Pub Date : 2023-06-08DOI: 10.1109/HORA58378.2023.10155772
I. Javed, H. Afzal
Social media platforms have become the go-to medium for connecting people in this era of the internet. Twitter has emerged as a popular platform that allowsusers to share their views on current events and political organizations, providing a wealth of political information. The aim of this study is to utilize natural language processing techniques to analyze a dataset extracted from Twitter. This involves retrieving data from Twitter, performing sentiment analysis using deeplearning approaches, and creating a Python library that classifiesinput texts as either positive or negative. The training data used in this study included the Roman-Urdu language, comprising 89793 entries. Various classification models were employed to categorize emotions, with the ensemble technique ultimately used to determine the results. The LSTM classifier achieved an accuracy of 87%, while the Bert model performed the best with 90% accuracy.
{"title":"Opinion Analysis of Bi-Lingual Event Data from Social Networks","authors":"I. Javed, H. Afzal","doi":"10.1109/HORA58378.2023.10155772","DOIUrl":"https://doi.org/10.1109/HORA58378.2023.10155772","url":null,"abstract":"Social media platforms have become the go-to medium for connecting people in this era of the internet. Twitter has emerged as a popular platform that allowsusers to share their views on current events and political organizations, providing a wealth of political information. The aim of this study is to utilize natural language processing techniques to analyze a dataset extracted from Twitter. This involves retrieving data from Twitter, performing sentiment analysis using deeplearning approaches, and creating a Python library that classifiesinput texts as either positive or negative. The training data used in this study included the Roman-Urdu language, comprising 89793 entries. Various classification models were employed to categorize emotions, with the ensemble technique ultimately used to determine the results. The LSTM classifier achieved an accuracy of 87%, while the Bert model performed the best with 90% accuracy.","PeriodicalId":247679,"journal":{"name":"2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131377053","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-08DOI: 10.1109/HORA58378.2023.10156733
Aws Khudhur, N. Ramaha
Predicting student performance is a crucial area of research in the field of education. To improve the accuracy and reliability of student performance prediction, machine learning (ML) techniques have been widely used. In this study, we propose a novel approach for predicting student performance using five ML techniques, which include data analysis, pre-processing techniques, and data augmentation using GAN. We evaluate the proposed approach using a real-world dataset of student academic records and compare the results to those obtained without data augmentation. Our findings demonstrate that data augmentation significantly improves the accuracy and reliability of student performance prediction. Specifically, the random forest classifier achieves the best accuracy of 99.8%. This research contributes to the field of education by providing a more comprehensive and accurate model for predicting student performance, which can support informed decision-making and improve educational outcomes.
{"title":"Students' Performance Prediction Using Machine Learning Based on Generative Adversarial Network","authors":"Aws Khudhur, N. Ramaha","doi":"10.1109/HORA58378.2023.10156733","DOIUrl":"https://doi.org/10.1109/HORA58378.2023.10156733","url":null,"abstract":"Predicting student performance is a crucial area of research in the field of education. To improve the accuracy and reliability of student performance prediction, machine learning (ML) techniques have been widely used. In this study, we propose a novel approach for predicting student performance using five ML techniques, which include data analysis, pre-processing techniques, and data augmentation using GAN. We evaluate the proposed approach using a real-world dataset of student academic records and compare the results to those obtained without data augmentation. Our findings demonstrate that data augmentation significantly improves the accuracy and reliability of student performance prediction. Specifically, the random forest classifier achieves the best accuracy of 99.8%. This research contributes to the field of education by providing a more comprehensive and accurate model for predicting student performance, which can support informed decision-making and improve educational outcomes.","PeriodicalId":247679,"journal":{"name":"2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115868649","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-08DOI: 10.1109/HORA58378.2023.10156710
Aoxin Ni, Nasser Kehtamavaz
Adaptive Dynamic Range Optimization (ADRO) is an amplification strategy which is used for hearing aids and other assistive hearing devices. To take into consideration hearing preferences of a specific user in the field, ADRO has been personalized by using maximum likelihood inverse reinforcement learning. A smartphone app is developed in this paper implementing the personalization of ADRO in real-world audio environments so that clinical studies can be carried out in the field. The developed app adjusts the comfort target parameter of ADRO by conducting paired audio comparisons in real-time to reach a personalized setting of gain values in five frequency bands. The audio processing steps taken to enable the app real-time functionality are discussed. The ADRO personalization results of the experiments carried out by using the app in different real-world environments are also presented.
{"title":"A Real- Time Smartphone App for Field Personalization of Hearing Enhancement by Adaptive Dynamic Range Optimization","authors":"Aoxin Ni, Nasser Kehtamavaz","doi":"10.1109/HORA58378.2023.10156710","DOIUrl":"https://doi.org/10.1109/HORA58378.2023.10156710","url":null,"abstract":"Adaptive Dynamic Range Optimization (ADRO) is an amplification strategy which is used for hearing aids and other assistive hearing devices. To take into consideration hearing preferences of a specific user in the field, ADRO has been personalized by using maximum likelihood inverse reinforcement learning. A smartphone app is developed in this paper implementing the personalization of ADRO in real-world audio environments so that clinical studies can be carried out in the field. The developed app adjusts the comfort target parameter of ADRO by conducting paired audio comparisons in real-time to reach a personalized setting of gain values in five frequency bands. The audio processing steps taken to enable the app real-time functionality are discussed. The ADRO personalization results of the experiments carried out by using the app in different real-world environments are also presented.","PeriodicalId":247679,"journal":{"name":"2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)","volume":"350 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133463151","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-08DOI: 10.1109/HORA58378.2023.10156803
Turgut Özseven, M. Arpacioglu
The increase in the rate of urbanization in recent years has led to an increase in environmental sound sources and, accordingly, an increase in noise pollution. Street noises, especially in big cities, pose some health problems. In terms of smart cities, accurate detection of street sounds is important in detecting unwanted sounds and responding to emergencies. In this study, research was carried out to select acoustic features of street sounds with meta-heuristic methods. In the experimental study, using the Urbansound8k dataset, feature extraction was done through openSMILE software, then feature selection was performed with PSO and WO algorithms. SVM and k-NN methods were applied for the classification process. Accuracy rates were obtained with SVM and k-NN classifiers as 88.12%, 69.32% in the PSO algorithm, 88.39%, and 70.51% in the WO algorithm, respectively.
{"title":"Classification of Urban Sounds with PSO and WO Based Feature Selection Methods","authors":"Turgut Özseven, M. Arpacioglu","doi":"10.1109/HORA58378.2023.10156803","DOIUrl":"https://doi.org/10.1109/HORA58378.2023.10156803","url":null,"abstract":"The increase in the rate of urbanization in recent years has led to an increase in environmental sound sources and, accordingly, an increase in noise pollution. Street noises, especially in big cities, pose some health problems. In terms of smart cities, accurate detection of street sounds is important in detecting unwanted sounds and responding to emergencies. In this study, research was carried out to select acoustic features of street sounds with meta-heuristic methods. In the experimental study, using the Urbansound8k dataset, feature extraction was done through openSMILE software, then feature selection was performed with PSO and WO algorithms. SVM and k-NN methods were applied for the classification process. Accuracy rates were obtained with SVM and k-NN classifiers as 88.12%, 69.32% in the PSO algorithm, 88.39%, and 70.51% in the WO algorithm, respectively.","PeriodicalId":247679,"journal":{"name":"2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)","volume":"25 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114007682","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-08DOI: 10.1109/HORA58378.2023.10156719
Ceren Dursun, Alper Ozcan
Ahstract-The integration of recommender systems contributes to the tourism industry as it provides tailored recommendations to users, assisting them in discovering and selecting the most suitable accommodation options based on their particular needs and preferences. By providing personalized recommendations that are tailored to each user's preferences and needs, hotel rec-ommendation systems could assist in reducing the time and effort required to find the best hotel options. In addition, users could discover new and relevant accommodation alternatives that they might not have previously considered. Despite the importance of the reasons underlying user preferences, existing review-based recommendation systems often neglect the importance of sentiment words linked to related item aspects. To address this need, this study presents a sentiment-enhanced hotel recommender system using neural collaborative filtering that incorporates information derived from both textual reviews and user-hotel relationships. This study employs a neural collaborative filtering approach to learn the relationship between user-hotel interactions and a sentiment-enhanced recommendation system. In regards to the experiment conducted in this study, our method enhances the model's ability to capture user preferences and item features through information from sentiment-enhanced text reviews in comparison to sub-ratings generated by users. Aspect-based sentiment analysis improves personalized hotel recommendations by taking into account the sentiment toward specific aspects of the hotel, such as cleanliness, service, or location.
{"title":"Sentiment-enhanced Neural Collaborative Filtering Models Using Explicit User Preferences","authors":"Ceren Dursun, Alper Ozcan","doi":"10.1109/HORA58378.2023.10156719","DOIUrl":"https://doi.org/10.1109/HORA58378.2023.10156719","url":null,"abstract":"Ahstract-The integration of recommender systems contributes to the tourism industry as it provides tailored recommendations to users, assisting them in discovering and selecting the most suitable accommodation options based on their particular needs and preferences. By providing personalized recommendations that are tailored to each user's preferences and needs, hotel rec-ommendation systems could assist in reducing the time and effort required to find the best hotel options. In addition, users could discover new and relevant accommodation alternatives that they might not have previously considered. Despite the importance of the reasons underlying user preferences, existing review-based recommendation systems often neglect the importance of sentiment words linked to related item aspects. To address this need, this study presents a sentiment-enhanced hotel recommender system using neural collaborative filtering that incorporates information derived from both textual reviews and user-hotel relationships. This study employs a neural collaborative filtering approach to learn the relationship between user-hotel interactions and a sentiment-enhanced recommendation system. In regards to the experiment conducted in this study, our method enhances the model's ability to capture user preferences and item features through information from sentiment-enhanced text reviews in comparison to sub-ratings generated by users. Aspect-based sentiment analysis improves personalized hotel recommendations by taking into account the sentiment toward specific aspects of the hotel, such as cleanliness, service, or location.","PeriodicalId":247679,"journal":{"name":"2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117205977","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-08DOI: 10.1109/HORA58378.2023.10156750
Erkan Tur
In the plastics industry, particularly in multistage extrusion processes, maintaining a consistent product quality is paramount. The extrusion process often involves converting granular raw material into a plastic film by heating and stretching it across multiple layers. Two significant aspects of the output product quality are product parameters such as film thickness and stretch, and the presence or absence of defects. Currently, product parameters are efficiently monitored using sensors, but defect identification largely relies on the manual visual inspection by the operator, which may not always occur in real time. This manual approach is prone to errors and can result in delayed defect detection. This study proposes to explore the application of deep learning to automate defect detection in the multi-stage plastic extrusion process. By training deep learning models on a rich dataset of process parameters of the output product, it is possible to enable realtime, automatic identification of defects. This can lead to a substantial improvement in the efficiency and accuracy of the quality control process. Various deep learning architectures will be employed and evaluated for their effectiveness in this task. Furthermore, this study also aims to investigate the correlation between various factors, including equipment performance and quality of incoming raw materials, and the occurrence of defects. Advanced deep learning techniques like Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks will be used to analyze the time-series data from the extrusion process. The findings from this analysis could provide valuable insights into the root causes of defects and guide efforts to minimize their occurrence. In conclusion, this research seeks to leverage the potential of deep learning to enhance the quality control process in the multi-stage plastic extrusion industry, with a focus on automated, real-time defect detection and root cause analysis.
{"title":"Applying Deep Learning for Automated Quality Control and Defect Detection in Multi-stage Plastic Extrusion Process","authors":"Erkan Tur","doi":"10.1109/HORA58378.2023.10156750","DOIUrl":"https://doi.org/10.1109/HORA58378.2023.10156750","url":null,"abstract":"In the plastics industry, particularly in multistage extrusion processes, maintaining a consistent product quality is paramount. The extrusion process often involves converting granular raw material into a plastic film by heating and stretching it across multiple layers. Two significant aspects of the output product quality are product parameters such as film thickness and stretch, and the presence or absence of defects. Currently, product parameters are efficiently monitored using sensors, but defect identification largely relies on the manual visual inspection by the operator, which may not always occur in real time. This manual approach is prone to errors and can result in delayed defect detection. This study proposes to explore the application of deep learning to automate defect detection in the multi-stage plastic extrusion process. By training deep learning models on a rich dataset of process parameters of the output product, it is possible to enable realtime, automatic identification of defects. This can lead to a substantial improvement in the efficiency and accuracy of the quality control process. Various deep learning architectures will be employed and evaluated for their effectiveness in this task. Furthermore, this study also aims to investigate the correlation between various factors, including equipment performance and quality of incoming raw materials, and the occurrence of defects. Advanced deep learning techniques like Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks will be used to analyze the time-series data from the extrusion process. The findings from this analysis could provide valuable insights into the root causes of defects and guide efforts to minimize their occurrence. In conclusion, this research seeks to leverage the potential of deep learning to enhance the quality control process in the multi-stage plastic extrusion industry, with a focus on automated, real-time defect detection and root cause analysis.","PeriodicalId":247679,"journal":{"name":"2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116406870","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-08DOI: 10.1109/HORA58378.2023.10156747
S. Mahmood, Swash Sami Mohammed, Ayad Ghany Ismaeel, Hülya Gükalp Clarke, Iman Nozad Mahmood, D. Aziz, Sameer Alani
This research presents a deep convolutional neural network (CNN) as a solution for identifying malarial cells that are infected. The AI model suggested in this work comprises a three-layered CNN and a two-layered dense neural network. The model can capture both minor and significant features by utilizing CNN, thereby extracting a maximum amount of information from the input data. The model is trained over 20 epochs and evaluated using the binary cross entropy loss function and accuracy metric to assess its performance. Remarkably, the proposed model achieved an impressive accuracy of 96% and maintained a loss value below 0.2 for both the training and validation datasets. Ultimately, this research demonstrates promising potential for automating the detection of malaria through parasite cell counting.
{"title":"Improved Malaria Cells Detection Using Deep Convolutional Neural Network","authors":"S. Mahmood, Swash Sami Mohammed, Ayad Ghany Ismaeel, Hülya Gükalp Clarke, Iman Nozad Mahmood, D. Aziz, Sameer Alani","doi":"10.1109/HORA58378.2023.10156747","DOIUrl":"https://doi.org/10.1109/HORA58378.2023.10156747","url":null,"abstract":"This research presents a deep convolutional neural network (CNN) as a solution for identifying malarial cells that are infected. The AI model suggested in this work comprises a three-layered CNN and a two-layered dense neural network. The model can capture both minor and significant features by utilizing CNN, thereby extracting a maximum amount of information from the input data. The model is trained over 20 epochs and evaluated using the binary cross entropy loss function and accuracy metric to assess its performance. Remarkably, the proposed model achieved an impressive accuracy of 96% and maintained a loss value below 0.2 for both the training and validation datasets. Ultimately, this research demonstrates promising potential for automating the detection of malaria through parasite cell counting.","PeriodicalId":247679,"journal":{"name":"2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115382839","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-08DOI: 10.1109/HORA58378.2023.10156805
Yagmur Selenay Selcuk, Elvin Çoban
Providing healthcare services at home is crucial for patients who require long-term care or face mobility or other health-related constraints that prevent them from traveling to healthcare facilities. Effective data analysis techniques are needed to optimize these services to understand patient needs and allocate resources efficiently. Machine learning algorithms can analyze big datasets generated from home healthcare services to identify patterns, trends, and predictive factors. By utilizing these techniques, predictive models for service time can be developed, leading to improved patient outcomes, increased efficiency, and reduced costs. This study explores the significance of various features in predicting service time for home healthcare services by analyzing real-life data using data analysis techniques. By developing a correlation matrix, healthcare providers can examine the relationships between features as well as their connections with the target value, thereby providing valuable managerial insights into improving the quality of home healthcare services through enhanced predictions of service time.
{"title":"Advancing Home Healthcare Through Machine Learning: Predicting Service Time for Enhanced Patient Care","authors":"Yagmur Selenay Selcuk, Elvin Çoban","doi":"10.1109/HORA58378.2023.10156805","DOIUrl":"https://doi.org/10.1109/HORA58378.2023.10156805","url":null,"abstract":"Providing healthcare services at home is crucial for patients who require long-term care or face mobility or other health-related constraints that prevent them from traveling to healthcare facilities. Effective data analysis techniques are needed to optimize these services to understand patient needs and allocate resources efficiently. Machine learning algorithms can analyze big datasets generated from home healthcare services to identify patterns, trends, and predictive factors. By utilizing these techniques, predictive models for service time can be developed, leading to improved patient outcomes, increased efficiency, and reduced costs. This study explores the significance of various features in predicting service time for home healthcare services by analyzing real-life data using data analysis techniques. By developing a correlation matrix, healthcare providers can examine the relationships between features as well as their connections with the target value, thereby providing valuable managerial insights into improving the quality of home healthcare services through enhanced predictions of service time.","PeriodicalId":247679,"journal":{"name":"2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123783234","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-08DOI: 10.1109/HORA58378.2023.10155771
I. Sierova, I. Aksonova, V. Shlykova, Tetiana Milevska
Based on the integration orientation of the development of the national economy, as the direction of its growth and competitiveness, general approaches to its analytical assessment are defined. The analysis of favorable conditions of integration is combined with the correctness of the implementation of analytical generalizations as a basis for the formation of legitimate conclusions. Based on the fact that the determination of real trends reflects the relative characteristics of the dynamics, a comparative analysis was conducted, which confirmed the relative stability of the export potential of the Ukrainian grain crops market. The calculation of the basic indicators of the economic openness by grain crops in comparison with the general level for the country indicated the similarity of trends, but a higher level of stability. The general conclusion regarding the impact of grain exports on the level of the country's competitiveness is confirmed by a comparative analysis of the Global competitiveness index trends and the economic openness of the Ukrainian grain market.
{"title":"Computer-mathematical support for analytical assessment of trends in the Ukrainian grain market development","authors":"I. Sierova, I. Aksonova, V. Shlykova, Tetiana Milevska","doi":"10.1109/HORA58378.2023.10155771","DOIUrl":"https://doi.org/10.1109/HORA58378.2023.10155771","url":null,"abstract":"Based on the integration orientation of the development of the national economy, as the direction of its growth and competitiveness, general approaches to its analytical assessment are defined. The analysis of favorable conditions of integration is combined with the correctness of the implementation of analytical generalizations as a basis for the formation of legitimate conclusions. Based on the fact that the determination of real trends reflects the relative characteristics of the dynamics, a comparative analysis was conducted, which confirmed the relative stability of the export potential of the Ukrainian grain crops market. The calculation of the basic indicators of the economic openness by grain crops in comparison with the general level for the country indicated the similarity of trends, but a higher level of stability. The general conclusion regarding the impact of grain exports on the level of the country's competitiveness is confirmed by a comparative analysis of the Global competitiveness index trends and the economic openness of the Ukrainian grain market.","PeriodicalId":247679,"journal":{"name":"2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123874840","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-08DOI: 10.1109/HORA58378.2023.10155786
Aniebiet Micheal Ezekiel, R. Obermaisser
Recent research on Artificial Neural Networks (ANNs) has shown significant improvement in machine learning over traditional algorithms in many disciplines. This paper contributes to the advances in medical science and AI technologies by exploring this promising technology for real-time cardiovascular complication detection and resuscitation during rescue missions. Previous studies have relied on cloud-based computing or specialized hardware such as graphics processing units (GPUs), which can be expensive and require significant power consumption. Additionally, existing AI models are often not optimized for low-latency processing, hindering their real-time applications. This study proposes a PyTorch-based ANN model with time optimization techniques on the field-programmable gate arrays (FPGAs) hardware platform, providing data privacy and hardware security. Our approach includes intermediate layer saving and layer parameter reuse, reducing computational complexity and memory requirements while maintaining accuracy. The prototype wearable utilizes a Trenz Electronics TE0802 FPGA board with custom PYNQ-Linux software, providing a low-cost, low-power, and high-performance hardware platform. Using the Apache TVM toolchain, our ANN model predicts cardiovascular disease risk and aids rescuers in making rapid and precise clinical decisions. The results demonstrate 95.9% accuracy in detecting cardiovascular complications, with an average execution time of 41.99ms using TVM. Additionally, our time optimization technique achieves reduced inference times of 33%, 55%, and 79% for reusing the saved output files of layers 1, 2, and 3, respectively, as validated through simulations and experiments.
{"title":"Time-Optimized Detection of Cardiovascular Complications with Artificial Intelligence in Rescue Operations using FPGA-based Wearable","authors":"Aniebiet Micheal Ezekiel, R. Obermaisser","doi":"10.1109/HORA58378.2023.10155786","DOIUrl":"https://doi.org/10.1109/HORA58378.2023.10155786","url":null,"abstract":"Recent research on Artificial Neural Networks (ANNs) has shown significant improvement in machine learning over traditional algorithms in many disciplines. This paper contributes to the advances in medical science and AI technologies by exploring this promising technology for real-time cardiovascular complication detection and resuscitation during rescue missions. Previous studies have relied on cloud-based computing or specialized hardware such as graphics processing units (GPUs), which can be expensive and require significant power consumption. Additionally, existing AI models are often not optimized for low-latency processing, hindering their real-time applications. This study proposes a PyTorch-based ANN model with time optimization techniques on the field-programmable gate arrays (FPGAs) hardware platform, providing data privacy and hardware security. Our approach includes intermediate layer saving and layer parameter reuse, reducing computational complexity and memory requirements while maintaining accuracy. The prototype wearable utilizes a Trenz Electronics TE0802 FPGA board with custom PYNQ-Linux software, providing a low-cost, low-power, and high-performance hardware platform. Using the Apache TVM toolchain, our ANN model predicts cardiovascular disease risk and aids rescuers in making rapid and precise clinical decisions. The results demonstrate 95.9% accuracy in detecting cardiovascular complications, with an average execution time of 41.99ms using TVM. Additionally, our time optimization technique achieves reduced inference times of 33%, 55%, and 79% for reusing the saved output files of layers 1, 2, and 3, respectively, as validated through simulations and experiments.","PeriodicalId":247679,"journal":{"name":"2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA)","volume":"273 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121361162","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：