Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
A reliable and secure demand side control for an IoT-enabled smart power system using machine learning
Get PDFAs the adoption of IoT-enabled smart power systems grows, the necessity for reliable and secure demand-side control becomes paramount. This paper introduces a robust demand-side management (DSM) engine that leverages machine learning to enhance both the reliability and security of smart grids. This paper presents a novel demand-side control system leveraging advanced machine learning techniques to optimize energy usage in smart power systems. The proposed system integrates IoT devices for data acquisition and employs machine learning algorithms to forecast energy demand, detect anomalies, and enable adaptive control strategies. By harnessing predictive analytics, the system anticipates consumption patterns and adjusts power distribution to maintain stability and prevent overloads. Moreover, robust security protocols are incorporated to protect the system against cyber threats and unauthorized access, ensuring data integrity and user privacy. Extensive simulation results demonstrate the system’s efficacy in reducing energy wastage, improving load balancing, and enhancing the overall reliability of the power grid. This research underscores the potential of combining IoT and machine learning to develop resilient and secure energy management solutions, paving the way for more sustainable and smart power systems
Elevating intelligent voice assistant chatbots with natural language processing, and OpenAI technologies
Get PDFBusinesses can offer support to customers outside of usual business hours and across time zones by employing chatbots, which can provide round-the-clock support. Chatbots can react to user inquiries quickly, reducing wait times and improving customer satisfaction. It becomes challenging for the chatbot to differentiate between two queries that users pose that carry the same meaning, making it harder for it to understand and react appropriately. The aim of this research is to develop a chatbot capable of understanding the semantic meaning of questions as well as recognizing various speech patterns, accents, and dialects to provide accurate responses. In this research, we have implemented a voice-enabled chatbot system where users can verbally pose questions, and the chatbot provides responses through voice assistance. The architecture incorporates several key components: a question-answer database, OpenAI embedding for semantic representation, and OpenAI text-to-speech (TTS) and speech-to-text (STT) for audio-to-text and text-to-audio conversion, respectively. Specifically, OpenAI embedding is utilized to encode questions and responses into vector representations, enabling efficient similarity calculations. Additionally, extreme gradient boosting (XGBoost) is trained on OpenAI embeddings to identify similarities between user queries and questions within the dataset. This framework allows for seamless interaction between users and the chatbot, leveraging state-of-the-art technologies in natural language processing (NLP) and speech recognition. The outcome demonstrates that the XGBoost model delivers excellent outcomes when it is trained on OpenAI embedding and tuned with the particle swarm optimizer (PSO). The OpenAI-generated embedding has good potential for capturing sentence similarity and provides excellent information for models trained on it
Interactive multimedia e-collaboration for innovative linguistics education
Get PDFThis study aims to investigate the needs of students and lecturers regarding interactive multimedia resources in linguistics at the Faculty of Teacher Training and Education, Universitas Borneo Tarakan, to facilitate further development. The findings reveal a significant gap between current instructional provisions and the specific needs of students and faculty, highlighting the necessity for pedagogical innovation to enhance interaction and understanding in linguistics. Utilizing a mixed-methods approach, the research included surveys and interviews with participants in linguistics courses. Results indicated that 86% of students sought in-depth knowledge of linguistics, and 73% felt that existing support was inadequate. It underscores a high demand for a focus on selected topics, simplified explanations, and multimedia interactivity. The findings demonstrate that instructional materials are poorly aligned with teaching needs, negatively impacting educational methodologies and failing to effectively address students' relevant needs. The implications of this study extend to practice and further research, urging faculty members to increasingly integrate multimedia elements into their teaching and develop tailored resources based on identified needs. Newly created materials should undergo practical evaluation to enhance student satisfaction and performance in linguistics studies
Accelerated framework for image compression and reconstruction based on compressive sensing
Get PDFImage compression is a crucial field driven by advancements in communication and imaging technologies. Its primary goal is to achieve low bit rates while maintaining high-quality image reconstruction. Compression is essential in digital image processing, multimedia applications, and medical imaging. Various algorithms exist for image compression and reconstruction, each differing in efficiency. Compressive sensing (CS) algorithms, commonly used for radar data reconstruction, require iterative computations that demand significant processing power and time, limiting real-time applications. To overcome these challenges, this study proposes a parallel-pipelined processing approach to enhance compression and reconstruction efficiency. The method accelerates processing speeds, increases data throughput, and optimizes performance by reducing data size. The proposed approach divides image data into multiple parallel processing branches, significantly reducing computational cycles. This results in faster execution and improved real-time applicability. MATLAB simulations and field-programmable gate array (FPGA) hardware implementations have been conducted to validate the system’s effectiveness. The results demonstrate that the parallel-pipelined method significantly enhances efficiency compared to traditional approaches, making it suitable for applications requiring high-speed image processing, such as satellite imaging and medical diagnostics
Hybrid TCP SYN attack detection model in SDN
Get PDFSoftware defined network (SDN) is a developing concept that emerged recently to overcome the constraints of traditional networks. The distinguishing characteristic of SDN is the uncoupling of the control plane from the data plane. This facilitates effective network administration and enables efficient programmability of the network. Nevertheless, the updated architecture is susceptible to cyberattacks including distributed denial of service (DDoS) attacks, that can impair network regular functions and hinder the SDN controller from assisting authorized users. This paper introduces hybrid deep learning model, to detect DDoS assaults triggered by TCP SYN attacks in SDN environments. Our proposed model integrates a temporal convolutional network (TCN) with a stacking classifier that leverages logistic regression, which is an innovative hybrid approach. We assessed the performance of our model by utilizing the benchmark CICDDoS2019 dataset. When compared to other benchmarking techniques, our model significantly improves attack detection. The experimental results indicate that the proposed hybrid model attains 99.9% accuracy for attack detection compared to the available approaches
Automated defect detection in submersible pump impellers using image classification
Get PDFCasting is a crucial manufacturing process used to produce complex metal parts, but it is often plagued by defects such as cracks, porosity, shrinkage, and cold shuts, which can compromise quality and lead to financial losses. Traditional visual inspection methods for detecting these defects are slow and prone to human error, making them inefficient for large-scale production. This project proposes automating the defect detection process using advanced AI-powered non-destructive testing (NDT) techniques. Specifically, convolutional neural networks (CNNs), a deep learning model, are employed for real-time visual inspection of castings. CNNs, trained on high-resolution images, can accurately identify surface defects such as cracks, scratches, and dimensional irregularities, significantly improving inspection speed and accuracy. The performance metrics of the system include defect detection accuracy, false positive and false negative rates, processing time, and scalability for high-volume production environments. By minimizing human intervention, this automated system reduces error rates, enhances product quality, and lowers production costs. Furthermore, the real-time capabilities of CNNs allow for rapid feedback, preventing defective parts from advancing through the production line. Overall, the integration of AI-based vision systems boosts efficiency, sustainability, and profitability in manufacturing, ensuring castings meet customer specifications with minimal errors
Query keyword extraction in discriminative marginalized probabilistic neural method for multi-document summarization
Get PDFThe large number of textual documents in the medical field makes it very difficult for readers to obtain comprehensive information. Users usually use a query approach to get the desired information. Using the correct query will produce relevant information. In the existing discriminative marginalized probabilistic neural method, referred to as DAMEN, used for multi-document summarization, a background sentence query is used to retrieve the top-K relevant documents and then generate a summary of these documents. However, the background sentence query used to retrieve the top-K documents did not provide accurate summary results. The author improved the DAMEN model by adding a keyword extraction process to the query background sentence. We call this model Q-DAMEN. Our model shows significant improvement over the original DAMEN method, with the best results achieved by the variation of using a keyword query entered into the discriminator component and a background sentence query entered into the generator component. The multipartieRank keyword extraction method shows the best results with a Rouge-1 value of 29.12, Rouge-2 of 0.79, and Rouge-L of 15.53. The results demonstrate that the more accurate the keywords extracted from the sentence background query, the more accurate the multi-document summaries generated
Abstractive and extractive based YouTube transcript summarization: a hybrid approach
Get PDFThe rapid advancement in the field of communication and ubiquitous access to computing has led to the proliferation of large amounts of video content on YouTube and other social media platforms. However, getting precise information from the video in concise textual manner remains a challenge. Different extractive and abstractive text summarization methods are prevalent in the literature. In this paper, classical extractive text summarization methods Luhn’s algorithm, TextRank algorithm and Keyword- based summarization are combined to develop a combined extractive (CE) method. To enhance its performance, bidirectional and auto-regressive transformers (BART) is investigated and integrated as a hybrid model. Further, we explore how Kmeans clustering algorithm can be used for text summarization in general and with the proposed hybrid approach for improvement in text summarization. Using CNN/DailyMail dataset, assessment of text summarization methods based on ROUGE scores and time taken for summary generation is carried out. Based on the ROUGE score, we observe that the proposed hybrid method - 0.2644 is better than traditional extractive summarization methods. The combination of hybrid method with K-means further improved the score to 0.3227. The time taken by them for summary generation are 138.09 and 142.16 seconds respectively. This work experimented with different classical and transformer-based text summarization techniques to explore the complementary aspects and the results obtained are comparable with that of existing models with less time for text summarization
Efficient model for cotton plant health monitoring via YOLO-based disease prediction
Get PDFProtecting plants from diseases involves recognizing the symptoms and identifying practical, safe, and reasonable treatment methods. Holistic approaches based on particular times or seasons can reduce plant resistance and minimize tedious work. Technological advancements have led to the development of microscopic examinations and computational methods using machine learning techniques to detect diseases automatically and quickly using leaf images. This study builds the prediction model using EfficientNet and YOLO neural network architectures from computer vision. The development of a model that assists farmers in identifying cotton disease so that they use pesticides that may treat it further utilizes this concept. In the physical world, the input is accepted from many different sources, so observing the model’s output is necessary. This work concentrates on model response to the inputs from physical devices, and analysis shows that the monitoring varies the results. A novel convolutional neural network (CNN) based on the EfficientNet architectures and variations of YOLO architectures is used to classify and identify the objects in cotton leaf. The EfficientNetB4 yielded 100% accuracy for healthy leaf and powdery mild leaf classes, and YOLO v4 version with 96%, 98.3%, 99.2%, and 0.70 for precision, recall, [email protected], mAP120.5:095 respectively. These results indicate that consequences vary in real-time per environmental parameters such as light effect and devices, and analysis shows that monitoring affects the results
Optimizing 2D-to-3D image conversion for precise flat surface detection using laser triangulation and HSV masking
Get PDFThis study tackles a critical challenge in converting two-dimensional (2D) images into three-dimensional (3D) representations, focusing on the precise detection of flat surfaces. The research utilizes a triangulation method involving laser and camera systems, emphasizing the optimization of laser shooting angles and camera positioning to accurately determine z-coordinates. The methodology employs hue, saturation, and value (HSV) color masking, which has proven superior to traditional red, green, blue (RGB) methods for isolating red line objects. Key findings indicate that the optimal laser angle, β1=70.65°, significantly minimizes root mean square (RMS) error, thereby enhancing the accuracy of 3D imaging. Additionally, the use of three laser lines at different angles enables a more comprehensive detection of z-coordinates by creating multiple reference points across the surface. This arrangement improves the robustness and precision of the 3D reconstruction process, as the intersecting laser lines generate detailed coordinate data that is critical for accurately mapping surface irregularities. These results not only support existing theories in digital feature extraction but also offer a robust framework for practical applications in manufacturing and quality control, particularly in surface defect detection. The study’s innovative approach advances the field of computer vision, providing new insights and methodologies for optimizing image conversion techniques