International Journal of Advances in Applied Sciences
Not a member yet
668 research outputs found
Sort by
Optical character recognition for Telugu handwritten text using SqueezeNet convolutional neural networks model
Get PDFOptical character recognition (OCR) is a process that recognizes and converts data from scanned images, including both handwritten and printed documents, into an accessible format. The challenges in Telugu OCR arise from compound characters, an extensive character set, limited datasets, character similarities, and difficulties in segmenting overlapping characters. To tackle these segmentation complexities, an algorithm has been developed, prioritizing the preservation of essential features during character segmentation. For distinguishing between structurally similar characters, we used convolutional neural networks (CNN) due to their feature-extracting properties. We have employed the CNN model, the SqueezeNet for feature extraction, resulting in an impressive character recognition rate of 94% and a word recognition rate of 80%
One-step phase transferring method on preparing CuInS2/ZnS QDs dispersion via ultrasonic treatment for bioimaging
Get PDFSonochemistry has many applications in biomedical engineering owing to its ability to break chemical bonds. It can also be used to facilitate the preparation of aqueous quantum dots (QDs) dispersions during microemulsion formati on. Cavitation, bubble growth and collapse during sonication facilitate the formation of microemulsions between a solvent and water. Sonication was employed in this study to prepare aqueous copper indium sulfide colloidal zinc sulfide (CuInS2/ZnS) QD dispersions using different solvent systems. The resultant QDs were characterized by UV-vis absorption, photoluminescence, and dynamic light scattering. Hexane was determined to be the best solvent for the preparation of QD dispersions with the resultant QDs (30 nm in diameter) retaining their chemical integrity and >30% of their quantum yield. CuInS2/ZnS QD dispersions in water can be directly prepared without any non-polar solvent and it has shown high quantum yield. Further, it was also shown that less-toxic, which is efficient in labeling tumor cells derived from human liver cancer (HepG2). It’s evident that, sonication was found to be an efficient alternative approach for preparing aqueous QD dispersions without the need for surface modification, or the use of non-polar solvents during microemulsion formation
Improving the BERT model for long text sequences in question answering domain
Get PDFThe text-based question-answering (QA) system aims to answer natural language questions by querying the external knowledge base. It can be applied to real-world systems like medical documents, research papers, and crime-related documents. Using this system, users don't have to go through the documents manually the system will understand the knowledge base and find the answer based on the text and question given to the system. Earlier state-of-the-art natural language processing (NLP) was recurrent neural network (RNN) and long short-term memory (LSTM). As a result, these models are hard to parallelize and poor at retaining contextual relationships across long text inputs. Today, bidirectional encoder representations from transformers (BERT) are the contemporary algorithm for NLP. BERT is not capable of handling long text sequences; it can handle 512 tokens at a time which makes it difficult for long context. Smooth inverse frequency (SIF) and the BERT model will be incorporated together to solve this challenge. BERT trained on the Stanford question answering dataset (SQuAD) and SIF model demonstrates robustness and effectiveness on long text sequences from different domains. Experimental results suggest that the proposed approach is a promising solution for QA on long text sequences
Employing transfer learning techniques for COVID-19 detection using chest X-ray
Get PDFCoronavirus 2 (SARS-COV-2) is a global emergency that continues to terrify the globe at an alarming rate. Some nations are still combating the virus, attempting to discover infected individuals early on to prevent the infection from spreading. In terms of identifying the pattern in the pictures, radiological patterns have been shown to have greater accuracy, sensitivity, and specificity. Publicly available datasets are used for the implementation. The data is divided into three categories: COVID, normal, and pneumonia patients. Transfer learning is a type of deep learning that allows pre-trained models to be used and achieves high accuracy by detecting various anomalies in limited medical datasets. An image dataset of 1109 pictures was used in this work, and training was done using two distinct models, ResNet50 and InceptionV3, to distinguish the patient categories. For ResNet and InceptionV3, the proposed model has an accuracy of 97.29 and 98.20, respectively, with a sensitivity of 100% for InceptionV3 and a specificity of 99.41% for ResNet50. With a 98.20% accuracy, complete sensitivity, and high specificity, this study presents a deep learning model that gives diagnostics for multiclass classification and attempts to discriminate COVID-19 patients using chest X-ray photos. Other illnesses can also be detected using the proposed model
Strategic electromagnetic interferences suppression in boost converters: zero-switch techniques
Get PDFThis article delves into the growing demand for efficient power conversion technologies accompanying the rise in electric vehicle (EV) adoption. Boost converters, essential for increasing the battery pack voltage to propel EV motors, pose a challenge due to the electromagnetic interference (EMI) generated by the high switching frequency of power devices. To address this issue, practitioners employ zero-voltage switching (ZVS) and zero-current switching (ZCS) techniques. In this comparative study, we systematically evaluate the effectiveness of these soft switching techniques in reducing conducted EMI in boost converters designed for EV applications. The results illuminate the potential of both ZVS and ZCS in significantly mitigating EMI emissions when compared to conventional hard-switching methods. Notably, ZVS soft switching emerges as more efficient and effective, particularly under higher loads, while ZCS soft switching excels in reducing EMI at lighter loads. In conclusion, the study asserts that ZVS soft switching presents a more promising solution for curtailing conducted EMI in boost converters for EV applications, particularly in high-load scenarios. However, it underscores the importance of considering specific operational conditions when deciding between the two techniques
Study of the modern reclamation and ecological state of soils in Azerbaijan
Get PDFFor many years in the Republic of Azerbaijan, numerous reclamation and ecological measures aimed at attracting agricultural lands of natural cenoses to use in agrocenoses were carried out. The soils of the studied territory are quite unique in terms of their soil-forming structure and ecological conditions, and they differ from other regions of the country in terms of irrigation water content. The article reflects some important results of soil and water studies conducted over the last years. Determining the qualitative composition of groundwater and irrigation water in different seasons of the year using modern methods is relevant from an environmental and reclamation point of view. Based on remote sensing and geographic information system (GIS) technologies, our research is based on the analysis of natural and anthropogenic landscape changes and changes in natural conditions as a result of agricultural development in the 20 km buffer zone along the banks of Kura and Araz rivers. As a result of the conducted research, it was determined that the rapid growth of the population has caused the reduction of the area of agro-landscapes, the salinization of the soil, the drying of the water in the water basins, the expansion of the construction areas. This has disrupted the natural balance
Enhanced human activity recognition through deep multi-layer perceptron on the UCI-HAR dataset
Get PDFUsing the UCI-HAR dataset, this paper examines human activity recognition (HAR) from the perspectives of data science and artificial intelligence. The primary objective is to present and evaluate the effectiveness of a multi-layer perceptron (MLP) model, concentrating on six different activity categories. We train and assess the MLP model using the UCI-HAR dataset, contrasting its results with those of convolutional neural networks (CNN). The MLP model shows competitive results, attaining an amazing 97% validation and testing accuracy, highlighting its efficiency for smaller datasets. An extensive study is carried out to assess the model's adaptation to a larger Motion Sense dataset using confusion matrices and cross-entropy, the model shows robustness with an accuracy of 89%. The MLP model performs admirably, demonstrating its capacity to pick up complex patterns. Results from comparative analysis with CNN are competitive, especially when dealing with smaller datasets. The suggested MLP model shows up as a practical and efficient way to advance HAR techniques. Its remarkable performance and versatility not only show its usefulness in real-world scenarios but also point to interesting directions for further study in the area of HAR
Novel cascaded switched-diode five level inverter for renewable energy integration
Get PDFThis research presents the usage of a unique five-level cascaded switching diode for medium voltage integration of renewable energy sources. Its primary purpose is to decrease the quantity of gate drivers and switches. In addition to that, the cost and space for the installation of multilevel inverters are less. The inverter topology of novel cascaded multilevel inverters and switched diodes will combine both benefits. One-cycle control (OCC), which is used for clock phase-shifting (CPS) to regulate a two-stage container security device (CSD) multilevel inverter of renewable energy integration, was created to address issues with the multilevel inverter's direct current (DC) source variations. The topology also provides efficiency, harmonic distortion reduction, and voltage output quality. Waveforms are created, and a robust resistance to DC source variations is attained. The viability of the unique five-level inverter using cascaded switched diodes is confirmed by discussing the results of both simulation and experiment
Advanced controller design for D-FACTS device in grid-connected photovoltaic system controller
Get PDFPhotovoltaic (PV) solar energy is considered one of the highest renewable energy (RE) resources worldwide. Hence, PV system installation capacity is increasing, triggering new power quality problems in grid systems. Some examples of these problems include unbalanced voltages, fluctuating power levels, harmonic distortions, or reverse power flow. To mitigate the adverse impacts of PV integration on voltage regulation and harmonic distortion in electrical distribution systems, a distribution static synchronous compensation (D-STATCOM) is considered a solution. A simulation study is performed by modeling a power system model with an integrated PV system and D-STATCOM. Two control schemes, proportional-integral (PI) and artificial neural network (ANN), were applied within the internal control of D-STATCOM to enhance the power quality of the power system. Two different inverter configurations were adapted, a sinusoidal pulse width modulation (SPWM) and a hysteresis current controller (HCC). Results are obtained as voltage profiles for all the considered control schemes with different inverter types under different contingency conditions. The performance is also evaluated by control characteristics evaluation for different controllers. The controller ANN has better performance than the PI controller and it can mitigate power quality problems and the impact of the PV integration on voltage regulation and harmonic distortion
Determining the possibility of passage through narrow alleys using ultrasonic sensors
Get PDFThe complex road topography of South Korea presents significant challenges to the timely arrival of emergency vehicles. Compounding the issue, obstacles such as legally or illicitly constructed structures, and improperly parked or stationary vehicles, frequently obstruct the path of emergency vehicles. To address these challenges, this study introduces a novel system aimed at enhancing emergency response times. The system employs ultrasonic sensors that can be integrated into personal devices to measure the width of the numerous narrow alleys prevalent in Korea's densely populated regions. Experiments demonstrate that within a 1-meter range in front of a narrow alley with widths varying between 270 cm and 450 cm where vehicle maneuvering is possible, it's possible to accurately gauge the width using two ultrasonic sensors, achieving a precision within a 5 cm margin of error. This level of accuracy enables the practical assessment of whether emergency vehicles can access the area in real-time by identifying the alley's narrowest point. The proposed system is a cost-effective method using easy-to-buy devices for augmenting emergency preparedness and enhancing emergency response times by ensuring that emergency vehicles can navigate through alleys, thereby fostering a safer living environment