Keywords: Sentiment analysis, Behavior engagement estimation, Emotion recognition and psychological analysis
Abstract: This paper introduces a new cyberbullying dataset that includes Twitter threads containing both the main posts and the replies from bystanders. The dataset is organized based on conversation ID and consists of 112 threads, totaling around 639 tweets. The unique aspect of this dataset is the inclusion of labels for bystanders’ roles, which provides a comprehensive understanding of the bullying incident and helps identify the level of aggressiveness in cyberbullying. This type of information is not available in existing datasets that only label isolated tweets. By incorporating bystanders’ roles, annotators gain a deeper understanding of real-world scenarios, leading to improved machine learning performance and better classification of cyberbullying. The dataset is freely available, promoting collaboration among researchers, ensuring result reliability, and enabling the reuse of Twitter datasets. It also offers a cost-effective way for non-technical researchers to leverage Twitter data in their scientific investigations.

Keywords: Emotion recognition and psychological analysis, Behavior engagement estimation
Abstract: This study presents findings from an analysis of a newly developed corpus focused on cyberbullying, aiming to comprehensively examine labeled cyberbullying threads on the social media platform Twitter, with a specific emphasis on the role of bystanders. Previous corpora used for automatic cyberbullying detection have primarily focused on the main posts, disregarding the threaded responses. Consequently, these studies have overlooked valuable information regarding the involvement of bystanders, which is crucial for enhancing the accuracy of cyberbullying detection. This study addresses this gap by incorporating bystander roles within the corpus, resulting in significant impact on annotators’ perception and classification of cyberbullying instances. The findings suggest promising prospects for improved automated cyberbullying detection. Notably, the most frequently observed bystander roles align with the content of the main post. Surprisingly, impartial bystanders are most prevalent in cyberbullying threads characterized by high levels of aggression. This article provides a detailed analysis of the annotation process and examines the influence of bystanders roles in greater depth.

Keywords: Behavior engagement estimation, Facial expression and gesture recognition and analysis, Emotion recognition and psychological analysis
Abstract: The final objective of this research project is to develop a lecture archive summarization platform that can extend learners' experience by automatically estimating and providing temporal and spatial ROI (Regions of Interest) according to multimodal features and learners' learning logs in lecture archives that record face-to-face lectures. To develop this platform, we (a) establish a method for extracting spatiotemporal multimodal features of lecture archives and (b) construct a method for estimating a learning style model based on the learning logs when watching the archives with the extracted features. Furthermore, to maximize the learning effect, we will (c) develop a prototype system to adaptively control the spatiotemporal ROI at the terminal side according to the learning style model as an adaptive summarization. This article describes the concept of the proposed platform and the initial analysis of learners' learning logs.

Keywords: Sentiment analysis, Emotion recognition and psychological analysis
Abstract: YouTube is a online video sharing service that is that many people view. If we can know the emotions of video viewers, we can use them to improve usability. In this study, we propose two method for estimating the emotion of YouTube videos from the comments of each video: a BERT-based method and a rule-based method. For the former, we use BERT with fine-turning by 350 video comments to estimate the emotion. In the rule-based method, emotion values are calculated using a Japanese emotional expression dictionary. To evaluate these methods, we obtained emotion values from 100 respondents who were asked to fill out questionnaires as ground truth. The results of the evaluation using cosine similarity showed that BERT-method was able to estimate emotions with higher accuracy than the rule-based method.

Keywords: Machine Learning, Data Mining, Neural Networks and Deep Learning
Abstract: Pineapple's size is very crucial in determining its market value. Size sorting is commonly done via visual inspection, which is usually subject to inconsistency and errors. Errors due to failed sorting may either lead to wastage or loss, or mispricing. This study presents incorporation of the machine learning techniques like Logistic Regression, K-Nearest Neighbors, Support Vector Machine, Decision Tree, and Random Forest in classifying pineapple sizes as small, medium, and large using the extracted features of images processed via OpenCV libraries as well as Python Programming. A total of 300 pineapples of different sizes were captured and processed to extract features such as the area, width, height, enclosed-circle radius, and perimeter. The models were optimized using GridSearchCV and were evaluated using accuracy and F1 score metrics. Based on the results, SVM was found to be the most suited classification model, having an optimized training and testing accuracy of 95.67 % and 96.67 %, respectively, and an F1 score of 96.67 %.

Keywords: Machine Learning, Neural Networks and Deep Learning, Data Mining
Abstract: The lack of a transport quality forecasting system in farming and sorting facilities of indigenous varieties of bananas is aiding the increase of food waste generation in the country. This in turn decreases agriculture sustainability and imposes economic losses to farmers. Musa acuminata ‘Señorita’ are diploid cultivars of bananas originating in the Philippines. This study aims to develop a machine learning-based system that classifies Musa acuminata ‘Señorita’ bananas into their transport tiers: (I) for interprovincial distribution, (II) for intra provincial distribution, or (III) subject for rejection. The model is trained and went through 7 machine learning classifiers to identify which model is the most compatible with the system design. The application of external parameters such as size, girth, weight, maturity stage, and RGB parameters can be the foundation to develop a machine learning-based banana transport tier classifier that accurately monitors and determines how long they can travel based on their maturity level. Among the seven models, Logistic Regression, Linear Discriminant Analysis, Decision Tree Classifier, Gaussian Naive Bayes, and Support Vector Machine attained a classification accuracy of 100 %. The development of this system can aid in the proper distribution of produce and help uplift the country’s agriculture and economic sustainability.

Keywords: Machine Learning, Neural Networks and Deep Learning
Abstract: The utilization of herbal medicinal plants dates to antiquity, and as human civilization has progressed and technology has advanced, a significant proportion of contemporary medicines have originated from herbal sources. The Philippines is renowned for its extensive utilization of herbal medicinal plants, exemplified by the Department of Health's recognition of ten prominent herbal medicines under the “Traditional and Alternative Medicine Act”, or the “Republic Act No. 8423”. This legislative measure not only enhances the healthcare system within the country but also underscores the nation's commitment to incorporating traditional healing methods. Many herbal medicinal plants possess valuable therapeutic properties; however, the lack of comprehensive research and clinical trials has resulted in limited knowledge regarding the specific benefits associated with each of these plants. This research endeavor culminated in developing a device capable of identifying the herbal medicinal plant name, scientific name, and medicinal purposes by capturing a leaf image. This study's initial phase entails pre-processing the captured image and extracting the leaf vein characteristics using the Histogram of Oriented Gradient feature extraction algorithm. Subsequently, the Convolutional Neural Network algorithm is employed to identify the leaf based on these extracted features. After conducting 25 trials for each plant, the research findings demonstrated an accuracy rate of 95%.

Keywords: Machine Learning, Neural Networks and Deep Learning
Abstract: Pests and diseases are significant threats to the agriculture industry, jeopardizing crop production and the livelihoods of farmers. Among these challenges is Coffee Leaf Rust (CLR), a devastating fungal disease that specifically affects coffee trees, leading to defoliation and yield loss. Early detection of diseases is crucial to mitigate the impact and preserve the value of crops. A proposed solution is a diagnostic model for Coffea liberica, a coffee species, by integrating Wireless Sensor Network and Deep Learning Algorithm. The system utilizes wireless sensors placed strategically in the plantation to collect real-time data on plant health and environmental parameters. This data is then processed using advanced Deep Learning algorithms that analyze patterns and accurately identify CLR and other diseases. By detecting diseases at an early stage, farmers can take proactive measures, such as targeted treatments and cultural practices, to mitigate the impact of CLR and improve overall plant health. This approach not only safeguards the livelihoods of farmers but also ensures the sustainable production of high-quality coffee. The integration of Wireless Sensor Network and Deep Learning empowers farmers with timely information, enabling them to make informed decisions and implement effective disease management strategies. This diagnostic model serves as a valuable tool for monitoring and safeguarding the health of coffee plants, enhancing crop productivity, and ultimately cont

Keywords: Optical Communications and Networks, Wireless Communications and Networks, Modulation and Coding Techniques
Abstract: Free space optics (FSO) communication operates over the unlicensed spectrum providing high throughput, high security, and ease of installation. FSO can solve the last mile problem of connectivity with wide applications, i.e., disaster recovery, campus connectivity, backhaul connection, etc. However, the deployment scenario is constrained due to the severe atmospheric conditions and a precise line of sight requirements. This research work analyzes the performance of FSO communication under the combined effect of pointing errors (PEs) and atmospheric turbulence. The effects of jitter and boresight for PEs are modelled with generalized Nakagami-m distribution. A closed-form average bit error rate (BER) is derived for the generalized phenomenon and the theoretical results are compared with simulation and experimental results. Also, an experimental testbed is designed and implemented under a controlled indoor environment to analyze the atmospheric turbulence affects on FSO communication.

Keywords: Optical Communications and Networks, Wireless Communications and Networks, Modulation and Coding Techniques
Abstract: Visible light communication (VLC) is secured and high-speed communication for a specified area location. This provides a green-fuzzy technology for optical wireless communication system design. The VLC channel included different parameters, where the variation of the parameter can be analyzed from different channel estimation techniques. The VLC channel is affected by both static and mobile environments. Previously most of the research is on static VLC channels but recently mobile VLC communication is in trend and has more applications in the real world. Thermal noise, ambient noise, shot noise, the distance of the transceiver, and the position of the receiver are some parameters, which has caused a major effect on the VLC Channel estimation system. So many estimation techniques for VLC channels are available to estimate the performance of communication links. In this paper, a detailed comparative analysis is given for both DC-biased optical orthogonal frequency division multiplexing (DCO OFDM) and asymmetrically clipped orthogonal frequency division multiplexing (ACO OFDM). Here, a novel VLC channel estimation technique for multiple-input-multiple-output (MIMO) ACO OFDM is designed, where the performance of linear minimum mean square error (LMMSE) channel estimation technique can be concluded with the bit error rate (BER) response. The paper considers least square (LS) and LMMSE channel estimation techniques to compare the proposed work with the verified work.

Keywords: Wireless Communications and Networks, Ad-hoc, Mesh and Sensor Networks
Abstract: The demand for firefighting has significantly risen in recent decades, accompanied by increased risks faced by firefighters. Tragic incidents, such as the Shanghai factory fires, have resulted in the loss of over thirty firefighter lives. One of the primary contributing factors is the abrupt breakdown in communication between firefighters inside a building and the commanding officer stationed outside, attributable to the harsh and complex indoor environment. This study aims to conduct a comparative analysis of different widely used wireless transmission methods, including Wi-Fi, Bluetooth Low Energy (BLE), and Long Range (LoRa). The experiments are conducted in a two-room setup, with two brick walls acting as a barrier. A wireless data transmitter is placed in one room, while smoke is generated. A receiver placed at varying distances collects the signal strengths. The findings indicate that LoRa exhibits the least drop in signal strength compared to the other methods. In contrast, BLE shows high signal strength variation for the same distances and is not recommended for firefighting communication purposes. This study provides valuable insights for selecting suitable wireless communication modules, particularly in the design of wearable devices for assessing safety risks faced by firefighters.

Keywords: Optical Communications and Networks, Vehicular Networks, Cellular Networks
Abstract: This paper considers mixed radio frequency (RF) and hybrid free space optics (FSO)/RF communications for information transfer between vehicles and the network where multiple infrastructures nodes, such as street lights, traffic lights, signboards, etc., act as relay units. Vehicles communicate with nearby infrastructure nodes in the first transmission phase using an RF link. In the next phase, an infrastructure node corresponding to the maximum signal-to-noise ratio (SNR) is selected to forward the signal to the base station using a hybrid FSO/RF link. An RF link is used as a backup connection to increase the system's reliability. To get practical insights, atmospheric turbulence-induced fading, pointing errors, and atmospheric attenuation, which may affect the FSO link's performance, are considered. In addition, the base station is deployed with multiple antennas. It exploits maximum-ratio combining when the RF link is active to improve end-to-end system performance. We derive closed-form expressions for the outage probability and system throughput to analyze the system's performance. We include the simulation results to verify that all the derived analytical findings are accurate.

Keywords: Micro Grids& Distributed Generation, Power Generation, Transmission and Distribution, Renewable Energy Sources and Technology
Abstract: A robust decentralized control method is proposed in this paper for an islanded DC microgrid. A state-space model of the islanded DC microgrid system is derived based on the small-signal model of the system. Parametric uncertainties like load resistance, filter inductance and capacitance are modelled using the upper linear fractional transformation technique. A loop-shaping H infinity controller is designed to ensure robust stability and satisfaction of desired performance criteria. The proposed control technique is applied to an islanded DC microgrid test system comprising a dispatchable distributed generation unit, a photovoltaic unit following the maximum power point tracking algorithm, and a battery energy storage system unit. Simulation studies validate the efficacy of the proposed control approach for an islanded DC microgrid.

Keywords: Micro Grids& Distributed Generation, Power Generation, Transmission and Distribution
Abstract: The penetration of embedded generation, including renewable power sources (wind and solar), is gradually increasing in power distribution networks. Also, the transition from conventional fossil fuel-based transportation to e-transportation has introduced electric vehicle charging stations as a new load class. The conventional distribution system architecture alteration has made the system operation rather challenging. Therefore, an efficient energy management scheme is crucial to the satisfactory operation of an active distribution system from techno-economic considerations. This paper proposes an optimal operating strategy to simultaneously minimize the operating cost, average voltage deviation, and line loadings and improve the voltage stability of an active distribution network. The distribution system is assumed to have a soft open point and smart transformer for smooth active and reactive power control. The demand response flexibility (offered by responsive electrical demands and public and residential electric vehicle charging stations) is coordinated by controlling a smart transformer and a soft open point to realize multiple objectives. The multi-objective problem is solved in the fuzzy domain using a combination of linear programming and particle swarm optimization. Simulation results on a sixty-nine-bus radial distribution system validate the proposed method’s effectiveness.

Keywords: Power Generation, Transmission and Distribution, Power System Monitoring, Control and Protection, Renewable Energy Sources and Technology
Abstract: With decreasing rooftop photovoltaic (PV) costs and increased incentives to generate electricity, more and more end-users are installing rooftop PV systems and availing net metering. However, as rooftop PV and net metering become more prevalent, electricity theft detection becomes more challenging. This paper investigates the performance of features and algorithms used in classification-based theft detection algorithms on systems with rooftop PV and net metering. We explore five features and four algorithms. We use the following features computed from check meter and individual customer meter readings: gamma deviance (GD), log cosh loss (LCL), percent loss error (PLE), Poisson deviance (PD), and squared error (SE). The meter readings were simulated using the IEEE European Low Voltage Test Feeder using OpenDSS across a wide range of PV, net metering, and theft penetration levels. We then used the extracted features to train classifiers using the following algorithms: support vector machine (SVM), artificial neural network (ANN), k-nearest neighbors (KNN), and decision tree (DT). Test results showed that KNN generally performed poorly, and DT generally performed well. Moreover, models using PD and LCL as features generally displayed robustness to varying levels of PV and net metering. And finally, ANN, SVM, and DT models that use LCL and PD as features are among the highest ranked models in terms of median accuracies and range of accuracy.

Keywords: Switching Circuits & Power Converters, Energy Storage System
Abstract: The effectiveness and efficiency of the method adopted for battery chargers with low input current distortion and better power factor are demanding for Electric Vehicles (EVs). On-Board-Charger (OBC) consists of two stage converters which includes the front end Power Factor Corrected (PFC) stage followed by a DC-DC converter. This paper presents a front end power factor corrected on-board battery charger for EVs. Bridgeless boost converter topology with less number of switches is selected as the PFC stage for battery charger and it is compared with other bridgeless topologies. Front end PFC converter followed by LLC resonant DC-DC converter is used here to develop the battery charger. A 48 V, 24Ah Li-ion battery pack is used as energy storage to drive EV and its charging through the proposed converter scheme is analysed using MATLAB/SIMULINK. The simulation results show the effectiveness of the proposed scheme.

Keywords: Switching Circuits & Power Converters, Renewable Energy Sources and Technology, Energy Storage System
Abstract: The primary goal of this article is to find out best current excitation for controlling maximum torque per amp (MTPA) in internal permanent magnet synchronous motors (IPMSM) with non-sinusoidal back emf. The optimum current elation for the mean torque in the IPMSM is presented in this work in closed form. The suggested work seeks to furnish a universal method for choosing the appropriate current for MTPA control for IPMSM used in EV drive train better than operating vector control for current harmonic injection. Hysteresis current control is also taken into account for current injection in IPMSM. Various examples from experimental investigations are shown, including the optimal current, an FFT analysis of the torque and current profile, and a back emf profile that takes into account the first and third order current harmonics. Finally, this investigation claims an efficient MTPA control strategy under specific operating environment employed in energy-efficient EV drive train.

Keywords: Biomedical and Health Informatics, Bioinformatics, Biomedical Imaging and Video analytics
Abstract: In this paper, we propose a deep learning based pipeline to predict the presence of Glaucoma, a neuro-degenerative eye condition, by applying object localization, image segmentation, and image classification techniques. The pipeline takes a full fundus image as input and provides Optic Disc and Optic Cup masks along with the likelihood measure to determining Glaucoma. The pipeline uses a lightweight U-Net model called ResNet based Disc Localization U-Net (RDLU-Net) for Optic Disc localization and Convolution mixed SegFormer Glaucoma Detection Network (CSGDN) for segmenting Optic Disc (OD) and Optic Cup (OC) structures and finally predicting the probability of the presence of Glaucoma by using the classification network in CSGDN. Our approach shows dice scores of 0.9871 and 0.9065 on OD and OC segmentation, respectively by using REFUGE dataset. The proposed framework obtains accuracy of 95.47% using MobileNet V2 as the detection network when applied on a custom Glaucoma dataset containing data from REFUGE and DRISHTI-GS datasets.

Keywords: Biomedical and Health Informatics, Biomedical Imaging and Video analytics, Biomedical Signal Processing and Instrumentation
Abstract: Functional MRI (fMRI) is widely used to examine brain functionality by detecting alteration in oxygenated blood flow that arises with brain activity. In this study, content-specific image categorization across different visual datasets is performed using fMRI time series (TS) to understand differences in neuronal activities related to vision. Publicly available BOLD5000 dataset is used for this purpose, containing fMRI scans while viewing 5254 images of diverse categories, drawn from three standard computer vision datasets: COCO, ImageNet and SUN. To understand vision, it is important to study how brain functions while looking at different images. To achieve this, spatial encoding of fMRI BOLD TS has been performed that uses classical Gramian Angular Field (GAF) and Markov Transition Field (MTF) to obtain 2D BOLD TS, representing images of COCO, Imagenet and SUN. For classification, individual GAF and MTF features are fed into regular CNN. Subsequently, parallel CNN model is employed that uses combined 2D features for classifying images across COCO, Imagenet and SUN. The result of 2D CNN models is also compared with 1D LSTM and Bi-LSTM that utilizes raw fMRI BOLD signal for classification. It is seen that parallel CNN model outperforms other network models with an improvement of 7% for multi-class classification. The obtained result of this analysis establishes a baseline in studying how differently human brain functions while looking at images of diverse complexities.

Keywords: Biomedical and Health Informatics, Biomedical Signal Processing and Instrumentation, Wearable Sensors for Health care monitoring
Abstract: Parkinson's disease (PD) is a neurological condition that is on the rise and disrupts the nervous system. However, there is no specific diagnosis for Parkinson's disease; only a variety of motor signs can be used to identify it. A speech impairment was found in more than 90% of PD patients. This study presents a voice and speech signal data-based model for PD identification. The PD is the speech data set used in this experiment has a great amount of dimension with very few data points. Different data pretreatment techniques, such as data standardization, multicollinearity diagnostic, and dimensionality reduction approach, were used in our suggested model to enhance the quality of the data. Different Machine Learning (ML) classifiers were employed to categorize PD, including k-nearest Neighbor, Support Vector Machine, Random Forest, AdaBoost, and Logistic Regression. In this experiment, grid search, cross-fold validation, and hyper-parameter tweaking were used to optimize classifier performance and maintain the class distribution of the unbalanced data set. Our suggested model outperformed the prior tests on the same data set by around 98% and reached a maximum accuracy of 98.10%.

Keywords: Biomedical and Health Informatics, Translational Engineering in Healthcare
Abstract: Human Activity Recognition (HAR) exploits WiFi signals to offer behavioral sensing for various practical applications, such as patient monitoring in hospitals, and children/elderly monitoring in smart homes. Particularly for medical applications, HAR is of considerable importance. HAR determines the type of activity precisely and effectively using the correlation between the Channel State Information (CSI) data and physical movements. However, various constraints often make the process challenging for usage in practical systems. Large feature dimensionality and variable activity signal duration are the two major difficulties for developing an efficient HAR framework. Additionally, the frequency components of the CSI activity signal are time-varying, which makes CSI data inherently non-stationary. In this work, we present a feature extraction approach using Discrete Cosine Transform (DCT) to address the aforementioned issues -- we integrate Principal Component Analysis (PCA) based dimensionality reduction and a solution to the non-stationarity problem. We assess the performance of our proposed approach on real data using some off-the-shelf reference models, such as support vector machine (SVM), random forest (RF), k-nearest neighbor (KNN), and convolutional neural network (CNN). Empirical results emphasize the quality of our proposed feature extraction approach by showing that even off-the-shelf classification models perform very well in challenging scenarios.

Keywords: Biomedical and Health Informatics
Abstract: Major depressive disorder (MDD) is a psychiatric disorder but currently defined by symptoms rather than biological mechanism. This in turn sets a huge barrier to effective diagnosis and treatment planning. Investigations were done through neuropathogenesis and neuroimaging analysis as an effort to identify discriminative biomarkers for MDD while understanding the biological dependencies. The literature suggested that microRNA or miRNA transcripts are more likely to deliver substantial predictive power in diagnosis and antidepressant treatment response (ATR) prediction. Yet, there presents discrepancy in unique markers, and such discrepancy might be due to the small sample size over some of the reported studies. This study utilized miRNA as a predictor to model MDD ATR using k-nearest neighbour (kNN). The shortlisted miRNA through feature selection techniques scored 71.20%, 68.13%, 72.13%, and 84.07% for three response levels in accuracy, sensitivity, specificity, and precision, respectively. Synthetic Minority Oversampling TEchnique (SMOTE) was then applied to the shortlisted miRNA and three response levels reported at least 98% in each of the mentioned performance metric.

Keywords: Sentiment analysis, Emotion recognition and psychological analysis
Abstract: Due to the COVID-19 pandemic, most universities endeavored to adopt online education as an alternative to conventional face-to-face classroom instruction. However, capturing students' Temporal Region of Interest (T-ROI) in long-duration video lectures poses a significant challenge. Therefore, lecture archive summarization becomes essential from an online perspective. The results of lecture archive summarization still require further improvement. This research aims to distinguish T-ROI using a speech processing approach h. Our plan is divided into collecting instructors'/presenters' voice datasets, clarifying the T-ROIs through sound processing technology, and building a suitable deep neural network architecture to detect the T-ROIs in the actual lecture archives automatically. We will inevitably encounter various challenges to achieve the objective, such as individual differences. This article describes the experimental dataset collection design considering individual differences and lecture room environments. It summarizes how such efforts will be effective in realizing personalized voice style detection and improving the accuracy of speech processing in real environments.

Keywords: human-computer interaction, pattern recognition and data analysis
Abstract: Vision-based hand gesture recognition (VGR) systems must provide the following functionalities or criteria to control a computer mouse: (i) hand tracking ability, (ii) continuous static and dynamic hand gesture recognition, and (iii) efficient resource management. Our motivation stems from the fact that only a few research so far has accommodated all these three criteria. In this paper, we developed a VGR system that accommodates these three criteria. We propose an algorithm that simultaneously detects and classifies hand gestures using RGB images and hand skeletons. To evaluate our work, we used the IPN dataset which consists of hand gestures that are suitable for mouse control. Compared to previous methods on the IPN dataset, our resulting VGR system achieves better performance in both isolated and continuous hand gesture recognition (HGR). For continuous HGR, we achieved 61.30% Levenshtein accuracy.

Keywords: human-computer interaction, pattern recognition and data analysis
Abstract: Human intention prediction (HIP) is one aspect of Human-Robot Interaction (HRI) that could facilitate understanding and improving how humans interact with robots and computers. However, current gaze-based intent prediction models that perform well often require invasive methods using specialized equipment. In this paper we present a non-invasive, contactless method for predicting human intentions using a multi-stage hybrid CNN-Transformer framework. The model consists of a depth estimator and two key components: a gazed object predictor and a human intent classifier. The gazed object predictor is a modified Detection-Transformer (DETR) and used a ResNet50 backbone for feature extraction and obtained an accuracy of 32.15% in the custom dataset. Meanwhile, the human intent classifier is a transformer-based classifier that achieved a 98% accuracy when predicting human intention based on a series of gazed objects. The resulting cascaded HIP system attained an accuracy of 54%.

Keywords: Facial expression and gesture recognition and analysis, Emotion recognition and psychological analysis, Behavior engagement estimation
Abstract: This study investigates the cultural dependence of a facial expression recognition (FER) system in an interactive agent by analyzing the performance of several recognition models in different cultural domains. A comprehensive cross-domain classification performance assessment reveals disparities in model performance across different cultural contexts, indicating challenges in cross-cultural FER. To further investigate these characteristics, several public datasets across regions and our cross-cultural dataset of facial expressions derived from Thai and Japanese TV shows are analyzed. By evaluating the capacity of existing FER models to interpret our newly collected data, we found significant variations in emotion interpretation across these cultural contexts, highlighting the necessity for culturally inclusive algorithms. These findings underscore the critical need for more consideration of cultural diversity in FER research, marking a crucial step toward more inclusive and culturally sensitive artificial intelligence technologies.

Keywords: Neural Networks and Deep Learning, Machine Learning, Crowd Sourcing & Social Intelligence
Abstract: Edible wild plants are an important source of food in many regions of the world, including in the Philippines, and their recognition is a key skill for survival in the wild. In this study, we propose a mobile platform for image recognition of edible wild plants using deep convolutional neural networks (CNNs). The proposed system is designed to be lightweight and easily deployable on mobile devices, allowing for real-time recognition of edible plants in the field. To develop the system, we first collected a dataset of images of various edible wild plants. The dataset was preprocessed and augmented for better generalization. We then trained a CNN model using transfer learning techniques on a custom specific dataset of edible wild plant images endemic to the Philippines to recognize the different species of edible plants. The trained model was then optimized for deployment on mobile devices, and the resulting mobile application was tested on a variety of wild plants. The results showed that the proposed system achieved high accuracy in identifying edible wild plants, with an average accuracy of 96.98%. The proposed system has many potential applications, including in the field of outdoor education, potential solutions to address food scarcity, and survival training. The proposed mobile platform for image recognition of edible wild plants using CNNs is a promising tool for enhancing the safety and sustainability of foraging and outdoor activities.

Keywords: Neural Networks and Deep Learning, Machine Learning, Human Computer Interface
Abstract: This study introduces a non-invasive egg sexing device that combines artificial intelligence (AI), spectroscopy, and computer vision technology to accurately determine the sex of duck embryos inside eggs. The device utilizes the plumage color as a reliable indicator of sex, employing a hybrid Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) model alongside spectroscopic analysis. Extensive simulations and experiments validate the proposed algorithm, achieving an impressive 98.68% accuracy rate in sex determination, with an average processing time of 37.46 milliseconds, significantly enhancing farming efficiency. Additionally, the research assesses the impact of spectroscopy on egg hatchability, demonstrating a higher hatchability rate of 74.80% within a population of 500 eggs. This finding indicates that spectroscopy does not adversely affect egg viability. Overall, this study presents a sustainable solution for effectively managing male ducklings in the industry, optimizing resource utilization, and mitigating wastage.

Keywords: Neural Networks and Deep Learning, Machine Learning
Abstract: Coffee production is a vital industry in many countries, but diseases affecting coffee leaves can lead to significant losses for farmers. To mitigate these losses, timely disease detection and accurate assessment of disease severity are crucial. This work proposes a deep learning approach for classifying coffee leaf diseases based on their severity levels. The proposed methodology involves several steps. Initially, U²Net removes the background from the coffee leaf images. Subsequently, the background-removed images are converted into BGR format to identify the diseased regions. DeepLabV3 is then trained to extract and mark the diseased portions of the leaves in red. Using these annotated images, various convolutional neural network (CNN) models, including VGG-16, InceptionV3, and MobileNetV2, were trained to classify the diseases based on their severity levels. These models are carefully modified and fine-tuned with hyperparameters to achieve the best performance metrics. Upon model training, the modified MobileNetV2 model performs better than the other CNN models, achieving an impressive F1-score of 97.99%. This outcome highlights the effectiveness of this paper's approach in accurately classifying coffee leaf diseases according to their severity levels. The proposed methodology has significant implications for coffee farmers, enabling them to swiftly detect diseases and assess their severity, allowing for timely and appropriate actions.

Keywords: Neural Networks and Deep Learning, Machine Learning
Abstract: Artificial Neural Networks (ANNs) are increasingly recognized as valuable tools for crop quality parameter measurement. This study investigates the ANNs effectiveness in the predictive measurement of the Chlorophyll-a levels of iceberg lettuce (Lactuca sativa var. capitata). This involved using ANNs to link the dataset of extracted RGB and HSV values with the Chlorophyll-a levels retrieved with UV-VIS spectroscopy. For the prediction model, the RGB and HSV values were used as the 6 input predictor values, while the Chlorophyll-a level was used as the 1 output response value. The ANNs were trained on this dataset using the Levenberg-Marquardt algorithm, where the training data comprised 70% of the dataset, the validation data 20% of the dataset, and the test data 10% of the dataset with a layer size of 15. The ANN model demonstrated a strong correlation between the predicted and target outputs, with an accuracy of 98.02% for the test data. This suggests that ANNs can be employed for an accurate and non-invasive monitoring of parameters in iceberg lettuce. The findings also open possibilities for other crops in the Philippines' agricultural industry.

Keywords: Neural Networks and Deep Learning
Abstract: Deep learning algorithms particularly Convolutional Neural Networks (CNNs) have emerged as state-of-the-art techniques for object detection, image classification, image segmentation and behaviour classification. These algorithms have extensive application across various domains including agriculture. However, cow identification in dairy farming still relies on methods like direct visual monitoring which are time consuming, costly and inaccurate; or use of invasive contact devices such as sensors which can cause discomfort during attachment or removal. This research aims at comparing three deep learning object detection models using cow images generated from video data in a housed dairy cattle barn. The dataset comprises of 5904 cow images and three YOLO variants i.e. YOLOv5, YOLOv7 and YOLOv8 were chosen based on their performance in object detection tasks. Cow images were annotated in YOLO format using makesense AI tool. The models were trained and validated to visualize bounding boxes of the predicted cow objects. Our approach demonstrates the efficiency of the YOLOv8 model, achieving an impressive object detection accuracy of 94.7%. This research makes a significant contribution by shedding light on the future research direction of one stage object detectors, in particular, YOLO algorithm, and highlights the practical implementation of deep learning models for real time cow detection, applicable in livestock management.

Keywords: Wireless Communications and Networks, Ad-hoc, Mesh and Sensor Networks
Abstract: This paper examines the effectiveness of a cognitive radio (CR) network using a conventional energy detector and the selective soft-information fusion rule when confronted with faulty control channels. Using a selective soft-information fusion scheme, we have formulated a mathematical expression that provides a closed-form solution for both the probability of a false alarm and the probability of a missed detection. Several new and extant results are presented as special scenarios in the proposed solution. We also investigate optimal values of the selective soft fusion’s threshold and spectrum-aware CR users by minimizing the CR network’s average error rate. Subsequently, numerical results illustrate the proposed CR network’s theoretical findings.

Keywords: Wireless Communications and Networks, Cellular Networks, Vehicular Networks
Abstract: We address the challenging problem of channel estimation in a narrowband multiuser communication that incorporates intelligent reflecting surfaces (IRS). To overcome the limitations of the two-hop channel in passive IRS system, we adopt a Hybrid IRS architecture, which involves integrating active sensors into a few elements of the IRS. By exploiting the sparse characteristics of the channel, we utilize parallel factor analysis (PARAFAC) tensor models to represent the training signals. We propose an algebraic algorithm to estimate the channel and evaluate the normalized mean square error (NMSE) between the estimated and actual parameters. Additionally, we explore the uniqueness of the canonical polyadic decomposition (CPD) and analyze the essential conditions necessary for achieving accurate channel estimation. Through simulation studies, we validate the effectiveness of our proposed algorithm in estimating channels in a multiuser IRS system.

Keywords: Wireless Communications and Networks
Abstract: Ambient backscatter communication (AmBC) is proposed as a prominent technology to support energy-efficient low power transmissions in 6G and beyond wireless networks. AmBC is capable of providing battery-free connectivity among large scale Internet-of-Things (IoT) devices. Also, it can satisfy the stringent delay and power saving requirements of intelligent transport systems (ITS) in seeking diverse vehicular applications. Motivated by these emerging trends, we consider an AmBCassisted tag-reader-based vehicular scenario and carried out the outage analysis by deriving the exact expressions for the reader’s outage probability (OP) considering tag being in reflective or nonreflective state over mixed Nakagami-m and double Nakagamim fading environment. We also examine the asymptotic OP in the high signal-to-noise (SNR) regime, which is perfectly aligned with the exact OP values even at moderate SNRs. Further, this asymptotic analysis reveals some meaningful insights into the system’s diversity order. Numerical and simulation studies are conducted to validate our mathematical framework.

Keywords: Wireless Communications and Networks
Abstract: This study is driven by the challenge of optimizing connectivity and performance in communication networks with multiple interconnected components. The primary objective is to ensure a consistent and reliable signal strength across the entire coverage area, with a particular focus on monitoring critical infrastructure systems like power transformers. To achieve this goal, an innovative approach is introduced, leveraging the optimization of RF repeaters within these systems. The study combines the power of a path planning algorithm and linear optimization techniques to identify the most suitable locations for repeaters within the RF network. The widely recognized A* (A-Star) path planning algorithm, implemented using the MATLAB Robotics Toolbox, guides this process. Additionally, the Longley-Rice model is utilized to calculate RF data, taking into account factors such as terrain, clutter, and other environmental variables. To enhance the usability and practicality of the optimization process, a user interface (UI) is developed using MATLAB App Designer. This UI enables users to access comprehensive simulation data, including metrics such as signal strength, link margin, SINR (Signal-to-Interference-plus-Noise Ratio), path loss, and distance. By providing detailed insights and analysis, the UI facilitates effective RF repeater placement and optimization within the system. Overall, this approach offers a valuable solution for achieving optimal performance in Power Transformer Mo

Keywords: Wireless Communications and Networks
Abstract: The coexistence of enhanced mobile broadband(eMBB) and ultra-reliable low-latency communication (URLLC) is common in 5G networks. 5G services require eMBB users to achieve higher data rate, and URLLC users to meet high reliability and low latency requirements. How to use limited resources to meet the heterogeneous requirements of eMBB and URLLC is a very meaningful problem. In this paper, We delve into the latency composition of URLLC packets and subsequently derive an expression to determine the number of mini-slots that URLLC packets can be queued. To solve the complex scheduling problem of URLLC packets with different delay requirements, we propose a novel greedy scheduling algorithm based on queuing theory. At each mini-slot, we dynamically schedule the URLLC packets that arrive using the proposed algorithm. We demonstrate the significant advantages of our algorithm through extensive simulations. Specifically, our algorithm significantly reduces the throughput loss of eMBB users, and also meets the high reliability requirements of URLLC in the case of high URLLC load.

Keywords: Power Generation, Transmission and Distribution, Energy Storage System, Renewable Energy Sources and Technology
Abstract: Unmanned Aerial Vehicle (UAV) propulsion system is significantly related to the UAV's flight performance as it is the core of UAV power produced. Propulsion systems include energy sources and power units such as engines and motors. Electric motors are used to generate thrust, which the thrust produced can affect the UAV performance. Thus, this study aims to investigate the best combination of electric motors and propellers for UAVs for high efficiency of thrust production. The experimental work involved electric motors and various small-scaled propellers (fixed and folding) with diameter ranges between 10 to 15 inches using a subsonic speed open loop wind tunnel. The wind tunnel was set to various speeds (0 m/s to 15m/s) to investigate the effects of static and dynamic thrust produced. Then, the efficiency of each motor and propeller combination is calculated based on the thrust produced and the best combination is determined. The significance of this work is to provide a reference when selecting a particular combination propeller (fixed or folding) and motors for specific uses of UAVs with the highest efficiency. As a result, a fixed propeller is proven to be more efficient than a folding propeller based on the thrust produced.

Keywords: Power System Monitoring, Control and Protection, Switching Circuits & Power Converters
Abstract: Researchers have been working on circuit protection in different ways for decades based on different loads, power, and requirements. Protection measures are particularly crucial for rover circuits due to their high voltage and high current consumption, advanced functionality, system reliability, and component protection needs. While implementing the rescue rover's circuit, we faced problems such as over-current, overheating, low current, short circuits, battery over-discharge, etc. When the Li-ion or Li-po battery in the rover circuit is connected in the wrong polarity, it can cause circuit damage due to its high discharge rating. This paper proposes a comprehensive circuit protection method (CPM) for the entire circuit protection of a rover, which is also applicable to other DC circuits. The proposed comprehensive circuit protection method offers reliable protection against various potential issues, including short circuits, overheating, over-voltage, over-current, low current, low voltage, battery over-discharge, reverse polarity, reverse current, and residual charge issues.

Keywords: Renewable Energy Sources and Technology
Abstract: Thin film solar cells (TFSCs) are one of the leading candidates to reduce the cost of photovoltaic production. However, low absorption coefficient due to relatively low light absorber layer thickness can limit the performance of TFSCs. To address these problems, this study computationally investigates the opto-electronic performance of Cadmium Telluride (CdTe) TFSCs with a metallic nano-grating structure embedded within the absorber layer. The finite-difference time-domain (FDTD) numerical analysis technique was used to computationally analyze different solar cell performance parameters with and without the nano-grating structure. Furthermore, an artificial intelligence (AI) technique, namely particle swarm optimization algorithm (PSO) was used to determine the optimum nano-grating configuration with respect to nano-grating height, angle and duty cycle. The results suggest that the short circuit current density increased by 20.72% while the solar cell efficiency yielded an increase of 21.66% for the Cadmium Telluride (CdTe) thin film solar cells (TFSCs) with the optimized nano-grating structure in comparison to a bare CdTe TFSC with no nano-grating. The results indicate that an important role can be played by such nano-grating structures to significantly enhance the opto-electronic performance of TFSCs and this process can be optimized by the use of AI techniques.

Keywords: Switching Circuits & Power Converters
Abstract: In this paper, an overview of interleaved DCM buck-boost PFC LED driver is presented. The proposed LED driver is designed to operate in discontinuous inductor current conduction mode (DCM) to make use of its obvious benefits such as inherent high power factor and low THDi, simple control (single loop control), excellence efficiency due to inherent zero current switching at turn-on time of the switches and inherent zero diode reverse recovery loss. The adoption of an interleaving technique can reduce the size of passive power components, which can enhance LED driver’s performance and efficiency. This benefit reduces costs, size and weight of LED drivers. The computer software simulations was used to simulate the performances of the proposed LED driver. Experimental results indicate that the characteristics and operations of the LED driver are similar to the simulation results. In addition, the prototype of interleaved DCM buck-boost PFC LED driver was implemented and tested in laboratory, PF 0.9918 and efficiency 92.67%.

Keywords: Switching Circuits & Power Converters
Abstract: This paper introduces a low-quiescent and reduced-power zero current detector (ZCD) designed specifically for a DC-DC switched-mode converter operating in discontinuous current mode. The ZCD incorporates control switches strategically placed to minimize power consumption and leakage current. The ZCD operation is enabled through a ZCD controller circuit, which efficiently manages the power consumption. The controller circuit generates a voltage to control the ZCD, ensuring that the control switches are triggered only when the inductor current is about to change polarity, typically occurring when the low-side power switch of the converter is active. Once the zero current is detected, the ZCD becomes inactive after a slight delay of 1ns, as defined in this paper. The ZCD controller circuit comprises flip-flops and logic gates, managing the ZCD's activation and deactivation. Control switches are strategically placed at each branch of the ZCD circuit to optimize its efficiency. To monitor the inductor current of the DC-DC converter, the voltage polarity at one end (input side) of the inductor is sensed in relation to the ground. The ZCD is configured with transistors in reverse back gate biasing, further enhancing its performance and efficiency. The simulation results showcase impressive outcomes, with the ZCD exhibiting a quiescent current of 1.13nA and an average current of 323nA at a 0.8V supply voltage.

Keywords: Machine Learning, Neural Networks and Deep Learning
Abstract: Radiation-induced lung injury (RILI) is a serious concern for patients affected by breast cancer. Preventing lung injury is impossible since radiotherapy is very effective for breast cancer when applied in the mammary glands close to lungs. Starting early treatment for preventing lung injury can improve the quality of life of the patients. This paper proposes artificial intelligence-based automation for classification of lung injury in retrospective patients. 1692 Position Emission Tomography images of injured lungs are taken and clustered with K-means clustering, since the data set is unlabeled. The entire dataset groups into two clusters-radiation pneumonitis and radiation fibrosis. Silhouette score is 0.1619 when applied on PET images of the lungs. VGG 16 based transfer learning is applied on the k-means clustering algorithm to improve the classification accuracy. The silhouette score is increased to 0.806836.

Keywords: Machine Learning
Abstract: In recent years, intelligent technologies have significantly contributed to enhancing patient care, reducing healthcare costs, and alleviating workload, particularly in telehealth settings. This research paper introduces a data-driven heart disease recommendation system aimed at evaluating the efficacy and accuracy of algorithms in delivering personalized medical test recommendations for individuals diagnosed with heart disease. By employing a sliding window technique, time series data from patients is processed to extract pertinent features. These features are utilized to train the models, enabling them to predict the patient’s condition for the subsequent day. The system incorporates three classifiers: Random Forest, Logistic Regression, and K-Nearest Neighbors. Experimental results demonstrate that the proposed system achieves a remarkable level of accuracy in providing recommendations. Furthermore, it presents a practical solution to mitigate the burden on individuals with heart disease by reducing the necessity for daily medical tests. The conclusive findings affirm the potential of the proposed system as a valuable tool for analyzing medical data, effectively offering accurate and reliable recommendations to patients with chronic heart diseases, thus improving their healthcare decision-making.

Keywords: Neural Networks and Deep Learning, Machine Learning
Abstract: This paper presents an self-supervised Siamese neural network (SNN) for identification and classification of fatty liver severity. SNN is used for self-supervision tasks for being influenced from model optimization property of supervised and manual annotation property of unsupervised learning. This technique is based on contrastive learning of the joint embedding network which can learn more subtle representations from the medical images for classification task, with just one or few number of labelled images required from each class for training. The efficiency of the proposed model is validated on our dataset of liver ultrasound to classify them into three stages of the fatty liver disease and normal liver. A two-class classifier (normal/grade-I, normal/grade-II and normal/grade-III fatty liver) and four-class classifier (normal, grade-I, grade-II, grade-III fatty liver disease) were trained by minimizing contrastive loss to obtain classification accuracy of 98.91% and 96.84% respectively.

Keywords: Neural Networks and Deep Learning, Machine Learning
Abstract: The COVID-19 pandemic has led to an overwhelming volume of scientific publications as researchers strive to address the crisis. To alleviate information overload, the COVID-19 Open Research Dataset (CORD-19) was released to help in analyzing large amounts of data and facilitate faster response. Most existing tools based on CORD-19 use extractive summarizers, which suffer from poor coherence and readability. Thus more abstractive summarizers for COVID-19 are needed. Specifically, using state-of-the-art (SOTA) transformers has shown to be successful in summarizing biomedical datasets like arXiv and PubMed.

In this study, we finetune two checkpoints of SOTA transformer PEGASUS-X on the CORD-19 dataset: PEGASUS-X-BASE-CORD19 and PEGASUS-X-BASE-arXiv-CORD19. Our results highlight the importance of finetuning summarizers on domain-specific datasets in the abstractive summarization of COVID-19 research: checkpoints finetuned on CORD-19 outperform other existing checkpoints and transformers finetuned on more general research datasets (e.g., arXiv and PubMed). After stopword removal in evaluation, we observe that PEGASUS-X-BASE-arXiv-CORD19 surpasses PEGASUS-X-BASE-CORD19 by a small margin. Our checkpoints still fall behind earlier sequence-to-sequence models; however, this limitation may be due to our constrained GPU resources. Future works, with access to more resources, can further improve our checkpoints for COVID-19 research summarization.

Keywords: Beyond CMOS device technology, Advanced CMOS devices and process, Device modeling & characterization
Abstract: In this work, the effect of process-induced inclined sidewalls on small-signal parameters of the nanowire (NW) gate-all-around (GAA) MOSFETs are explored using calibrated TCAD simulation results. The non-quasi-static (NQS) small-signal instead of the quasi-static (QS), parameters are extracted for better accuracy at higher operating frequencies. The distributed channel resistances (Rgd and Rgs) and the intrinsic terminal capacitances (Cgd, and Cgs) are susceptible to the inclination of sidewalls angle (θ). The increase in θ causes the distributed channel resistances to increase and the terminal capacitances to decrease.

Keywords: Beyond CMOS device technology, Device modeling & characterization
Abstract: Reversible computing has a remarkable ability to reduce heat dissipation in computing machinery. It can be broadly applied in various fields, which include Digital signal processing, Cryptography, DNA computing, Network congestion, Database transactions, Quantum computing, etc. Fault detection is a complicated and demanding problem in reversible circuits. Fault detection is an essential process in the field of testing to ensure the reliability and integrity of the circuit. This paper proposes a straightforward approach for Single Missing Gate Fault (SMGF), Multiple Missing Gate Fault (MMGF), Repeated Gate Fault (RGF) and Partial Missing Gate Fault (PMGF) under the Missing Gate Fault (MGF) model. The method includes the process of binary to gray code conversion in order to determine the total number of test vectors to detect the respective MGFs. Experimental results are performed on reversible benchmark circuits to evaluate the number of test vectors required to recognize all the MGFs. The comparative analysis of the proposed work with the existing work is also presented.

Keywords: Device modeling & characterization, Electronic devices, materials and fabrication process, Materials and Structures
Abstract: This study focuses on synthesizing MnO2 thin films(TF) and Au-decorated MnO2 TF using the E-beam evaporation technique to examine its wettability application. The X-ray diffraction (XRD) analysis demonstrates the preferred orientation and polycrystalline growth of the MnO2 film. A contact angle goniometer was used to measure its contact angle, MnO2 TF and MnO2 TF/Au NP showed a contact angle of 112.2 and 104.2, respectively. Optical analysis revealed a significant UV and visible region enhancement for Au-decorated MnO2 TF. Photoinduced hydrophilicity showed that MnO2 TF/Au NP surpassed the bare MnO2 TF. Introducing noble metal nanoparticles onto the MnO2 TF contributed to this enhancement, which holds promise for various self-cleaning applications.

Keywords: Materials and Structures, Device modeling & characterization, Electronic devices, materials and fabrication process
Abstract: The aim of this study is to evaluate the effect of process parameters on resistance and plastic deformation of ultrasonic aluminum ribbon bond on the molybdenum back contact layer of copper indium gallium selenide (CIGS) thin film photovoltaic (TFPV) solar panel. The aluminium ribbon was ultrasonically bonded on molybdenum with two process parameter settings with constant pressure and energy while varying the amplitude. The resistance measurement of the samples was conducted with two techniques which are the transmission line method (TLM) and micro-ohmmeter to evaluate the conductivity of the interconnection. Moreover, the plastic deformation of the aluminium bond from longitudinal and transverse cross-sections was examined by measuring the thickness of the aluminum bond. The resistance of the samples is directly proportional to the amplitude applied while the thickness of the aluminum for both longitudinal and transverse cross-sections is inversely proportional to the amplitude employed. By applying adequate pressure (3.5 bar) and energy (20 J), with the lower amplitude applied which is 7.7 µm, less plastic deformation occurs to the aluminum bond with lower resistance measured.

Keywords: Materials and Structures, Electronic devices, materials and fabrication process, Device modeling & characterization
Abstract: The present work extends the 2D DiamondMesh to 3D homogeneous and heterogeneous DiamondMesh architectures. By incorporating diagonal links into the conventional mesh topology, the 2D DiamondMesh improves network performance while retaining the regular, simple, and scalable properties of the Mesh topology. To further improve the performance of the network, 2D DiamondMesh has been extended to 3D DiamondMesh by stacking the 2D layers vertically and interconnecting them with silicon vias (TSVs). In this work, 3D DiamondMesh has been evaluated for different network sizes. Also, five heterogeneous 3D DiamondMesh architectures have been proposed and evaluated. The results have inferred that the heterogeneity in topology across the layers has shown a remarkable reduction in latency with a slight area overhead or a reduction in the area with a slight penalty in performance. An average reduction of 7.83%, 9.59%, 13.18%, 4.80%, 8.85%, 10.50%, 6.56% and 10.63% in APL can be observed with 4Layer-64node XDMesh, DiamondMesh, DMesh, DiamondMesh+Mesh, DiamondMesh+XDMesh, DiamondMesh+DMesh, DMesh+Mesh, DMesh+XDMesh architectures respectively over the conventional 4Layer-64node Mesh architecture. Similarly, an average reduction of 9.26%, 21.21%, 25.00%, 10.47%, 15.21%, 23.00%, 12.50%, and 16.90% can be observed with 4Layer-256node XDMesh, DiamondMesh, DMesh, DiamondMesh+Mesh, DiamondMesh+XDMesh, DiamondMesh+DMesh, DMesh+Mesh, DMesh+XDMesh architectures respectively over the conventional.

Keywords: Materials and Structures, MEMS and semiconductor sensors
Abstract: We proposed an extremely sensitive E. Coli sensor based on a hyperbolic metamaterial structure combining ultrathin Ag-Al2O3 layers to minimize metallic optical loss. The principle relied on detecting the change in the resonance wavelength due to the interaction of bacteria with the surrounding aqueous environment by utilizing the finite-difference time-domain numerical technique. Our proposed hyperbolic metamaterial E. Coli sensor operated in the range from visible to near-infrared wavelengths exhibiting strong bulk plasmon polaritons at the hyperbolic regime (λ ≥ 460 nm). An anisotropic hyperbolic range was obtained theoretically by solving the effective medium theory. An outstanding sensitivity of 9000 nm per bacteria was achieved for a bulk plasmon-polariton mode. The hyperbolic metamaterial was the origin of obtaining such extremely high sensitivity; no bulk plasmon polaritons were found without hyperbolic metamaterial. We analyzed the effect of different shapes in two-dimensional Ag differential grating on sensing performance. Additionally, we compared the performance parameters of our proposed E. Coli sensor with recently demonstrated sensors. Our proposed hyperbolic metamaterial structure has the potential as a highly sensitive E. Coli sensor operating in a wide range of wavelengths for label-free detection.

Keywords: Materials and Structures
Abstract: This work aims to concisely overview the relationship between metal nanoparticles and MnO2 thin films. The electron beam evaporation technique deposited MnO2 Thin Film (TF) decorated with titanium (Ti), Gold (Au), and silver (Ag) nanoparticles (NP). The study involved the analysis of the structural and optical properties of a MnO2 thin film decorated with nanoparticles (Au, Ag, Ti). XRD reveals that the MnO2 TF and the metal nanoparticles are crystalline. The AFM analysis of Ti NP/MnO2 TF, Ag NP/MnO2 TF, and Au NP/MnO2 TF with a value of 0.19 nm, 0.18 nm, and 0.17 nm, respectively Reflectance measurement unveils the Au NP/MnO2 TF, Ag NP/ MnO2 TF, and Ti NP/MnO2 TF bandgap, which are 3.64 eV, 3.37 eV, and 3.12 eV at 340 nm, 367 nm, and 397 nm. PL emission shows that Au NP/MnO2 TF has more emission than Ag NP/MnO2 TF and Ti NP/MnO2 TF.

Keywords: Data, Text, Web Mining, & Visualization, Data Modeling & Semantic Engineering, Big Data Analytics
Abstract: Dengue, a mosquito-borne viral infection, poses a significant threat, especially in warm, tropical climate countries like Bangladesh, India, Thailand, Malaysia, Laos, etc. This study is solely focused on the dengue data of Bangladesh as it explores the historical dengue data spanning 23 years (2000 to 2022) for EDA purposes, with a focus on 9 years (2014-2022) divisional data for model performance analysis. Additionally, climate data was collected for the same period to examine the potential correlation between dengue cases and climate factors. Machine learning (ML) and Deep learning (DL) models, including Random Forest Regression (RFR), Long Short-Term Memory (LSTM), and LSTM with Artificial Neural Networks (ANN), were implemented and validated against ground truth data. The results reveal notable differences in performance between ML and DL models when handling imbalanced datasets with outliers, with RFR outperforming LSTM when compared to the ground truth data. The study uncovers significant correlations between dengue cases and climate factors like humidity, temperature, and precipitation. The insights gained from this research have practical implications for dengue prevention and control efforts in Bangladesh and beyond, paving the way for more effective strategies and interventions.

Keywords: Data, Text, Web Mining, & Visualization, Data Modeling & Semantic Engineering, Data Centric Programming
Abstract: The purpose of this study is to develop a data analytics and visualization system for detecting falls in elderly and disabled individuals. In this study, analytic techniques and intuitive data visualization methods are employed to overcome challenges associated with real-time fall detection. The system is designed to provide immediate alert notifications, enabling prompt assistance. A user-centered approach takes into consideration the unique needs and capabilities of the target audience. Furthermore, the research emphasizes the importance of data visualization in presenting fall-related information in a clear and understandable way, thereby facilitating better decision-making and patient recommendations. By addressing this gap, the study aims to enhance the overall usability of user interface design for fall detection and response systems for elderly and disabled individuals.

Keywords: Data, Text, Web Mining, & Visualization
Abstract: In recent years, the growing use of biometric recognition systems in various applications has increased the need to protect the biometric templates recorded in multiple databases. Due to their consistency and uniqueness, iris recognition systems have significantly outperformed other biometrics. Directly stored Iris templates on a central server constitute a privacy and security risk. To address this, we will generate a cancelable template that can be stored instead of the original. In the event of a security breach, we will discard the stored template and generate a new iris template. This research employs the Weber Local Descriptor (WLD) technique to create a multi-instance iris biometric system. Left and right iris images are initially acquired and normalized using the USIT toolkit. We generate a feature vector from the normalized image using WLD. The obtained feature vector is then normalized using L1 normalization. The vector of normalized features is then projected onto a median filter to generate a cancelable template. Experiments are conducted on the IIT Delhi iris database, and the results are optimistic compared to previously published research.

Keywords: Domain Specific Data Management
Abstract: Database plays an important role in software architecture and greatly influences system performance. In this paper, we evaluate how two databases, Cassandra and Aerospike, adapt the data sponsor concurrency problem in the telecom sector. To obtain the evaluation result, we provide a method by building the test cases in different scenarios.

Keywords: Crowd Sourcing & Social Intelligence, Human Computer Interface, Neural Networks and Deep Learning
Abstract: The COVID-19 pandemic has greatly affected the country, particularly in executing health and safety protocols. In response, the researchers designed and proposed a project which aims to implement a real-time masked face recognition for logging system with health and temperature monitoring. This includes image processing, machine learning model training, masked and unmasked face detection and recognition, facemask classification, body temperature checking, health declaration form verification for health status validation, health data integration, and log entry recording. The project implementation involved MSI-Z97M System Unit, Logitech HD Pro Webcam C920, FLIR Lepton 3.5 with PureThermalV2 smart I/O board, Dual-display Monitor, FaceNet with Single Shot MultiBox Detector training model, and Microsoft Office 365 applications. The results showed that the researchers successfully implemented the system and achieved the objectives of the study. This research offers a promising solution for real-time tracking and management of health protocols, amidst the ongoing concern brought by the pandemic.

Keywords: Machine Learning, Data Mining, Human Computer Interface
Abstract: Currently, lung cancer is a very common form of cancer. This is because many people are chain smokers nowadays, and many are affected due to their work hazards. The pollution level in modern cities is also a major cause of this type of cancer. This model is built to predict the chances of the occurrence of lung cancer in an individual with the help of certain conditions. The acquired dataset used in this study contains multiple features, but not all are necessary for predicting the risk of lung cancer. Hence, an embedding feature importance model Light Gradient Boosting Machine (LGBM) is used to find the impact of every feature, and the model had trained using the features with maximum influence. The dataset has been divided into two parts for training and testing the model. The models achieve a k-fold mean accuracy of 97.63% and above with all the features and more than 93% on the reduced features for all the deployed models. The models are developed based on a resource-constrained device perspective over the reduced features low resource dataset and use an algorithm to measure the execution time taken for every model to run and complete its prediction after fitting with the respective classifiers. The model developed on medical data should have maximum accuracy and is necessary for time efficiency that reflects on all the deployed models with stability and efficacy, indicating the robustness and non-overfitted model.

Keywords: Machine Learning, Data Mining, Neural Networks and Deep Learning
Abstract: Water is an essential natural resource for life on Earth, and it is the foundation of all living things. However, water pollution is a growing environmental concern caused by human activities, such as improper waste disposal and the discharge of untreated sewage. The consequences of this problem on human health and aquatic life highlight the need for effective supervision and administration of water reserves. This research paper aims to utilize a machine learning approach to predict water quality and identify the most influential features affecting water potability. These features were obtained from three methods, namely Univariate Selection, Recursive Feature Elimination, and Feature Importance, to identify the most influential features. The study compares the performance of various classification algorithms, including K-Nearest Neighbor, Decision Tree, Random Forest, AdaBoost, XGBoost, Linear Discriminant Analysis, Gaussian Naïve Bayes, Logistic Regression, MLPClassifier, and ExtraTree Classifier, using evaluation criteria such as accuracy, precision, recall, F1 score, and computational efficiency. After conducting all these processes, ExtraTree Classifier achieved the highest accuracy of 89 % among the compared machine learning models. Overall, the results of this research may contribute to better public health outcomes and improved management of water resources.

Keywords: Machine Learning, Neural Networks and Deep Learning
Abstract: Weightlifting is a demanding sport requiring power, flexibility, and the correct technique. The snatch and the clean and jerk involve the fast lifting of weight to an overhead position. Incorrect technique or posture may lead to inefficient lifts or even injury. This paper presents a new framework for biomechanics analysis and barbell trajectory tracking in weightlifting by leveraging the capabilities of Keypoint-RCNN and YOLOv7 deep learning models. The proposed framework extracts skeletal information from weightlifting video sequences using a pre-trained Keypoint-RCNN model for human pose estimation and a custom YOLOv7 model to detect and track barbell trajectories. The Keypoint-RCNN model estimates human pose without manual annotation or specialised apparatus, while the YOLOv7 model provides real-time, non-intrusive barbell tracking. The efficacy of barbell trajectory tracking with YOLOv7 on a public weightlifting dataset of 973 images (70-30 train-test ratio) was evaluated, obtaining high precision (0.9214), recall (0.9678), and mAP@0.5 of 0.9792 and mAP@0.5:0.95 of 0.7765, indicating the applicability of this model to weight training applications. The proposed framework presents a cost-effective, user-friendly, and easily accessible alternative to conventional motion capture and analysis systems, making it accessible for lifters of all skill levels and training environments.

Keywords: Machine Learning, Neural Networks and Deep Learning
Abstract: Alzheimer’s disease is a neurodegenerative disease which affects the brain and causes memory loss. Today, this is the most affected disease in the world and has been a real struggle to survive. It is a condition which affects a large number of individuals globally, and it's important to diagnose it early and distinguish between different cognitive states such as AD, MCI, and NC. Personality changes, hallucinations, and difficulties speaking and walking may be symptoms as the illness worsens. It is important to ensure that a patient get treatment on time and management of the disease. The accuracy of AD classification can be greatly improved by incorporating different forms of data, including clinical, genetic, imaging and electroencephalogram(EEG) data, according to a recent study. One area of research involves using biomarkers to identify the disease early, before symptoms are evident. In this paper, we suggest a multimodal approach to identifying cognitive states in Alzheimer's disease. Recent research has demonstrated that multimodal classification techniques for AD diagnosis offer both unique advantages and drawbacks. In conclusion, our study highlights the importance of multimodal data in improving the accuracy of AD classification and provides a promising approach for early diagnosis and management of the disease, ultimately offering hope for better outcomes and quality of life for affected individuals.