Articles

Access the latest knowledge in applied science, electrical engineering, computer science and information technology, education, and health.

Advanced Search
Filter Icon

Filters article

Years

FAQ Arrow
Single Year
Range Year
0
0

Source Title

FAQ Arrow

Authors

FAQ Arrow

31,456 Article Results

Multi-task deep learning for Vietnamese capitalization and punctuation recognition

10.11591/ijai.v14.i2.pp1605-1615
Phuong-Nhung Nguyen , Nguyen Thu-Hien , Truong-Thang Nguyen , Nguyen Thi Thu-Nga , Nguyen Thi Anh-Phuong , Tuan-Linh Nguyen
Speech recognition is the process of converting the speech signal of a particular language into a sequence of corresponding content words in text format. The output text of automatic speech recognition (ASR) systems often lacks struc- ture, such as punctuation, capitalization of the first letter of a sentence, proper nouns, and names of locations. This absence of structure complicates compre- hension and restricts the utility of ASR-generated text in various applications, such as creating movie subtitles, generating transcripts for online meetings, and extracting customer information. Therefore, developing standardization solu- tions for the output text from ASR is necessary to improve the overall quality of ASR systems. In this article, we use the idea of multitask deep learning for the task of capitalization and punctuation recognition (CPR) for the output text of Vietnamese ASR, with the aim of the named entity recognition (NER) task as a supplement to help the CPR model perform better, and proposed to use text-to- speech (TTS) to create a dataset for CPR-NER multitask model training. The experiment results show that the multi-task deep learning model has improved CPR results by 6.2% of F1 score with ASR output and 7.1% on raw text.
Volume: 14
Issue: 2
Page: 1605-1615
Publish at: 2025-04-01

Model of semiconductor converters for the simulation of an asymmetric loads in an autonomous power supply system

10.11591/ijece.v15i2.pp1332-1347
Saidjon Tavarov , Mihail Senyuk , Murodbek Safaraliev , Sergey Kokin , Alexander Tavlintsev , Andrey Svyatykh
This article is devoted to the development of computer model with semiconductor converters for the simulation of asymmetric loads allowing to solve the voltage symmetry problems under asymmetric loads (active and active-inductive) for isolated electric networks with renewable energy sources (mini hydroelectric power plants). A model of a symmetry device has been developed in the MATLAB/Simulink environment based on a proportional-integral controller and a relay controller - P. The effectiveness of their use depends on the load's nature. The implementation of a voltage converter is presented considering a three-phase inverter with discrete key switching at 120, 150, and 180 degrees with a purely active load. Based on the harmonic analysis of the three-phase voltage at discrete conversion, the value of the first harmonic is determined. Voltage transformations under active-inductive load at 120, 150, and 180 degrees are mathematically described. To determine the harmonic spectrum, an analysis of the fast Fourier transform for the three-phase voltage of a MATLAB/Simulink semiconductor converter was carried out. It is established that the alternating current output voltage is generated on the output side of the inverter of a three-phase voltage source through a three-phase load connected by a star with a harmonic suppression method.
Volume: 15
Issue: 2
Page: 1332-1347
Publish at: 2025-04-01

Application of machine learning methods to analysis and evaluation of distance education

10.11591/ijece.v15i2.pp2172-2180
Ainur Mukhiyadin , Manargul Mukasheva , Ulzhan Makhazhanova , Aislu Kassekeyeva , Gulmira Azieva , Zhanat Kenzhebayeva , Alfiya Abdrakhmanova
In recent decades, distance learning has become an essential component of the modern educational system, providing students with flexibility and access to knowledge regardless of location. This paper discusses creating a hybrid machine-learning model for assessing the quality of distance learning based on survey data. The model combines two feature extraction methods: Term frequency-inverse document frequency (TF-IDF) and Word2Vec. Combining these methods allows for a more complete and accurate representation of text data, improving the quality of machine learning models. The study aims to develop and evaluate the effectiveness of the proposed hybrid model for analyzing survey data and assessing the quality of distance learning. The paper considers the tasks of collecting and preprocessing text data, experimentally comparing various feature extraction methods and their combinations, training and evaluating a machine learning model based on a combination of TF-IDF and Word2Vec features, as well as analyzing the results and assessing the effectiveness of the proposed model using various metrics. In conclusion, the prospects for further development and application of the proposed model in educational institutions to improve the quality of distance learning are discussed.
Volume: 15
Issue: 2
Page: 2172-2180
Publish at: 2025-04-01

Object retrieval analysis on plastic bottle waste recycling-based image control using convex hull algorithm and autoregressive integrated moving average prediction method

10.11591/ijece.v15i2.pp2055-2069
marisa Marisa , Azizul Azhar Ramli , Mohd Farhan Md Fudzee , Zubaile Abdullah
In Indonesia, plastic garbage bottles are the most common sort of waste. Given that waste is expected to grow annually, managing plastic waste is a major challenge. The results of the study were achieved by comparing the reference, which was a collection of manually created contour images, with 50 sets of vortex images with different forms and vortex areas as experimental objects. The results indicate that the suggested approach reports a mean error of 2.84%, a correlation coefficient of 0.9965, and a root mean square error of 0.2903 when compared to the manual extraction method. These findings imply that the extract area determined by the procedure outlined in this research is more accurate and nearer to the actual values. The proposed method can therefore be used in place of the traditional process for investigating cooling parameters through manual testing. With measurement values mean absolute percentage error (MAPE)=121,842, mean absolute deviation (MAD)=20,140, and mean squared deviation (MSD)=776,712, the trend analysis of plastic bottles for autoregressive integrated moving average (ARIMA) modeling leads to the conclusion that the waste from plastic bottles will continue to rise annually and that efforts must be made to address this trend with knowledge and waste recycling technology. Plastic that is advantageous to industry and society.
Volume: 15
Issue: 2
Page: 2055-2069
Publish at: 2025-04-01

Graphene-based THz antenna with a wide bandwidth for future 6G short-range communication

10.12928/telkomnika.v23i2.26562
Md. Kawsar; Daffodil International University Ahmed , Md. Sharif; Daffodil International University Ahammed , Md. Ashraful; Daffodil International University Haque , Narinderjit Singh; INTI International University Sawaran Singh , Jamal; Daffodil International University Hossain Nirob , Redwan; Daffodil International University A. Ananta , Kamal; Daffodil International University Hossain Nahin , Liton; Pabna University of Science and Technology Chandra Paul
In this study, we present the design and investigation of a terahertz (THz) frequency antenna optimized for the 2-10 THz range, featuring both single-element and multiple-input multiple-output (MIMO) configurations, with a focus on industrial and innovative applications to enhance future 6G communication systems. The antenna, constructed on a polyimide substrate with dimensions of 90×30 µm, achieves a bandwidth from 4.0328 to 10 THz. The MIMO configuration, which includes two ports, demonstrates excellent isolation with a value of -27 dB. The proposed antenna system achieves a gain of 12.38 dB and an efficiency of 89%, making it highly appropriate for THz communication applications. Furthermore, the envelope correlation coefficient (ECC) of 0.002 and diversity gain (DG) of 9.99 affirm the antenna’s effectiveness in MIMO systems. A resistance inductance capacitance (RLC) circuit model was employed to accurately represent the S11 curve, ensuring precise characterization of the antenna’s performance. These results underscore the probability of the proposed antenna for high-speed, short-range communication systems.
Volume: 23
Issue: 2
Page: 306-315
Publish at: 2025-04-01

Human–robot collaboration with mixed reality for interactive and safe workspaces

10.12928/telkomnika.v23i2.26275
Sanghun; Changwon National University Nam
Realizing seamless collaboration between humans and robots in shared workspaces requires advanced systems that can ensure safety and efficiency while considering the inherent unpredictability of human movement. This paper proposes a system that integrates mixed reality (MR) and robotics through a unified coordinate system to facilitate real-time interaction and collaboration. By leveraging a MR interface, human collaborators can visualize and interact with the projected paths of the robotic arms, thereby enhancing both spatial awareness and task coordination. The proposed system adapts the robot’s movement path dynamically using the Voronoi diagram algorithm to modify trajectories in response to the detection of a human hand within a predefined caution zone. This mechanism reduces the risk of collisions, which ensures safer collaborative environments. The proposed system’s ability to exchange motion information between the operator and the robot supports real-time adjustments and promotes an intuitive and efficient collaborative experience. Our findings suggest that integrating MR technology in human–robot collaboration systems can improve safety protocols and operational fluidity dramatically, thereby representing a significant step forward in the development of safe, efficient, and effective interactive robot systems.
Volume: 23
Issue: 2
Page: 526-532
Publish at: 2025-04-01

A system architecture for mixed reality systems in vocational schools in Indonesia

10.11591/ijict.v14i1.pp207-215
Satrio Pradono Suryodiningrat , Harjanto Prabowo , Arief Ramadhan , Harry Budi Santoso
In Indonesia, vocational schools are less favored compared to K -12 schools. Unfortunately, graduates from vocational schools do not fulfill the minimum requirements set by industries, particularly in the current era of industry revolution 5.0. This revolution aims to establish society 5.0, where humans and robots collaborate closely to achieve improved work outcomes. One technique to enhance the proficiency of graduates and prepare them for the workforce is by implementing a mixed-reality system. that will effectively address a multitude of issues and significantly enhance the caliber of graduates and before the implementation of mixed reality (MR) systems, it is necessary to create system architecture diagrams to ensures that the system can be utilized not only in specific schools but also in any vocational school in Indonesia. This study comprises 5 participants, including experts from both the professional and academic fields, who possess extensive knowledge in the domains of metaverse, MR systems, and information systems. The methodology employed in this study draws inspiration from James Martin’s rapid application development (RAD). The result of this study is a validated system architectural diagram, endorsed by experts, which depicts a metaverse-based MR system designed specifically for vocational schools in Indonesia
Volume: 14
Issue: 1
Page: 207-215
Publish at: 2025-04-01

OCNet-23: a fine-tuned transfer learning approach for oral cancer detection from histopathological images

10.11591/ijece.v15i2.pp1826-1833
Amatul Bushra Akhi , Abdullah Al Noman , Sonjoy Prosad Shaha , Farzana Akter , Munira Akter Lata , Rubel Sheikh
Oral squamous cell carcinoma (OSCC) is emerging as a significant global health concern, underscoring the need for prompt detection and treatment. Our study introduces an innovative diagnostic method for OSCC, leveraging the capabilities of artificial intelligence (AI) and histopathological images (HIs). Our primary objective is to expedite the identification process for medical professionals. To achieve this, we employ transfer learning and incorporate renowned models such as VGG16, VGG19, MobileNet_v1, MobileNet_v2, DenseNet, and InceptionV3. A key feature of our approach is the meticulous optimization of the VGG19 architecture, paired with advanced image preprocessing techniques such as contrast limited adaptive histogram equalization (CLAHE) and median blur. We conducted an ablation study with optimized hyperparameters, culminating in an impressive 95.32% accuracy. This groundbreaking research ensures accurate and timely diagnoses, leading to improved patient outcomes, and represents a significant advancement in the application of AI for oral cancer diagnostics. Utilizing a substantial dataset of 5,192 meticulously categorized images into OSCC and normal categories, our work pioneers the field of OSCC detection. By providing medical professionals with a robust tool to enhance their diagnostic capabilities, our method has the potential to revolutionize the sector and usher in a new era of more effective and efficient oral cancer treatment.
Volume: 15
Issue: 2
Page: 1826-1833
Publish at: 2025-04-01

Enhanced intrusion detection through dual reduction and robust mean

10.11591/ijai.v14.i2.pp1260-1270
Archi Taha , Mohammed Benattou
The exponential growth of online networks necessitates a paradigm shift in intrusion detection systems (IDS). Traditionalmethods falter under the massive influx of data, resulting in high false positives and reduced detection accuracy. This research introduces a novel approach combining principal component analysis (PCA) and linear discriminant analysis (LDA), augmented by robust generalized sample mean, to enhance IDS performance. PCA efficiently reduces data dimensionality, while LDA extracts critical features that differentiate normal network traffic from anomalies. The robust generalized sample mean counteracts the influence of outliers, ensuring accurate and reliable analysis. Implemented on the UNSW-NB15 dataset, our method achieves an average 6% reduction in false positives and a 10% increase in detection rate. Additionally, our testing methodology closely mirrors real-world conditions, making the results more representative of practical scenarios compared to existing work. These advancements demonstrate substantial improvements in IDS performance and robustness over existing techniques.
Volume: 14
Issue: 2
Page: 1260-1270
Publish at: 2025-04-01

A constrained convolutional neural network with attention mechanism for image manipulation detection

10.11591/ijece.v15i2.pp2304-2313
Kamagate Beman Hamidja , Fatoumata Wongbé Rosalie Tokpa , Vincent Mosan , Souleymane Oumtanaga
The information disseminated by online media is often presented in the form of images, in order to quickly captivate readers and increase audience ratings. However, these images can be manipulated for malicious purposes, such as influencing public opinion, undermining media credibility, disrupting democratic processes or creating conflict within society. Various approaches, whether relying on manually developed features or deep learning, have been devised to detect falsified images. However, they frequently prove less effective when confronted with widespread and multiple manipulations. To address this challenge, in our study, we have designed a model comprising a constrained convolution layer combined with an attention mechanism and a transfer learning ResNet50 network. These components are intended to automatically learn image manipulation features in the initial layer and extract spatial features, respectively. It makes possible to detect various falsifications with much more accuracy and precision. The proposed model has been trained and tested on real datasets sourced from the literature, which include MediaEval and Casia. The obtained results indicate that our proposal surpasses other models documented in the literature. Specifically, we achieve an accuracy of 87% and a precision of 93% on the MediaEval dataset. In comparison, the performance of methods from the literature on the same dataset does not exceed 84% for accuracy and 90% for precision.
Volume: 15
Issue: 2
Page: 2304-2313
Publish at: 2025-04-01

Key management for bitcoin transactions using cloud based key splitting technique

10.11591/ijece.v15i2.pp1861-1867
Amar Buchade , Nakul Sharma , Varsha Jadhav , Jagannath Nalavade , Suhas Sapate , Rajani Sajjan
Bitcoin wallet contains the information which is required for making transactions. To access this information, user maintains the secret key. Anyone with the secret key can access the records stored in bitcoin wallet. The compromise of the key such as physical theft, side channel attack, sybil attack, DoS attack and weak encryption can cause the access of transactional details and bitcoins stored in the wallet to the attacker. The cloud-based key split up technique is proposed for securing the key in blockchain technology. The key shares are distributed across virtual machines in cloud computing. The approach is compared to the existing key management approaches such as local key storage, keys derived from password and hosted wallet. It is observed that our approach is most suitable among the other key management approaches.
Volume: 15
Issue: 2
Page: 1861-1867
Publish at: 2025-04-01

Buffers balancing of buffer-aided relays in 5G non-orthogonal multiple access transmission internet of things networks

10.11591/ijece.v15i2.pp1774-1782
Mohammad Alkhwatrah , Nidal Qasem
Buffer-aided cooperative non-orthogonal multiple access (NOMA) enhances the efficiency of utilizing the spectral by allowing more users to share the same re- sources to establish massive connectivity. This is remarkably attractive in the fifth generation (5G) and beyond systems, where a massive number of links is essential like in the internet of things (IoT). However, the capability of buffer co-operation in reducing the outage is limited due to empty and full buffers, where empty buffers can not transmit and full buffers can not receive data packets. Therefore, in this paper, we propose balancing the buffer content of the inter-connected relays, so the buffers that are more full send packets to the emptier buffers, hence all buffers are more balanced and farther from being empty or full. The simulations show that the proposed balancing technique has improved the network outage probability. The results show that the impact of the balancing is more effective as the number of relays in the network is increased. Further- more, utilizing the balancing with a lower number of relays may lead to better performance than that of more relays without balancing. In addition, giving the balancing different levels of priorities gives different levels of enhancement.
Volume: 15
Issue: 2
Page: 1774-1782
Publish at: 2025-04-01

Predicting stock prices using ensemble learning techniques

10.11591/ijece.v15i2.pp1783-1792
Salma Elsayed , Ahmad Salah , Ibrahim Elhenawy , Marwa Abdellah
Stock price prediction has grown in importance due to its role in determining the future worth of business shares. There are several approaches for stock price prediction that can be classified into machine learning, deep learning, and ensemble learning methods. To predict stock prices, we proposed collecting a dataset for different well-known stocks, e.g., Microsoft. The utilized datasets consist of two parts; the first part contains a set of tweets for the stocks under investigation in this study which were collected from the X social media platform and the other part contains the stock prices. Sentimental features of the tweets were extracted and merged with the stock price changes. Then, we framed the problem as a regression task. we aim to analyze the performance gap between ensemble learning and other machine learning (ML) and deep learning (DL) models for predicting stock prices based on tweets. In this context, different ensemble learning models were proposed to predict the price change of each stock. Besides, several machine learning and deep learning models were used for comparison purposes. Several evaluation metrics were utilized to evaluate the performance of the proposed models. The experimental results proved that the stacking regressor model outperformed the other models.
Volume: 15
Issue: 2
Page: 1783-1792
Publish at: 2025-04-01

Kafka-machine learning based storage benchmark kit for estimation of large file storage performance

10.11591/ijece.v15i2.pp1990-1999
Sanjay Kumar Naazre Vittal Rao , Anitha Chikkanayakanahalli Lokesh Kumar , Subhash Kamble
Efficient storage and maintenance of big data is important with respect to assuring accessibility and cost-friendliness to improve risk management and achieve an effective comprehension of the user requirements. Managing the extensive data volumes and optimizing storage performance poses a significant challenge. To address this challenge, this research proposes the Kafka-machine learning (ML) based storage benchmark kit (SBK) designed to evaluate the performance of the file storage system. The proposed method employs Kafka-ML and a drill-down feature to optimize storage performance and enhance throughput. Kafka-ML-based SBK has the capability to optimize storage efficiency and system performance through space requirements and enhance data handling. The drill-down search feature precisely contributes through reducing disk space usage, enabling faster data retrieval and more efficient real-time processing within the Kafka-ML framework. The SBK aims to provide transparency and ease of utilization for benchmarking purposes. The proposed method attains maximum throughput and minimum latency of 20 MBs and 70 ms, respectively on the number of data bytes is 10, as opposed to the existing method SBK Kafka.
Volume: 15
Issue: 2
Page: 1990-1999
Publish at: 2025-04-01

Optimization techniques applied on image segmentation process by prediction of data using data mining techniques

10.11591/ijece.v15i2.pp2161-2171
Ramaraj Muniappan , Srividhya Selvaraj , Rani Vanathi Gurusamy , Velumani Thiyagarajan , Dhendapani Sabareeswaran , David Prasanth , Varadharaj Krithika , Bhaarathi Ilango , Dhinakaran Subramanian
The research work presents an enhanced method that combines rule-based color image segmentation with fuzzy density-based spatial clustering of applications with noise (FDBSCAN). This technique enhances super-pixel robustness and improves overall image quality, offering a more effective solution for image segmentation. The study is specifically applied to the challenging and novel task of predicting the age of tigers from camera trap images, a critical issue in the emerging field of wildlife research. The task is fraught with challenges, particularly due to variations in image scale and thickness. Proposed methods demonstrate that significant improvements over existing techniques through the broader set of parameters of min and max to achieve superior segmentation results. The proposed approach optimizes segmentation by integrating fuzzy clustering with rule-based techniques, leading to improved accuracy and efficiency in processing color images. This innovation could greatly benefit further research and applications in real-world scenarios. Additionally, the scale and thickness variations of the present barracuda panorama knowledge base offer many advantages over other enhancement strategies that have been proposed for the use of these techniques. The experiments show that the proposed algorithm can utilize a wider range of parameters to achieve better segmentation results.
Volume: 15
Issue: 2
Page: 2161-2171
Publish at: 2025-04-01
1 2 3
...
2097 2098
Show 336 of 2098

Discover Our Library

Embark on a journey through our expansive collection of articles and let curiosity lead your path to innovation.

Explore Now
Library 3D Ilustration