempts to address overfitting, a frequent problem with multimodal emotion identification models. This study proposes model optimization using various hyperparameter approaches, such as dropout layer, l2 kernel regularization, batch normalization, and learning rate schedule, and discovers which approach yields the most impact for optimizing the model from overfitting. For the emotion dataset, this research utilizes the interactive emotional dyadic motion capture (IEMOCAP) dataset and uses the motion capture and speech audio data modality. The models used in this experiment are convolutional neural network (CNN) for the motion capture data and CNN-bidirectional long short-term memory (CNN-BiLSTM) for the audio data. This study also applied a smaller model batch size in the experiment to accommodate the limited computing resources. The result of the experiment is that the optimization using hyperparameter tuning raises the validation accuracy to 73.67% and the f1-score to 73% on audio and motion capture data, respectively, from the base model of this research and can competitively compete with another research model result. It is hoped that the optimization experiment results in this study can be useful for future emotion recognition research, especially for those who have encountered overfitting problems
Improved quantum inspired evolution algorithm with ResNet50 for spectrum sensing in cognitive radio networks
Get PDFSpectrum is considered one of the most highly regulated and limited natural resources. Cognitive radio (CR) relies on cutting-edge technology which helps to rectify the issues related to spectrum shortage in wireless communication systems. The CR technology allows the secondary user to accomplish the process related to spectrum sensing for identifying the usage of spectrum in the cognitive radio network (CRN). Though various spectrum sensing approaches are introduced, they exhibit complexity during spectrum sensing. To overcome the issues related to spectrum sensing and utilization, this research introduces improved quantum inspired evolution (IQISE) algorithm with ResNet 50 architecture. The IQISE-ResNet 50 which helps to enhance the spectrum efficiency is used in spectrum sensing. The detection of occupied and unoccupied users in CRN is performed using ResNet 50 architecture, while the IQISE is utilized in the process of training the model and optimizing the weights to enhance spectrum sensing efficiency. The experimental results show that the results achieved by the proposed approach are more effective than S-QRNN and honey badger remora optimization-based AlexNet (HBRO-based AlexNet). For example, the probability of correct classification of the proposed approach at -10 dB for binary phase shift keying (BPSK) modulation is 0.55, whereas the S-QRNN achieves an accuracy of 0.49
EXIT chart analysis of regular and irregular LDPC convolutional codes on AWGN channel
Get PDFLow-density parity-check (LDPC) codes are widely recognized for their excellent forward error correction, near-Shannon-limit performance, and support for high data rates with effective hardware parallelization. Their convolutional counterpart, LDPC convolutional codes (LDPC-CCs), offer additional advantages such as variable codeword lengths, unlimited parity-check matrices, and simpler encoding and decoding. These features make LDPC-CCs particularly suitable for practical implementations with varying channel conditions and data frame sizes. This paper investigates the performance of LDPC-CCs using the extrinsic information transfer (EXIT) chart, a graphical tool for analyzing iterative decoding. EXIT charts visualize mutual information exchange and help predict convergence behavior, estimate performance thresholds, and optimize code design. Starting with the EXIT chart principles for LDPC codes, we derived the mutual information functions for variable and check nodes in regular and irregular LDPC-CC tanner graphs. This involved adapting existing EXIT functions to the periodic parity-check matrix of LDPC-CCs. We compare regular and irregular LDPC-CC constructions, examining the impact of degree distributions and the number of periods in the parity-check matrix on convergence behavior. Our simulations show that irregular LDPC-CCs consistently outperform regular ones, and the EXIT chart analysis confirms that LDPC-CCs demonstrate superior bit error rate (BER) performance compared to equivalent LDPC block codes
Enhanced student attendance and communication in educational management systems
Get PDFThe internet of things (IoT) and radio frequency identification (RFID) technology were explored to devise a beneficial approach for managing student attendance. The research developed a system that uses RFID tags embedded in student bracelets to gather presence data via strategically placed sensors. The system leverages real-time databases and Google technologies to enhance the student experience through an online platform, while also utilizing RFID for authentication. Focusing on improving user experience (UX) through effective design, the proposed system offers a pleasurable and cost-effective solution. Developed using popular web technologies such as Firebase, React.js, and Tailwind, along with Arduino chips and sensors, the system provides a practical solution for managing student attendance, academic performance, and administrative communication. The research highlights the potential of RFID technology in improving student management and academic performance. By decreasing the effort needed by traditional systems and proving cost-effective in the long term, it could act as a potential choice for implementation in educational institutions worldwide
Optimization of perovskite solar cell with MoS2-based HTM layer using hybrid L27 Taguchi-GRA based genetic algorithm
Get PDFThis article proposes an optimization method to predictively model the perovskite solar cell with molybdenum disulfide (MoS2) based inorganic hole transport material (HTM) for improved fill factor (FF) and power conversion efficiency (PCE) by finding the most optimum thickness and donor/acceptor concentration for each layer via a hybrid L27 Taguchi grey relational analysis (GRA) based genetic algorithm (GA). Numerical simulation of the device is carried out by employing one-dimensional solar cell capacitance simulator (SCAPS-1D) while the optimization procedures are developed based on combination of multiple methods; L27 Taguchi orthogonal array, GRA, multiple linear regression (MLR), and GA. The results of post-optimization reveal that the most optimum layer parameters for improved FF and PCE are predicted as follows; SnO2F thickness (0.855 μm), SnO2F donor concentration (9.206×1018 cm-3), TiO2 thickness (0.011 μm), TiO2 donor concentration (9.306×1016 cm-3), CH3NH3PbI3 thickness (0.897 μm), CH3NH3PbI3 donor concentration (0.906×1013 cm-3), MoS2 thickness (0.154 μm), and MoS2 acceptor concentration (9.373×1017 cm-3). Both FF and PCE of the device are improved by ~1.1% and ~12.6% compared to the pre-optimization