International Journal of Innovations in Science & Technology
Not a member yet
813 research outputs found
Sort by
Towards End-to-End Speech Recognition System for Pashto Language Using Transformer Model
Get PDFThe conventional use of Hidden Markov Models (HMMs), and Gaussian Mixture Models (GMMs) for speech recognition posed setup challenges and inefficiency. This paper adopts the Transformer model for Pashto continuous speech recognition, offering an End-to-End (E2E) system that directly represents acoustic signals in the label sequence, simplifying implementation. This study introduces a Transformer model leveraging its state-of-the-art capabilities, including parallelization and self-attention mechanisms. With limited data for Pashto, the Transformer is chosen for its proficiency in handling constraints. The objective is to develop an accurate Pashto speech recognition system. Through 200 hours of conversational data, the study achieves a Word Error Rate (WER) of up to 51% and a Character Error Rate (CER) of up to 29%. The model\u27s parameters are fine-tuned, and the dataset size increased, leading to significant improvements. Results demonstrate the Transformer\u27s effectiveness, showcasing its prowess in limited data scenarios. The study attains notable WER and CER metrics, affirming the model\u27s ability to recognize Pashto speech accurately. In conclusion, the study establishes the Transformer as a robust choice for Pashto speech recognition, emphasizing its adaptability to limited data conditions. It fills a gap in ASR research for the Pashto language, contributing to the advancement of speech recognition technology in under-resourced languages. The study highlights the potential for further improvement with increased training data. The findings underscore the importance of fine-tuning and dataset augmentation in enhancing model performance and reducing error rates
Cluster Analysis of COVID-19 Through Genome Sequences Using Python Bioinformatics Library
Get PDFIntroduction and Importance of Study:
During the COVID-19 pandemic, mortality rates varied across different regions of the world. To better understand the virus\u27s behavior, it\u27s important to gain in-depth knowledge of the nucleotide records of the COVID-19 genomic sequence.
Novelty Statement:
In the study, researchers analyzed clusters of highly affected countries to find similar codons in countries with a similar effect of the virus through Python based library.
Material And Method:
Nucleotide records were extracted from the NCBI database in FASTA format. Further Python Bioinformatics library was used to form the clusters of each country using the K-means clustering technique.
Result and Discussion:
The study focuses on finding the similarities between the codons of amino acids in different countries that are affected in a similar way during COVID-19. For instance, China and the EU have a lower mortality rate and have Leucine, Methionine, Isoleucine, and Valine amino acids in common. On the other hand, countries like Pakistan and India have Leucine, Isoleucine, Valine, and Threonine in common and an average death rate. Moreover, Brazil and the US have a higher mortality rate and share similar codons such as Leucine, Glutamine, and Amber.
Concluding Remarks:
The study shows that countries affected by COVID-19 in a similar way share some common amino acids and their respective codons
Optimizing UAV Wing Performance: A Computational Analysis with Computer-Based Algorithms for Composite Material Integration
Get PDFIntroduction/Importance of Study: The aircraft wing, a vital component, demands intricate design to balance lift generation, drag reduction, and weight minimization. In advanced UAVs (Unmanned Aerial Vehicles), prioritizing stealth and low weight, a pioneering solution involves replacing traditional metallic wing components with composite materials, offering superior lightweight properties, strength, durability, and flexibility.
Novelty Statement: Since most of the studies focus on fuselage, wing ribs, and skin, this research emphasizes spars which are a primary component of the wing.
Material and Method: Composite material T800S/3900-2 is a widely used carbon fiber material in the aerospace industry, which is proposed to be utilized in wing spars. The finite element method is used to carry out the investigation and verification of this transition of materials from metals to composite materials.
Result and Discussion: By varying ply orientations and thicknesses of composite materials to match the stiffness and strength of metal spars, our findings demonstrate that composite wing spars exhibit equivalent stiffness, greater strength, and reduced weight compared to traditional metallic counterparts.
Concluding Remarks: The shift to composite materials in UAV wing design offers a transformative solution. This research shows that for optimal structural performance and achieving lower weight objective composite materials, composite materials are the most suitable materials for UAV wing spars
Fine-Tuning Audio Compression: Algorithmic Implementation and Performance Metrics
Get PDFIntroduction/Importance of Study:
This study introduces a comprehensive evaluation of audio compression algorithms to address the increasing demand for efficient data compression techniques in various audio processing applications.
Novelty statement:
Our research contributes novel insights into the comparative analysis of audio compression algorithms, offering a systematic approach to assess performance across multiple dimensions.
Material and Method:
The research methodology involved the selection of a diverse dataset comprising five audio files, rigorous implementation of four prominent compression algorithms, and systematic evaluation of performance metrics.
Results and Discussion:
The abstract primarily focuses on presenting the findings of the comparative analysis, highlighting the performance of MP3, LPC, Wavelet, and Sub band algorithms across various evaluation parameters.
Concluding Remarks:
In conclusion, our study identifies Wavelet compression as the optimal choice among the evaluated algorithms, offering exceptional accuracy, perceptual quality, and minimal distortion in audio compression
Comparative Analysis on the Effect of Crack Location and Orientation on Crack Growth in Boiler Tube
No full textIntroduction/Importance of Study:
Safety is the paramount concern in the operations and inspections of pressure vessels, such as water tube boilers. Defects in the boiler tubes can lead to the development of cracks.
Novelty Statement:
The investigation focuses on the effect of crack location and orientation on crack growth under cyclic loading which has been analyzed computationally using Separate Morphing and Re-meshing Technology (SMART) in ANSYS. The effect of location on crack growth is primarily focused which is theoretically investigated as well using Simpson’s Integration of Paris’s Law.
Materials and Method:
The tube in focus is a component of a D-type water-tube industrial boiler, crafted from low-carbon steel (SA 178 A). For the effect of location, semi-elliptical cracks on inner and outer tube surfaces have been studied both theoretically and computationally.
Results and Discussion:
Theoretical investigation revealed that cracks on the inner tube surface exhibit a 30.28% higher accumulative growth rate compared to the outer surface, attributed to hoop stress distribution. For investigating the effect of orientation elliptical embedded cracks at certain orientations have been examined computationally and the critical plane orientation for crack growth is identified as perpendicular to the hoop stress.
Concluding Remarks:
In conclusion, the study underscores that cracks grow faster when located at the inner surface and oriented perpendicular to the hoop stress
Combatting Illegal Logging with AI-powered IoT Devices for Forest Monitoring
Get PDFThis research presents a comprehensive strategy for tackling illegal logging by leveraging Artificial Intelligence (AI) and Internet of Things (IoT) technologies. In high-risk forestry areas, sensors-equipped Internet of Things devices are used to continuously monitor and detect the sound of the surroundings. The AI component uses machine learning methods to identify potential unlawful logging activities by accurately detecting and distinguishing sound patterns associated with chainsaw and logging operations such as tree cutting and also detecting natural disasters like wildfires. When such activities are detected by these smart AI-powered IoT devices installed in the forest, real-time notifications are generated after such activity which allows surrounding enforcement agencies, such as the forest department, to intervene promptly. By providing a targeted and prompt solution to the issue of illicit logging, this strategy supports biodiversity preservation and sustainable forest management
Relevance Classification of Flood-Related Tweets Using XLNET Deep Learning Model
Get PDFFloods, being among nature\u27s most significant and recurring phenomena, profoundly impact the lives and properties of tens of millions of people worldwide. As a result of such events, social media structures like Twitter often emerge as the most essential channels for real-time information sharing. However, the total volume of tweets makes it hard to manually distinguish between those relating to floods and those that are not. This poses a large obstacle for responsible government officials who need to make timely and well-knowledgeable decisions. This study attempts to overcome this challenge by utilizing advanced techniques in natural language processing to effectively sort through the extensive volume of tweets. The outcome we obtained from this process is promising, as the XLNET model achieved an extraordinary F1 rating of 0.96. This high degree of overall performance illustrates the model’s usefulness in classifying flood-related tweets. By leveraging the abilities of the XLNET model, we aim to provide a valuable guide for responsible governance, aiding in making timely and well-informed choices during flood situations. This, in turn, will assist reduce the impact of floods on the lives and property-affected communities around the world
Osteochondroma Identification Through Transfer Learning and Convolutional Neural Networks
Get PDFAccurate and timely diagnosis of musculoskeletal conditions like osteochondroma is pivotal in ensuring effective treatment and improved patient outcomes. However, traditional diagnostic methods relying on manual interpretation of medical images can be susceptible to human errors, potentially leading to misdiagnosis or delayed detection. Previous studies have explored Deep Learning (DL) techniques for automated disease detection, but they often face challenges such as limited dataset availability and generalization capabilities across diverse imaging modalities. This research addresses these gaps by proposing a robust Convolutional Neural Network (CNN) framework for osteochondroma identification, leveraging transfer learning and data augmentation techniques. The ResNet-50 architecture, pre-trained on a large dataset, is fine-tuned with dense layers and an output layer for binary classification. Extensive data pre-processing and offline augmentation strategies enhance model performance and generalizability. The proposed model achieves an impressive 97.67% accuracy on the test dataset, demonstrating its effectiveness in distinguishing between normal and osteochondroma cases. Furthermore, its generalizability is validated by training and testing on the publicly available Potato Leaf Disease dataset, showcasing consistent performance in multi-class classification scenarios. While the model exhibits promising results, future work could explore integrating more extensive and diverse datasets and investigating advanced architectures for improved accuracy and computational efficiency. The implications of this research extend to empowering medical practitioners with accurate and swift osteochondroma diagnostics, ultimately contributing to enhanced patient care in orthopaedics
Prediction of Elective Patients and Length of Stay in Hospital
Get PDFThe efficient management of hospital resources and the optimization of patient care are critical tasks in healthcare systems worldwide. One of the key challenges in hospital management is predicting the duration of a patient\u27s stay and accurately determining their location within the hospital, such as whether they are in the Intensive Care Unit (ICU) or the Operating Theater (OT). In this study, we address this problem statement by employing machine learning algorithms to predict both the stay duration and the location of patients within the hospital. The methods applied in this study include Random Forest, Support Vector Machine (SVM), and K-nearest neighbors (KNN) algorithms. These algorithms utilize patient demographic information such as age, weight, and severity of disease as features to predict the stay duration and location. The dataset used for this study consists of a revised dataset containing relevant patient information. Upon applying the machine learning algorithms, we obtained promising results. The Random Forest algorithm achieved the highest accuracy of 88.6% in predicting patient locations, followed by SVM with an accuracy of 60.8% and KNN with an accuracy of 58.1%. Additionally, Random Forest exhibited superior precision, recall, and F1-scores for both ICU and OT classifications compared to SVM and KNN. The results obtained from this study have several practical implications and potential uses. Firstly, accurate predictions of patient stay duration and location can aid hospital administrators in resource allocation and planning, enabling them to efficiently manage bed occupancy and staffing levels. Additionally, healthcare providers can use these predictions to anticipate patient needs and allocate resources accordingly, thereby enhancing patient care and satisfaction. Moreover, the machine learning algorithms utilized in this study can be integrated into hospital information systems to automate the prediction process, providing real-time insights to healthcare professionals. In conclusion, the application of machine learning algorithms in predicting patient stay duration and location within the hospital offers promising results and valuable insights for hospital management. By leveraging patient demographic information and advanced predictive models, healthcare institutions can improve operational efficiency, enhance patient care delivery, and ultimately optimize resource utilization
Epidemiological Insights and Statistical Analysis of a Recent Conjunctivitis Outbreak in Lahore, Pakistan
Get PDFThis study presents a comprehensive epidemiological analysis of a recent outbreak of conjunctivitis, known as pink eye disease, in Lahore, Pakistan. Conjunctivitis is a highly contagious eye infection that poses a significant public health concern, particularly in social environments. The research focuses on understanding the prevalence and influencing factors of this ailment through a statistical analysis of patient data. The gender distribution among patients revealed a slightly higher prevalence among males (52.5%) as compared to females (47.5%). Young adults (age 18-25) comprised the highest affected group (89%), emphasizing the higher infection\u27s prevalence among this demographic. Symptom analysis highlights moderate to severe manifestations as predominant, significantly impacting patients\u27 daily routines. Males exhibit a higher severity, potentially associated with increased social engagement compared to females. Notably, the infection commonly affects both eyes (86%), and individuals with a history of prior eye infections demonstrate a reduced likelihood of contracting conjunctivitis (11%). The onset of symptoms is typically sudden (85%), with a gradual presentation in some cases (15%). Despite the contagious nature of the infection, its spread to family members’ remains relatively limited (36.8%). Remarkably, although symptoms are severe, the duration of the infection is brief, with most patients recovering within 2-5 days, even without medical consultation. Moreover, the spatial distribution showed that redness and itchiness were very severe in location 1(latitude 31.4972, and longitude 74.2735) and severe in location 4 (latitude 31.508, and longitude 74.327). In conclusion, this study is the first to report on the rapid yet severe nature of a conjunctivitis outbreak in Lahore. Key trends, including gender disparities, previous eye infection history, sudden onset of symptoms, and limited familial transmission, have emerged. Understanding these dynamics is crucial for implementing targeted preventive measures and developing effective management strategies for this contagious eye infection. The findings contribute valuable epidemiological insights that can guide public health interventions in similar scenarios