Pub Date : 2022-11-05DOI: 10.1109/iSAI-NLP56921.2022.9960273
Mako Komatsu, Masato Takeuchi, Teruhiko Unoki, M. Shikida
The impact of COVID-19 has led to the shift of job interviews online. There is now a return to face-to-face interviews in important situations, such as the final interview. However, it is still difficult to practice face-to-face interviews, and there is a growing need to practice face-to-face interviews alone or remotely. The problems with practicing interviews alone are that there is no listener in front of the practitioner, so the practitioner does not feel the nervousness about being watched and evaluated. In this paper, we aim to support these issues by using a small communication robot. We conduct experiments under six conditions: practicing alone, with a person face-to-face, with an autonomous robot, with a teleoperated robot, with an avatar remotely, and with a person remotely. Then we examine the influence of the practice style, such as the practitioner's nervousness. The results suggest that the most effective practice is possible when practicing with a person, regardless of whether it is face-to-face or remotely, but that the interview practice support with a small communicative robot is useful in the current social situation.
{"title":"The Evaluation of Interviewer's Presentation Styles for Interview Practice with a Communicative Robot","authors":"Mako Komatsu, Masato Takeuchi, Teruhiko Unoki, M. Shikida","doi":"10.1109/iSAI-NLP56921.2022.9960273","DOIUrl":"https://doi.org/10.1109/iSAI-NLP56921.2022.9960273","url":null,"abstract":"The impact of COVID-19 has led to the shift of job interviews online. There is now a return to face-to-face interviews in important situations, such as the final interview. However, it is still difficult to practice face-to-face interviews, and there is a growing need to practice face-to-face interviews alone or remotely. The problems with practicing interviews alone are that there is no listener in front of the practitioner, so the practitioner does not feel the nervousness about being watched and evaluated. In this paper, we aim to support these issues by using a small communication robot. We conduct experiments under six conditions: practicing alone, with a person face-to-face, with an autonomous robot, with a teleoperated robot, with an avatar remotely, and with a person remotely. Then we examine the influence of the practice style, such as the practitioner's nervousness. The results suggest that the most effective practice is possible when practicing with a person, regardless of whether it is face-to-face or remotely, but that the interview practice support with a small communicative robot is useful in the current social situation.","PeriodicalId":399019,"journal":{"name":"2022 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114479553","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 : 2022-11-05DOI: 10.1109/iSAI-NLP56921.2022.9960259
Thazin Myint Oo, T. Tanprasert, Ye Kyaw Thu, T. Supnithi
Weighted Finite State Transducers (WFST) can be very efficient to implement Burmese dialects translation. We illustrate this on two Burmese dialect language pairs, Burmese-Beik and Burmese-Rakhine. In this study, we examine syllable and word segmentation schemes and their effect on alignment and transducing between dialect language pairs. We performed alignments with Anymalign, fastalign, pialign, Hieralign, eflomal and GIZA ++ approaches and implemented WFST based machine translation system with OpenFst library. From the overall results, syllable segmentation achieved higher BLEU and chrF scores for Burmese-Rakhine and Rakhine-Burmese translations. However, word segmentation achieved better translation performance for Burmese-Beik and Beik-Burmese translation directions. Alignment techniques fast align, Hieralign, eflomal and GIZA ++ are working well for low-resource Burmese dialects.
{"title":"Syllable-to-Syllable and Word-to-Word Transducers for Burmese Dialect Translation","authors":"Thazin Myint Oo, T. Tanprasert, Ye Kyaw Thu, T. Supnithi","doi":"10.1109/iSAI-NLP56921.2022.9960259","DOIUrl":"https://doi.org/10.1109/iSAI-NLP56921.2022.9960259","url":null,"abstract":"Weighted Finite State Transducers (WFST) can be very efficient to implement Burmese dialects translation. We illustrate this on two Burmese dialect language pairs, Burmese-Beik and Burmese-Rakhine. In this study, we examine syllable and word segmentation schemes and their effect on alignment and transducing between dialect language pairs. We performed alignments with Anymalign, fastalign, pialign, Hieralign, eflomal and GIZA ++ approaches and implemented WFST based machine translation system with OpenFst library. From the overall results, syllable segmentation achieved higher BLEU and chrF scores for Burmese-Rakhine and Rakhine-Burmese translations. However, word segmentation achieved better translation performance for Burmese-Beik and Beik-Burmese translation directions. Alignment techniques fast align, Hieralign, eflomal and GIZA ++ are working well for low-resource Burmese dialects.","PeriodicalId":399019,"journal":{"name":"2022 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"116 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124325647","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 : 2022-11-05DOI: 10.1109/iSAI-NLP56921.2022.9960264
Anurak Yutthanawa, Janya Onpans
In Thailand and numerous other Southeast Asian countries, Rice is one of the most income country products. Rice leaf disease control must be improved in order to enhance rice production. But it is a complicated process dependent on the farmer's experience and local knowledge. Artificial intelligence solutions will become one of the options for resolving this problem and informing all new and existing farmers about the diseases of their products. Big Transfer (BiT) is a deep learning model proposed in this paper for identifying rice leaf disease. BiT-M prediction performance is notable, with 100% prediction accuracy after 19 epochs of training.
{"title":"Rice Leaf Diseases Identify Using Big Transfer","authors":"Anurak Yutthanawa, Janya Onpans","doi":"10.1109/iSAI-NLP56921.2022.9960264","DOIUrl":"https://doi.org/10.1109/iSAI-NLP56921.2022.9960264","url":null,"abstract":"In Thailand and numerous other Southeast Asian countries, Rice is one of the most income country products. Rice leaf disease control must be improved in order to enhance rice production. But it is a complicated process dependent on the farmer's experience and local knowledge. Artificial intelligence solutions will become one of the options for resolving this problem and informing all new and existing farmers about the diseases of their products. Big Transfer (BiT) is a deep learning model proposed in this paper for identifying rice leaf disease. BiT-M prediction performance is notable, with 100% prediction accuracy after 19 epochs of training.","PeriodicalId":399019,"journal":{"name":"2022 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127828317","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}
Pumpkins contain a significant amount of beta-carotene. Beta-Carotene has numerous biological functions in the human body and because human is not able to synthesize any of them, it is necessary to supply these valuable compounds with food or pharmaceuticals. Internet of thing (IoT) in agriculture is not only reduce the man efforts but also improve the productivity and the efficiency. This research is primarily about the study of effect of beta-carotene in pumpkin between polyhouse and outdoor, because polyhouse is a closed structure protect the plants from weather conditions, insect and pest attacks. The irrigation in polyhouse using automatic drip irrigation, which operate according to the soil moisture threshold. Air temperature control using Fan based on temperature threshold. Analysis of beta carotene contain in pumpkin samples with polyhouse and outdoor process were determined by the samples were collected for three time periods found that the linear regression equation of the curve was y = 0.2111x-0.09161, with a coefficient r2 = 0.9975. The result show that, plants growing in the green house most are higher than the outdoor.
{"title":"The Effect of Beta-Carotene contain in The Pumpkin using IoT Technology in Polyhouse","authors":"Kanitha Homjun, Kasree Namkane, Sirilux Kaewsirirung, Nongnuch Ketui, Worawit Fankam-ai","doi":"10.1109/iSAI-NLP56921.2022.9960283","DOIUrl":"https://doi.org/10.1109/iSAI-NLP56921.2022.9960283","url":null,"abstract":"Pumpkins contain a significant amount of beta-carotene. Beta-Carotene has numerous biological functions in the human body and because human is not able to synthesize any of them, it is necessary to supply these valuable compounds with food or pharmaceuticals. Internet of thing (IoT) in agriculture is not only reduce the man efforts but also improve the productivity and the efficiency. This research is primarily about the study of effect of beta-carotene in pumpkin between polyhouse and outdoor, because polyhouse is a closed structure protect the plants from weather conditions, insect and pest attacks. The irrigation in polyhouse using automatic drip irrigation, which operate according to the soil moisture threshold. Air temperature control using Fan based on temperature threshold. Analysis of beta carotene contain in pumpkin samples with polyhouse and outdoor process were determined by the samples were collected for three time periods found that the linear regression equation of the curve was y = 0.2111x-0.09161, with a coefficient r2 = 0.9975. The result show that, plants growing in the green house most are higher than the outdoor.","PeriodicalId":399019,"journal":{"name":"2022 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128541875","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 : 2022-11-05DOI: 10.1109/iSAI-NLP56921.2022.9960249
Viet Tinh Nguyen, Rujira Chaysiri
In the last decade, the threat of climate change and energy insecurity has put pressure on governments to search for alternatives energy sources to replace fossil fuels. As such, when more and more renewable energy projects have been developed, the number of related decision-making problems also increase. For solar energy projects, location selection is one of the most important and complex decision-making problems which involve both quantitative and qualitative criteria. This study aims to introduce a Spherical Fuzzy based MCDM model, utilizing Analytic Hierarchy Process (AHP) and Višekriterijumsko kompromisno rangiranje (VIKOR) methods. The proposed model is applied to case study in Vietnam to demonstrate its feasibility. The results suggests that, among the eights potential locations, Soc Trang (SP06) is the optimal location.
{"title":"Spherical Fuzzy AHP-VIKOR Model Application in Solar Energy Location Selection Problem: A Case Study in Vietnam","authors":"Viet Tinh Nguyen, Rujira Chaysiri","doi":"10.1109/iSAI-NLP56921.2022.9960249","DOIUrl":"https://doi.org/10.1109/iSAI-NLP56921.2022.9960249","url":null,"abstract":"In the last decade, the threat of climate change and energy insecurity has put pressure on governments to search for alternatives energy sources to replace fossil fuels. As such, when more and more renewable energy projects have been developed, the number of related decision-making problems also increase. For solar energy projects, location selection is one of the most important and complex decision-making problems which involve both quantitative and qualitative criteria. This study aims to introduce a Spherical Fuzzy based MCDM model, utilizing Analytic Hierarchy Process (AHP) and Višekriterijumsko kompromisno rangiranje (VIKOR) methods. The proposed model is applied to case study in Vietnam to demonstrate its feasibility. The results suggests that, among the eights potential locations, Soc Trang (SP06) is the optimal location.","PeriodicalId":399019,"journal":{"name":"2022 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131344439","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 : 2022-11-05DOI: 10.1109/iSAI-NLP56921.2022.9960244
Prapavarin Buranananont, A. Dumrongsiri, P. Chanvarasuth, Pornpimol Chongphaisal
In Thailand, the coffee shop business grows continuously. A coffee shop is where coffee is served as the primary beverage with food, and other drinks are only available as sub-component, The coffee shop can be described as a third place besides the working place and home where people go to meet, relax, and socialize with others. This research aims to study the factors that affect consumers' purchase intention in the coffee shop. The sample group of the study was 385 respondents. The data was collected through an online questionnaire survey given. This research was analyzed using a multiple regression method with the IBM SPSS Statistics (Statistical Package for the Social Science) version 26 to collect the data to produce the statistical analysis result. More importantly, this research has value for the organizations that want to maintain consumers for their coffee shops with a better understanding.
在泰国，咖啡店业务持续增长。咖啡店是咖啡作为食物的主要饮料，其他饮料只能作为次要成分提供的地方。咖啡店可以被描述为除了工作场所和家之外的第三个地方，人们去那里见面，放松，与他人交往。本研究旨在研究影响消费者在咖啡店购买意愿的因素。该研究的样本组为385名受访者。数据是通过在线问卷调查收集的。本研究使用IBM SPSS Statistics (Statistical Package for the Social Science) version 26收集数据，采用多元回归方法进行分析，得出统计分析结果。更重要的是，这项研究对那些想要更好地了解他们的咖啡店的消费者的组织有价值。
{"title":"Factors Affecting Purchase Intention to Coffee Shop","authors":"Prapavarin Buranananont, A. Dumrongsiri, P. Chanvarasuth, Pornpimol Chongphaisal","doi":"10.1109/iSAI-NLP56921.2022.9960244","DOIUrl":"https://doi.org/10.1109/iSAI-NLP56921.2022.9960244","url":null,"abstract":"In Thailand, the coffee shop business grows continuously. A coffee shop is where coffee is served as the primary beverage with food, and other drinks are only available as sub-component, The coffee shop can be described as a third place besides the working place and home where people go to meet, relax, and socialize with others. This research aims to study the factors that affect consumers' purchase intention in the coffee shop. The sample group of the study was 385 respondents. The data was collected through an online questionnaire survey given. This research was analyzed using a multiple regression method with the IBM SPSS Statistics (Statistical Package for the Social Science) version 26 to collect the data to produce the statistical analysis result. More importantly, this research has value for the organizations that want to maintain consumers for their coffee shops with a better understanding.","PeriodicalId":399019,"journal":{"name":"2022 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123257913","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 : 2022-11-05DOI: 10.1109/iSAI-NLP56921.2022.9960274
Romrawin Chumpu, Pitchayagan Temniranrat, S. Marukatat
Facial identification has recently been a legal con-cern for protecting one's identity and personal confidentiality. Many face synthesis techniques were used to safeguard individual users' data. This work presents a technique for generating synthetic faces from in-the-wild face components. The face components, such as the eyes, eyebrows, nose, and mouth, were extracted from a facial landmark of in-the-wild images and ran-domly replaced with the original image. Generative Adversarial Networks (GANs) for face restoration were then used to denoise the swapped image while preserving the original colorization. The experiments on face swapping with ten thousand of wild images demonstrate an average of 0.723 difference from the source image. The result shows that our face component swapping technique could be an effective lawful way to use facial data in the future.
{"title":"Synthetic face generation from in-the-wild face components swapping","authors":"Romrawin Chumpu, Pitchayagan Temniranrat, S. Marukatat","doi":"10.1109/iSAI-NLP56921.2022.9960274","DOIUrl":"https://doi.org/10.1109/iSAI-NLP56921.2022.9960274","url":null,"abstract":"Facial identification has recently been a legal con-cern for protecting one's identity and personal confidentiality. Many face synthesis techniques were used to safeguard individual users' data. This work presents a technique for generating synthetic faces from in-the-wild face components. The face components, such as the eyes, eyebrows, nose, and mouth, were extracted from a facial landmark of in-the-wild images and ran-domly replaced with the original image. Generative Adversarial Networks (GANs) for face restoration were then used to denoise the swapped image while preserving the original colorization. The experiments on face swapping with ten thousand of wild images demonstrate an average of 0.723 difference from the source image. The result shows that our face component swapping technique could be an effective lawful way to use facial data in the future.","PeriodicalId":399019,"journal":{"name":"2022 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126429981","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 : 2022-11-05DOI: 10.1109/iSAI-NLP56921.2022.9960242
I. Chatunapalak, W. Kongprawechnon, J. Kudtongngam
This research has proposed to utilize the combination of Machine Learning models (ML models) to optimally forecast the energy demand in Thailand. The various ML models are explored in which the individual and the combination of ML models are each optimized and evaluated for their best achievable performances. Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) are utilized to compare models' performances. A total of 4 ML models are executed, which include Artificial Neural Network (ANN), Decision Tree (DT), Random Forest (RF) Ensemble and proposed Vote Ensemble models. The results show that, by means of ensemble or model combination, the Vote Ensemble model could perform well with the lowest RMSE for training and testing of 613.63 and 666.52 and the lowest MAPE of 3.59% accordingly while also using less execution time of 3 minutes and 56 seconds.
{"title":"Long-Term Energy Demand Forecasting in Thailand with Ensemble Prediction Model","authors":"I. Chatunapalak, W. Kongprawechnon, J. Kudtongngam","doi":"10.1109/iSAI-NLP56921.2022.9960242","DOIUrl":"https://doi.org/10.1109/iSAI-NLP56921.2022.9960242","url":null,"abstract":"This research has proposed to utilize the combination of Machine Learning models (ML models) to optimally forecast the energy demand in Thailand. The various ML models are explored in which the individual and the combination of ML models are each optimized and evaluated for their best achievable performances. Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) are utilized to compare models' performances. A total of 4 ML models are executed, which include Artificial Neural Network (ANN), Decision Tree (DT), Random Forest (RF) Ensemble and proposed Vote Ensemble models. The results show that, by means of ensemble or model combination, the Vote Ensemble model could perform well with the lowest RMSE for training and testing of 613.63 and 666.52 and the lowest MAPE of 3.59% accordingly while also using less execution time of 3 minutes and 56 seconds.","PeriodicalId":399019,"journal":{"name":"2022 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"226 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121303252","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 : 2022-11-05DOI: 10.1109/iSAI-NLP56921.2022.9960270
S. Vorapojpisut, Karishma Agrawal
Temporal behaviors, e.g., cycle time and throughput, are among essential key performance indicators for the management of manufacturing processes. This paper presents a statistical model that captures the processing time spent throughout a production line using RSSI data acquired from Bluetooth Low Energy (BLE) network. First, a Hidden Semi-Markov Model (HSMM) is formulated based on the characteristics of production processes. Then, a learning problem is discussed for the re-estimation of state duration probability distribution using the forward-backward algorithm. The Kullback- Leibler Divergence is used to verify the accuracy by comparing between the original and estimated state duration probability distribution with a score of 0.0573. Finally, physical experiment was performed to evaluate the proposed method.
{"title":"Modeling of Manufacturing Processes using Hidden Semi-Markov Model and RSSI data","authors":"S. Vorapojpisut, Karishma Agrawal","doi":"10.1109/iSAI-NLP56921.2022.9960270","DOIUrl":"https://doi.org/10.1109/iSAI-NLP56921.2022.9960270","url":null,"abstract":"Temporal behaviors, e.g., cycle time and throughput, are among essential key performance indicators for the management of manufacturing processes. This paper presents a statistical model that captures the processing time spent throughout a production line using RSSI data acquired from Bluetooth Low Energy (BLE) network. First, a Hidden Semi-Markov Model (HSMM) is formulated based on the characteristics of production processes. Then, a learning problem is discussed for the re-estimation of state duration probability distribution using the forward-backward algorithm. The Kullback- Leibler Divergence is used to verify the accuracy by comparing between the original and estimated state duration probability distribution with a score of 0.0573. Finally, physical experiment was performed to evaluate the proposed method.","PeriodicalId":399019,"journal":{"name":"2022 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"653 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122962180","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}
Human activity recognition (HAR) and fall detection using smartphone sensors are currently popular because they can be extended to many useful applications especially when a person needs an urgent treatment such as a fall. Several methods based on machine learning (ML) and deep learning (DL) have been proposed to improve classification performances. In this work, we propose hybrid models of convolutional neural network (CNN) models and ML algorithms to classify human activities and falls using smartphone-sensor data. The CNN model will be used as feature extraction to extract a set of features. Thereafter, the ML algorithm will apply this set of features to predict the corresponding activity and fall. Several combinations of CNN models and ML algorithms are investigated on two public datasets: UniMiB SHAR and UMAFall. Their accuracy scores are compared in order to determine the best hybrid model. On the UniMiB SHAR dataset, the hybrid model based on the AlexN et model and the extra trees algorithm achieves the highest accuracy score of 95.27%. On the UMAFall dataset, the hybrid model based on the Xception model and the support vector machine/k-nearest neighbors/extra trees algorithms offer the highest accuracy score of 82.24 %.
{"title":"Smartphone-Based Human Activity and Fall Recognition Using Deep Feature Extraction and Machine-Learning Classifiers","authors":"Laksamee Nooyimsai, Onnicha Pakdeepong, Supajitra Chatchawalvoradech, Tipkasem Phiakhan, Seksan Laitrakun","doi":"10.1109/iSAI-NLP56921.2022.9960250","DOIUrl":"https://doi.org/10.1109/iSAI-NLP56921.2022.9960250","url":null,"abstract":"Human activity recognition (HAR) and fall detection using smartphone sensors are currently popular because they can be extended to many useful applications especially when a person needs an urgent treatment such as a fall. Several methods based on machine learning (ML) and deep learning (DL) have been proposed to improve classification performances. In this work, we propose hybrid models of convolutional neural network (CNN) models and ML algorithms to classify human activities and falls using smartphone-sensor data. The CNN model will be used as feature extraction to extract a set of features. Thereafter, the ML algorithm will apply this set of features to predict the corresponding activity and fall. Several combinations of CNN models and ML algorithms are investigated on two public datasets: UniMiB SHAR and UMAFall. Their accuracy scores are compared in order to determine the best hybrid model. On the UniMiB SHAR dataset, the hybrid model based on the AlexN et model and the extra trees algorithm achieves the highest accuracy score of 95.27%. On the UMAFall dataset, the hybrid model based on the Xception model and the support vector machine/k-nearest neighbors/extra trees algorithms offer the highest accuracy score of 82.24 %.","PeriodicalId":399019,"journal":{"name":"2022 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125080918","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：