Journal of ICT Research and Applications
Not a member yet
359 research outputs found
Sort by
WSN-IoT Forecast: Wireless Sensor Network Throughput Prediction Framework in Multimedia Internet of Things
Get PDFAccurate throughput predictions can significantly improve the quality of experience (QoE), where QoE denotes a network’s capacity to provide satisfactory service. By increasing the results of good throughput predictions, the best strategy can be planned for managing data transmission networks with the aim of better and faster data transmission, thereby increasing QoE. Consequently, this paper investigates how to predict the throughput of wireless sensor networks utilizing multimedia data. First, we conducted a comparative analysis of relevant prior research on the topic of throughput prediction in Multimedia Internet of Things (Multimedia IoT). We developed a throughput prediction framework for wireless sensor networks based on what we learned from these studies using machine learning. The Throughput Prediction Framework identifies historical throughput data and employs these traits to predict throughput. In the final phase, multiple camera nodes and local servers are utilized to test a framework for throughput prediction. Our analysis demonstrates that WSN-IoT predictions are quite precise. For a 1-second time breakdown, the average absolute percentage error for all investigated scenarios ranges from 1 to 8 percent
Utilizing Generative Adversarial Network for Synthetic Image Generation to Address Imbalance Challenges in Chest X-Ray Image Classification
Get PDFDeep learning-based classifiers need lots of image data to train. Unfortunately, not all real-world cases are supported by a huge amount of image data. One of the cases are images for classification of pneumonia infections with chest X-rays images. This study proposes a way of synthesizing chest X-rays with abnormal conditions in order to use the synthesized images for classification purposes. A GAN-based technique can generate synthetic images with greater quality that resemble original images thus can provide a more balanced data distribution than other approaches. To indirectly evaluate the quality of our GAN-based synthetic images, we used CNN-based classification architectures on diverse datasets. Three scenarios examined the effects of synthetic picture categorization. Scenario-1: adding 90% of synthesized images to the original images into the training dataset. Scenario-2: adding 50% of synthesized images to the original images. Scenario-3: adding 10% of synthesized image to the original images. The classification test revealed significantly increased F1 scores in all scenarios. Our study also emphasizes the significance of addressing the problem of imbalanced collections of chest X-ray images and the capability of GANs to alleviate this issue
Sociable Robot ‘Lometh’: Exploring Interactive Regions of a Product-Promoting Robot in a Supermarket
Get PDFThe robot ‘Lometh’ is an information-presenting robot that naturally interacts with people in a supermarket environment. In recent years, considerable effort has been devoted to the implementation of robotic interfaces to identify effective behaviors of communication robots focusing only on the social and physical factors of the addresser and the hearer. As attention focus and attention target shifting of people differs based on the human visual focus and the spatiality, this study considered four interactive regions, considering the visual focus of attention as well as the interpersonal space between robot and human. The collected primary data revealed that 56% attention shifts occurred in near peripheral field of view regions and 44% attention shifts in far peripheral field of view regions. Using correspondence analysis, we identified that the bodily behaviors of the robot showed the highest success rate in the left near peripheral field of view region. The verbal behaviors of the robot captured human attention best in the right near peripheral field of view region. In this experiment of finding a socially acceptable way to accomplish the attention attracting goals of a communication robot, we observed that the robots’ affective behaviors were successful in shifting human attention towards itself in both left and right far- peripheral field of view regions, so we concluded that for far field of view regions, designing similar interaction interventions can be expected to be successful
Improving the Performance of Low-resourced Speaker Identification with Data Preprocessing
Get PDFAutomatic speaker identification is done to tackle daily security problems. Speech data collection is an essential but very challenging task for under-resourced languages like Burmese. The speech quality is crucial to accurately recognize the speaker’s identity. This work attempted to find the optimal speech quality appropriate for Burmese tone to enhance identification compared with other more richy resourced languages on Mel-frequency cepstral coefficients (MFCCs). A Burmese speech dataset was created as part of our work because no appropriate dataset available for use. In order to achieve better performance, we preprocessed the foremost recording quality proper for not only Burmese tone but also for nine other Asian languages to achieve multilingual speaker identification. The performance of the preprocessed data was evaluated by comparing with the original data, using a time delay neural network (TDNN) together with a subsampling technique that can reduce time complexity in model training. The experiments were investigated and analyzed on speech datasets of ten Asian languages to reveal the effectiveness of the data preprocessing. The dataset outperformed the original dataset with improvements in terms of equal error rate (EER). The evaluation pointed out that the performance of the system with the preprocessed dataset improved that of the original dataset
Enhancing Skin Disease Diagnosis Through Fine-Tune Convolutional Neural Network: A Comparative Study with Multi-class Approach
Get PDFDue to their similar appearance, skin disorders frequently disguise their early warning signs from our skin, which is the defense system of the body. Preventing serious disorders requires their early detection. This work investigated the use of fine-tune transfer learning as a fast and accurate way to diagnose skin diseases. To classify different skin issues, we used pre-trained models, i.e., InceptionV3, DenseNet201, and Xception. This work examined 17,500 photos from three sources. It was found that fine-tune Xception performed exceptionally well, with an accuracy rate of 99.14%. It was closely followed by DenseNet201 and InceptionV3, each with different processing speeds, 98.74% and 98.46%, respectively. We used transfer learning with data sets validated by medical experts, outperforming earlier research in precision. This more accurate detection of skin diseases could greatly improve patient outcomes and expedite medical procedures. This approach is new in that it fine-tunes transfer learning by utilizing a vast number of data to increase accuracy compared to other researcher works
The Utility of Decision Tree and Analytics Hierarchy Process in Prioritizing of Social Aid Distribution due to Covid-19 Pandemic in Indonesia
Get PDFThe Indonesian government provided various social assistance programs to local governments during Covid-19. One of the difficulties for the local governments in determining candidates for social aid is ensuring that the number of candidates is in balance with the available quota. Therefore, the local governments must select the most eligible candidates. This study proposes a priority model that can provide recommendations for candidates who meet the criteria for social assistance. The six parameters used in this study were: number of dependents, occupation, income, age, Covid status, and citizen status. The model operates in two stages, namely classification followed by ranking. The classification stage is conducted using a decision tree, while the ranking stage is performed conducted using the Analytical Hierarchy Process (AHP) algorithm. The decision tree separates two classes, namely, eligible and non-eligible. In addition, the classification process is also used to determine the dominant attributes and played a role in the modeling. The proposed model generates a list of the most eligible candidates based on our research. These are sorted by weight from greatest to most eligible using five dominant parameters: number of dependents, income, age, Covid status, and citizen status
Emergency Data Transmission Mechanism in VANETs using Improved Restricted Greedy Forwarding (IRGF) Scheme
Get PDFOne of the most critical tasks in Vehicular Ad-hoc Networks (VANETs) is broadcasting Emergency Messages (EMs) at considerable data delivery rates (DDRs). The enhanced spider-web-like Transmission Mechanism for Emergency Data (TMED) is based on request spiders and authenticated spiders to create the shortest route path between the source vehicle and target vehicles. However, the adjacent allocation is based on the DDR only and it is not clear whether each adjacent vehicle is honest or not. Hence, in this article, the Improved Restricted Greedy Forwarding (IRGF) scheme is proposed for adjacent allocation with the help of trust computation in TMED. The trust and reputation score value of each adjacent vehicle is estimated based on successfully broadcast emergency data. The vehicles’ position, velocity, direction, density, and the reputation score, are fed to a fuzzy logic (FL) scheme, which selects the most trusted adjacent node as the forwarding node for broadcasting the EM to the destination vehicles. Finally, the simulation results illustrate the TMED-IRGF model’s efficiency compared to state-of-the-art models in terms of different network metrics
Enhanced Relative Comparison of Traditional Sorting Approaches towards Optimization of New Hybrid Two-in-One (OHTO) Novel Sorting Technique
Get PDFIn the world of computer technology, sorting is an operation on a data set that involves ordering it in an increasing or decreasing fashion according to some linear relationship among the data items. With the rise in the generation of big data, the concept of big numbers has come into existence. When the number of records to be sorted is limited to thousands, traditional sorting approaches can be used; in such cases, complexities in their execution time can be ignored. However, in the case of big data, where processing times for billions or trillions of records are very long, time complexity is very significant. Therefore, an optimized sorting technique with efficient time complexity is very much required. Hence, in this paper an optimized sorting technique is proposed, named Optimized Hybrid Two-in-One Novel Sorting Technique (OHTO, a mixed approach of the Insertion Sort technique and the Bubble Sort technique. The proposed sorting technique uses the procedure of both Bubble Sort and Insertion Sort, resulting in fewer comparisons, fewer data movements, fewer data insertions, and less time complexity for any given input data set compared to existing sorting techniques
Machine Learning-based Early Detection and Prognosis of the Covid-19 Pandemic
Get PDFThe outbreak of Covid-19 has caused a global health crisis, presenting numerous challenges to the healthcare system with its severe respiratory symptoms and variable presentation. Early and accurate diagnosis of the virus is critical in controlling its spread and reducing the burden on healthcare facilities. To address this issue and relieve the strain on the healthcare system, this paper proposes a machine learning-based approach for Covid-19 diagnosis. Four algorithms were used for analyzing early Covid-19 detection, i.e., logistic regression, random forest, decision tree, and naive Bayes, using a data set of basic symptoms such as fever, shortness of breath, etc. for predicting positive and negative Covid-19 cases. Furthermore, development of a web portal that provides information on global vaccine distribution and the most widely used vaccines by country along with Covid-19 predictions. Our evaluation results demonstrate that the decision tree model outperformed the other models, achieving an accuracy of 97.69%. This study provides a practical solution to the ongoing Covid-19 crisis through an improved diagnosis method and access to vaccination information
The Potential of a Low-Cost Thermal Camera for Early Detection of Temperature Changes in Virus-Infected Chili Plants
Get PDFOne effect of viral infection on plant physiology is increased stomata closure so that the transpiration rate is low, which in turn causes an increase in leaf temperature. Changes in plant leaf temperature can be measured by thermography using high-resolution thermal cameras. The results can be used as an indicator of virus infection, even before the appearance of visible symptoms. However, the higher the sensor resolution of the thermal camera, the more expensive it is, which is an obstacle in developing the method more widely. This article describes the potential of thermography in detecting Tobacco mosaic virus infection in chili-pepper plants using a low-cost camera. A FLIR C2 camera was used to record images of plants in two treatment groups, non-inoculated (V0) and virus-inoculated plants (V1). Significantly, V1 had a lower temperature at 8 and 12 days after inoculation (dai) than those of V0, but their temperature was higher than V0 before symptoms were visible, i.e., at 17 dai. Thermography using low-cost thermal cameras has potency to detect early viral infection at 8 dai with accuracy levels (AUC) of 80.0% and 86.5% based on k-Nearest Neighbors and Naïve Bayes classifiers, respectively