Yeong-Yuh Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yeong-Yuh Xu has a total of 29 co-authors. Here are the top co-authors by number of shared publications:
- Angelica Te-Hui Hao
Department of Nursing, Hungkuang University, Taichung, Taiwan
2 shared publications - Chen-Yu Wang
Department of Critical Care Medicine, Taichung Veterans General Hospital, Taichung 407219, Taiwan
2 shared publications - Hsiao-Tien Pao
Department of Management Science, National Chiao Tung University, Hsinchu, Taiwan, ROC
2 shared publications - Hui-Chiao Yang
Department of Chest Medicine, Division of Respiratory Therapy, Taichung Veterans General Hospital, Taichung, Taiwan
2 shared publications - Ming-Cheng Chan
Department of Critical Care Medicine, Taichung Veterans General Hospital, Taichung, Taiwan
2 shared publications - Shao-Jen Weng
Department of Industrial Engineering and Enterprise Information, Tunghai University, Taichung, Taiwan
2 shared publications
Latest Publications
- 28 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (11) , 17822-17834
Mobility-Aware Hand Gesture Recognition With Inductive Proximity Sensors
Hand gesture recognition is widely regarded as one of the most promising human–machine interfaces (HMIs). In wrist-worn devices, conventional gesture recognition methods primarily rely on contact-based sensing techniques such as electromyography (EMG) and inertial measurement units (IMUs). However, contact-based approaches are inherently susceptible to environmental disturbances—including sweat and dust—and may fail under loose-wear conditions. Moreover, mobility-related issues, such as rewearing and cross-user, remain critical challenges for the real-world deployment of wrist-based gesture recognition. This study employs an inductive-proximity sensing technique to capture wrist-contour–induced mechanomyography (MMG) information, enabling a contactless wrist-worn gesture detection approach. We further propose a mobility-aware learning framework that achieves online gesture recognition under both cross-user and re-wearing scenarios. The contributions of this work are threefold: (1) development of a mobility-aware learning framework comprising motion detection, unsupervised clustering, and cluster–class mapping; (2) incorporation of virtual data generation into the mobility-aware learning to address data-streaming challenges in real-time gesture recognition; and (3) design of a hardware circuit integrating an inductive-proximity sensor, and implementation of a complete MMG watch prototype with firmware-level mobility-aware learning. Experimental results obtained from the MMG watch prototype demonstrate that, the proposed model achieves an average accuracy of 92.93% in the re-wearing scenario, yielding a 31% improvement compared to models that do not address mobility issues. In the cross-user scenario, the system achieves an average accuracy above 93%, with a mean per-gesture recognition latency of 9.66 ms. Overall, the highly integrated sensing and learning system are compatible with modern smartwatch architectures and provide real-time, stable gesture-streaming capabilities.
- 28 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (11) , 17822-17834
Mobility-Aware Hand Gesture Recognition With Inductive Proximity Sensors
Hand gesture recognition is widely regarded as one of the most promising human–machine interfaces (HMIs). In wrist-worn devices, conventional gesture recognition methods primarily rely on contact-based sensing techniques such as electromyography (EMG) and inertial measurement units (IMUs). However, contact-based approaches are inherently susceptible to environmental disturbances—including sweat and dust—and may fail under loose-wear conditions. Moreover, mobility-related issues, such as rewearing and cross-user, remain critical challenges for the real-world deployment of wrist-based gesture recognition. This study employs an inductive-proximity sensing technique to capture wrist-contour–induced mechanomyography (MMG) information, enabling a contactless wrist-worn gesture detection approach. We further propose a mobility-aware learning framework that achieves online gesture recognition under both cross-user and re-wearing scenarios. The contributions of this work are threefold: (1) development of a mobility-aware learning framework comprising motion detection, unsupervised clustering, and cluster–class mapping; (2) incorporation of virtual data generation into the mobility-aware learning to address data-streaming challenges in real-time gesture recognition; and (3) design of a hardware circuit integrating an inductive-proximity sensor, and implementation of a complete MMG watch prototype with firmware-level mobility-aware learning. Experimental results obtained from the MMG watch prototype demonstrate that, the proposed model achieves an average accuracy of 92.93% in the re-wearing scenario, yielding a 31% improvement compared to models that do not address mobility issues. In the cross-user scenario, the system achieves an average accuracy above 93%, with a mean per-gesture recognition latency of 9.66 ms. Overall, the highly integrated sensing and learning system are compatible with modern smartwatch architectures and provide real-time, stable gesture-streaming capabilities.
- Open Access
- 21 May 2025
- Published by JMIR Publications Inc. in JMIR Medical Informatics
- Vol. 13 (1) , e64592
Prediction of Spontaneous Breathing Trial Outcome in Critically Ill-Ventilated Patients Using Deep Learning: Development and Verification Study
Background: Long-term ventilator-dependent patients often face problems such as decreased quality of life, increased mortality, and increased medical costs. Respiratory therapists must perform complex and time-consuming ventilator weaning assessments, which typically take 48-72 hours. Traditional disengagement methods rely on manual evaluation and are susceptible to subjectivity, human errors, and low efficiency. Objective: This study aims to develop an artificial intelligence–based prediction model to predict whether a patient can successfully pass a spontaneous breathing trial (SBT) using the patient’s clinical data collected before SBT initiation. Instead of comparing different SBT strategies or analyzing their impact on extubation success, this study focused on establishing a data-driven approach under a fixed SBT strategy to provide an objective and efficient assessment tool. Through this model, we aim to enhance the accuracy and efficiency of ventilator weaning assessments, reduce unnecessary SBT attempts, optimize intensive care unit resource usage, and ultimately improve the quality of care for ventilator-dependent patients. Methods: This study used a retrospective cohort study and developed a novel deep learning architecture, hybrid CNN-MLP (convolutional neural network–multilayer perceptron), for analysis. Unlike the traditional CNN-MLP classification method, hybrid CNN-MLP performs feature learning and fusion by interleaving CNN and MLP layers so that data features can be extracted and integrated at different levels, thereby improving the flexibility and prediction accuracy of the model. The study participants were patients aged 20 years or older hospitalized in the intensive care unit of a medical center in central Taiwan between January 1, 2016, and December 31, 2022. A total of 3686 patients were included in the study, and 6536 pre-SBT clinical records were collected before each SBT of these patients, of which 3268 passed the SBT and 3268 failed. Results: The model performed well in predicting SBT outcomes. The training dataset’s precision is 99.3% (2443/2460 records), recall is 93.5% (2443/2614 records), specificity is 99.3% (2597/2614 records), and F1-score is 0.963. In the test dataset, the model maintains accuracy with a precision of 89.2% (561/629 records), a recall of 85.8% (561/654 records), a specificity of 89.6% (586/654 records), and an F1-score of 0.875. These results confirm the reliability of the model and its potential for clinical application. Conclusions: This study successfully developed a deep learning–based SBT prediction model that can be used as an objective and efficient ventilator weaning assessment tool. The model's performance shows that it can be integrated into clinical workflow, improve the quality of patient care, and reduce ventilator dependence, which is an important step in improving the effectiveness of respiratory therapy.
- 21 May 2025
- Published by JMIR Publications Inc. in JMIR Medical Informatics
- Vol. 13 (1) , e64592
Open AccessPrediction of Spontaneous Breathing Trial Outcome in Critically Ill-Ventilated Patients Using Deep Learning: Development and Verification Study
Background: Long-term ventilator-dependent patients often face problems such as decreased quality of life, increased mortality, and increased medical costs. Respiratory therapists must perform complex and time-consuming ventilator weaning assessments, which typically take 48-72 hours. Traditional disengagement methods rely on manual evaluation and are susceptible to subjectivity, human errors, and low efficiency. Objective: This study aims to develop an artificial intelligence–based prediction model to predict whether a patient can successfully pass a spontaneous breathing trial (SBT) using the patient’s clinical data collected before SBT initiation. Instead of comparing different SBT strategies or analyzing their impact on extubation success, this study focused on establishing a data-driven approach under a fixed SBT strategy to provide an objective and efficient assessment tool. Through this model, we aim to enhance the accuracy and efficiency of ventilator weaning assessments, reduce unnecessary SBT attempts, optimize intensive care unit resource usage, and ultimately improve the quality of care for ventilator-dependent patients. Methods: This study used a retrospective cohort study and developed a novel deep learning architecture, hybrid CNN-MLP (convolutional neural network–multilayer perceptron), for analysis. Unlike the traditional CNN-MLP classification method, hybrid CNN-MLP performs feature learning and fusion by interleaving CNN and MLP layers so that data features can be extracted and integrated at different levels, thereby improving the flexibility and prediction accuracy of the model. The study participants were patients aged 20 years or older hospitalized in the intensive care unit of a medical center in central Taiwan between January 1, 2016, and December 31, 2022. A total of 3686 patients were included in the study, and 6536 pre-SBT clinical records were collected before each SBT of these patients, of which 3268 passed the SBT and 3268 failed. Results: The model performed well in predicting SBT outcomes. The training dataset’s precision is 99.3% (2443/2460 records), recall is 93.5% (2443/2614 records), specificity is 99.3% (2597/2614 records), and F1-score is 0.963. In the test dataset, the model maintains accuracy with a precision of 89.2% (561/629 records), a recall of 85.8% (561/654 records), a specificity of 89.6% (586/654 records), and an F1-score of 0.875. These results confirm the reliability of the model and its potential for clinical application. Conclusions: This study successfully developed a deep learning–based SBT prediction model that can be used as an objective and efficient ventilator weaning assessment tool. The model's performance shows that it can be integrated into clinical workflow, improve the quality of patient care, and reduce ventilator dependence, which is an important step in improving the effectiveness of respiratory therapy.
- Open Access
- 21 July 2024
- Published by JMIR Publications Inc. in JMIR Preprints
Prediction of Spontaneous Breathing Trial Outcome in Critically Ill-Ventilated Patients Using Deep Learning: Development and Verification Study (Preprint)
BACKGROUND Long-term ventilator-dependent patients often face problems such as decreased quality of life, increased mortality, and increased medical costs. Respiratory therapists must perform complex and time-consuming ventilator weaning assessments, which typically take 48-72 hours. Traditional disengagement methods rely on manual evaluation and are susceptible to subjectivity, human errors, and low efficiency. OBJECTIVE This study aims to develop an artificial intelligence–based prediction model to predict whether a patient can successfully pass a spontaneous breathing trial (SBT) using the patient’s clinical data collected before SBT initiation. Instead of comparing different SBT strategies or analyzing their impact on extubation success, this study focused on establishing a data-driven approach under a fixed SBT strategy to provide an objective and efficient assessment tool. Through this model, we aim to enhance the accuracy and efficiency of ventilator weaning assessments, reduce unnecessary SBT attempts, optimize intensive care unit resource usage, and ultimately improve the quality of care for ventilator-dependent patients. METHODS This study used a retrospective cohort study and developed a novel deep learning architecture, hybrid CNN-MLP (convolutional neural network–multilayer perceptron), for analysis. Unlike the traditional CNN-MLP classification method, hybrid CNN-MLP performs feature learning and fusion by interleaving CNN and MLP layers so that data features can be extracted and integrated at different levels, thereby improving the flexibility and prediction accuracy of the model. The study participants were patients aged 20 years or older hospitalized in the intensive care unit of a medical center in central Taiwan between January 1, 2016, and December 31, 2022. A total of 3686 patients were included in the study, and 6536 pre-SBT clinical records were collected before each SBT of these patients, of which 3268 passed the SBT and 3268 failed. RESULTS The model performed well in predicting SBT outcomes. The training dataset’s precision is 99.3% (2443/2460 records), recall is 93.5% (2443/2614 records), specificity is 99.3% (2597/2614 records), and F1-score is 0.963. In the test dataset, the model maintains accuracy with a precision of 89.2% (561/629 records), a recall of 85.8% (561/654 records), a specificity of 89.6% (586/654 records), and an F1-score of 0.875. These results confirm the reliability of the model and its potential for clinical application. CONCLUSIONS This study successfully developed a deep learning–based SBT prediction model that can be used as an objective and efficient ventilator weaning assessment tool. The model's performance shows that it can be integrated into clinical workflow, improve the quality of patient care, and reduce ventilator dependence, which is an important step in improving the effectiveness of respiratory therapy. CLINICALTRIAL
- 21 July 2024
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessPrediction of Spontaneous Breathing Trial Outcome in Critically Ill-Ventilated Patients Using Deep Learning: Development and Verification Study (Preprint)
BACKGROUND Long-term ventilator-dependent patients often face problems such as decreased quality of life, increased mortality, and increased medical costs. Respiratory therapists must perform complex and time-consuming ventilator weaning assessments, which typically take 48-72 hours. Traditional disengagement methods rely on manual evaluation and are susceptible to subjectivity, human errors, and low efficiency. OBJECTIVE This study aims to develop an artificial intelligence–based prediction model to predict whether a patient can successfully pass a spontaneous breathing trial (SBT) using the patient’s clinical data collected before SBT initiation. Instead of comparing different SBT strategies or analyzing their impact on extubation success, this study focused on establishing a data-driven approach under a fixed SBT strategy to provide an objective and efficient assessment tool. Through this model, we aim to enhance the accuracy and efficiency of ventilator weaning assessments, reduce unnecessary SBT attempts, optimize intensive care unit resource usage, and ultimately improve the quality of care for ventilator-dependent patients. METHODS This study used a retrospective cohort study and developed a novel deep learning architecture, hybrid CNN-MLP (convolutional neural network–multilayer perceptron), for analysis. Unlike the traditional CNN-MLP classification method, hybrid CNN-MLP performs feature learning and fusion by interleaving CNN and MLP layers so that data features can be extracted and integrated at different levels, thereby improving the flexibility and prediction accuracy of the model. The study participants were patients aged 20 years or older hospitalized in the intensive care unit of a medical center in central Taiwan between January 1, 2016, and December 31, 2022. A total of 3686 patients were included in the study, and 6536 pre-SBT clinical records were collected before each SBT of these patients, of which 3268 passed the SBT and 3268 failed. RESULTS The model performed well in predicting SBT outcomes. The training dataset’s precision is 99.3% (2443/2460 records), recall is 93.5% (2443/2614 records), specificity is 99.3% (2597/2614 records), and F1-score is 0.963. In the test dataset, the model maintains accuracy with a precision of 89.2% (561/629 records), a recall of 85.8% (561/654 records), a specificity of 89.6% (586/654 records), and an F1-score of 0.875. These results confirm the reliability of the model and its potential for clinical application. CONCLUSIONS This study successfully developed a deep learning–based SBT prediction model that can be used as an objective and efficient ventilator weaning assessment tool. The model's performance shows that it can be integrated into clinical workflow, improve the quality of patient care, and reduce ventilator dependence, which is an important step in improving the effectiveness of respiratory therapy. CLINICALTRIAL
- 14 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 374-377
Machine Learning-Based Outcome Prediction of Acute Ischemic Stroke After Endovascular Treatment
This paper presents a convolutional neural network (CNN) model, underpinned by deep learning principles, tailored to predict endovascular intervention outcomes in patients with acute anterior circulation ischemic stroke. Recognizing the crucial role of predictive models in guiding patient prognosis and treatment selection, we partitioned our dataset into training and testing subsets. The proposed CNN model exhibited outstanding performance during its training phase, reaching a stellar accuracy rate of 100%. This performance was maintained during the testing phase, with an accuracy surpassing 94.12%. Such minimal deviation between training and testing accuracies underscores the robustness of our model and suggests a reduced likelihood of overfitting, enhancing its relevance for real-world clinical scenarios. These promising results signify a leap forward in harnessing the power of deep learning for medical prognostication, particularly in stroke intervention outcomes.
- 14 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 374-377
Machine Learning-Based Outcome Prediction of Acute Ischemic Stroke After Endovascular Treatment
This paper presents a convolutional neural network (CNN) model, underpinned by deep learning principles, tailored to predict endovascular intervention outcomes in patients with acute anterior circulation ischemic stroke. Recognizing the crucial role of predictive models in guiding patient prognosis and treatment selection, we partitioned our dataset into training and testing subsets. The proposed CNN model exhibited outstanding performance during its training phase, reaching a stellar accuracy rate of 100%. This performance was maintained during the testing phase, with an accuracy surpassing 94.12%. Such minimal deviation between training and testing accuracies underscores the robustness of our model and suggests a reduced likelihood of overfitting, enhancing its relevance for real-world clinical scenarios. These promising results signify a leap forward in harnessing the power of deep learning for medical prognostication, particularly in stroke intervention outcomes.
- Open Access
Game Fun Prediction Based on Frequency Domain Physiological Signals: Observational Study
Traditionally, the subjective questionnaire collected from game players is regarded as a primary tool to evaluate a video game. However, the subjective evaluation result may vary due to individual differences, and it is not easy to provide real-time feedback to optimize the user experience. This paper aims to develop an objective game fun prediction system. In this system, the wearables with photoplethysmography (PPG) sensors continuously measure the heartbeat signals of game players, and the frequency domain heart rate variability (HRV) parameters can be derived from the inter-beat interval (IBI) sequence. Frequency domain HRV parameters, such as low frequency(LF), high frequency(HF), and LF/HF ratio, highly correlate with the human’s emotion and mental status. Most existing works on emotion measurement during a game adopt time domain physiological signals such as heart rate and facial electromyography (EMG). Time domain signals can be easily interfered with by noises and environmental effects. The main contributions of this paper include (1) regarding the curve transition and standard deviation of LF/HF ratio as the objective game fun indicators and (2) proposing a linear model using objective indicators for game fun score prediction. The self-built dataset in this study involves ten healthy participants, comprising 36 samples. According to the analytical results, the linear model’s mean absolute error (MAE) was 4.16%, and the root mean square error (RMSE) was 5.07%. While integrating this prediction model with wearable-based HRV measurements, the proposed system can provide a solution to improve the user experience of video games.
Open AccessGame Fun Prediction Based on Frequency Domain Physiological Signals: Observational Study
Traditionally, the subjective questionnaire collected from game players is regarded as a primary tool to evaluate a video game. However, the subjective evaluation result may vary due to individual differences, and it is not easy to provide real-time feedback to optimize the user experience. This paper aims to develop an objective game fun prediction system. In this system, the wearables with photoplethysmography (PPG) sensors continuously measure the heartbeat signals of game players, and the frequency domain heart rate variability (HRV) parameters can be derived from the inter-beat interval (IBI) sequence. Frequency domain HRV parameters, such as low frequency(LF), high frequency(HF), and LF/HF ratio, highly correlate with the human’s emotion and mental status. Most existing works on emotion measurement during a game adopt time domain physiological signals such as heart rate and facial electromyography (EMG). Time domain signals can be easily interfered with by noises and environmental effects. The main contributions of this paper include (1) regarding the curve transition and standard deviation of LF/HF ratio as the objective game fun indicators and (2) proposing a linear model using objective indicators for game fun score prediction. The self-built dataset in this study involves ten healthy participants, comprising 36 samples. According to the analytical results, the linear model’s mean absolute error (MAE) was 4.16%, and the root mean square error (RMSE) was 5.07%. While integrating this prediction model with wearable-based HRV measurements, the proposed system can provide a solution to improve the user experience of video games.
- 1 October 2008
- Published by Elsevier in Expert Systems with Applications
- Vol. 35 (3) , 1444-1450
Constructing and application of multimedia TV-news archives
- 1 October 2008
- Published by Elsevier in Expert Systems with Applications
- Vol. 35 (3) , 1444-1450
Constructing and application of multimedia TV-news archives
- 1 June 2008
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 21578672,p. 233-236
Multimodal search for effective image retrieval
In this paper we present a multimodal approach for effectively searching and retrieving images. The proposed multimodal image query and retrieval (MIQR) method uses two or more types of query for accessing images - textual annotation associated with images and visual appearance, such as color, texture and positional features of objects in sample images. A user places a keyword-based query first and then retrieves desired images by visual content-based query. A prototype MIQR system was implemented and is available at http://140.113.216.66/WebImageSearch for online demo, and public evaluation. We also conducted experiments over a categorized Corel image collection and a non-categorized WWW image collection to show the performance of the proposed MIQR method.
- 1 June 2008
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 21578672,p. 233-236
Multimodal search for effective image retrieval
In this paper we present a multimodal approach for effectively searching and retrieving images. The proposed multimodal image query and retrieval (MIQR) method uses two or more types of query for accessing images - textual annotation associated with images and visual appearance, such as color, texture and positional features of objects in sample images. A user places a keyword-based query first and then retrieves desired images by visual content-based query. A prototype MIQR system was implemented and is available at http://140.113.216.66/WebImageSearch for online demo, and public evaluation. We also conducted experiments over a categorized Corel image collection and a non-categorized WWW image collection to show the performance of the proposed MIQR method.
- 28 February 2005
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4, 2789-2794
Forecasting electricity market prices: a neural network based approach
- 28 February 2005
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4, 2789-2794
Forecasting electricity market prices: a neural network based approach