Junyong You
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Junyong You has a total of 35 co-authors. Here are the top co-authors by number of shared publications:
- Jari Korhonen
School of Natural and Computing Sciences, University of Aberdeen, Aberdeen, U.K.
4 shared publications - Andrew Perkis
Norwegian University of Science and Technology, NTNU
3 shared publications - Arngeir Berge
National Centre for Emergency Primary Health Care, Health and Social Sciences, NORCE Research AS, Bergen, Norway
2 shared publications - Bin Hu
Kean University
2 shared publications - Christoph Trattner
MediaFutures, University of Bergen, Bergen, Norway
2 shared publications - Erik Zakariassen
National Centre for Emergency Primary Health Care, NORCE, Norwegian Research Centre
2 shared publications
Latest Publications
- 11 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. PP (1530437X) , 1
Hybrid-DP Diffusion for Privacy-Preserving WiFi CSI Sensing and Biometric Template Generation
WiFi Channel State Information (CSI) enables device-free human activity recognition (HAR) and RF sensing but may expose user- and environment-specific information when shared or used for model training. Recent studies further show that CSI can leak sensitive user attributes, such as identity and physical characteristics, motivating the need for privacy-preserving sensing frameworks. We propose Hybrid-DP Diffusion, a selective differentially private framework for synthesizing privacy-preserving CSI. The framework combines non-private diffusion pretraining with selective DP-SGD applied to identity-sensitive components, mitigating the fidelity degradation observed in prior differentially private generative models. Hybrid-DP is evaluated using both MLP and UNet backbones on UT-HAR and Widar 3.0 under matched privacy budgets. Results show that architectural inductive bias is critical: UNet-based diffusion achieves improved generative fidelity (lower FID, higher SSIM) and more stable optimization compared to MLP models under identical privacy constraints. Compared with Full-DP training, Hybrid-DP produces CSI samples with fewer artifacts while maintaining formal differential privacy guarantees and stable convergence behavior. Membership inference attacks remain close to random guessing, indicating limited exploitable leakage. Overall, the proposed framework achieves a balanced trade-off between generative fidelity and privacy protection, enabling practical privacy-preserving CSI synthesis for RF sensing and collaborative IoT data sharing.
- 11 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. PP (1530437X) , 1
Hybrid-DP Diffusion for Privacy-Preserving WiFi CSI Sensing and Biometric Template Generation
WiFi Channel State Information (CSI) enables device-free human activity recognition (HAR) and RF sensing but may expose user- and environment-specific information when shared or used for model training. Recent studies further show that CSI can leak sensitive user attributes, such as identity and physical characteristics, motivating the need for privacy-preserving sensing frameworks. We propose Hybrid-DP Diffusion, a selective differentially private framework for synthesizing privacy-preserving CSI. The framework combines non-private diffusion pretraining with selective DP-SGD applied to identity-sensitive components, mitigating the fidelity degradation observed in prior differentially private generative models. Hybrid-DP is evaluated using both MLP and UNet backbones on UT-HAR and Widar 3.0 under matched privacy budgets. Results show that architectural inductive bias is critical: UNet-based diffusion achieves improved generative fidelity (lower FID, higher SSIM) and more stable optimization compared to MLP models under identical privacy constraints. Compared with Full-DP training, Hybrid-DP produces CSI samples with fewer artifacts while maintaining formal differential privacy guarantees and stable convergence behavior. Membership inference attacks remain close to random guessing, indicating limited exploitable leakage. Overall, the proposed framework achieves a balanced trade-off between generative fidelity and privacy protection, enabling practical privacy-preserving CSI synthesis for RF sensing and collaborative IoT data sharing.
- Open Access
- 7 May 2026
- Published by JMIR Publications Inc. in JMIR Formative Research
- Vol. 10, e87082
Impact of a Prototype Combining Recommender Functionality With Structured Documentation on Operator Performance in Calls to Medical Communication Centers: Quasi-Experimental Feasibility Study
Background: Management of contacts to medical communication centers relies heavily on clinical judgment, contextual understanding, and communication skills. Decision support systems, intended to complement medical expertise, may, due to their rigidity, impede effective caller interaction and may, together with the obligatory documentation of calls, contribute to a workflow that draws attention away from the communication. Recommender systems have demonstrated potential in supporting decision-making across various domains by nudging individuals toward better choices without undermining autonomy. We built a prototype that combined artificial intelligence–based question recommendations with structured documentation (hereafter: the prototype) and conducted a feasibility study to test its influence on operators’ performance. Objective: This study aimed to examine whether the prototype influenced the operators’ performance during telephone triage. We hypothesized that the prototype would affect medical quality without affecting communication quality. Methods: A quasi-experimental pre- and posttest feasibility study was conducted in a simulated setting. Twenty-five operators were voluntarily recruited from 5 Norwegian medical communication centers, in which 22 operators contributed to both the pretest (before the prototype) and the posttest (with the prototype). The operators handled the same 15 medical cases presented by simulated callers, with a 5-month interval between the 2 sessions. The question recommender was trained on other data and then fine-tuned on the 15 scenarios used. Audio recordings of the calls were rated using the tool Assessment of Quality in Telephone Triage. Pre- and posttest values were compared, with overall medical and communication quality as the primary outcomes. Secondary outcomes included specific items related to medical content and communication, accuracy of triage, patient safety, call duration, and efficiency. Results: A total of 320 paired calls were analyzed. Overall medical quality improved significantly with use of the prototype, from a mean of 6.83 points pretest to 7.16 points posttest rated on a 10-point scale (difference 0.34, 95% CI 0.11-0.57; P=.004). The effect size was small (Cohen dz=0.16). No significant change was observed in overall communication quality, with a mean of 7.06 points pretest and 6.97 points posttest (difference –0.09 points, 95% CI –0.28 to 0.10; P=.35). A significant decrease from pre‑ to posttest was observed in the specific items “Collects information about the patient’s location” (P<.001) and “Ensures that the triage decision is understandable and feasible” (P=.002). None of the remaining secondary outcomes showed significant changes. Conclusions: The prototype yielded a modest improvement in medical quality within the scenario‑based test environment. Although overall communication quality remained unchanged, aspects of the interaction were negatively affected. Artificial intelligence–based question recommendations combined with structured documentation may serve as useful functionalities within a decision support system, but each functionality requires further testing and development before such technology can be implemented in the triage of unselected, real‑world calls.
- 7 May 2026
- Published by JMIR Publications Inc. in JMIR Formative Research
- Vol. 10, e87082
Open AccessImpact of a Prototype Combining Recommender Functionality With Structured Documentation on Operator Performance in Calls to Medical Communication Centers: Quasi-Experimental Feasibility Study
Background: Management of contacts to medical communication centers relies heavily on clinical judgment, contextual understanding, and communication skills. Decision support systems, intended to complement medical expertise, may, due to their rigidity, impede effective caller interaction and may, together with the obligatory documentation of calls, contribute to a workflow that draws attention away from the communication. Recommender systems have demonstrated potential in supporting decision-making across various domains by nudging individuals toward better choices without undermining autonomy. We built a prototype that combined artificial intelligence–based question recommendations with structured documentation (hereafter: the prototype) and conducted a feasibility study to test its influence on operators’ performance. Objective: This study aimed to examine whether the prototype influenced the operators’ performance during telephone triage. We hypothesized that the prototype would affect medical quality without affecting communication quality. Methods: A quasi-experimental pre- and posttest feasibility study was conducted in a simulated setting. Twenty-five operators were voluntarily recruited from 5 Norwegian medical communication centers, in which 22 operators contributed to both the pretest (before the prototype) and the posttest (with the prototype). The operators handled the same 15 medical cases presented by simulated callers, with a 5-month interval between the 2 sessions. The question recommender was trained on other data and then fine-tuned on the 15 scenarios used. Audio recordings of the calls were rated using the tool Assessment of Quality in Telephone Triage. Pre- and posttest values were compared, with overall medical and communication quality as the primary outcomes. Secondary outcomes included specific items related to medical content and communication, accuracy of triage, patient safety, call duration, and efficiency. Results: A total of 320 paired calls were analyzed. Overall medical quality improved significantly with use of the prototype, from a mean of 6.83 points pretest to 7.16 points posttest rated on a 10-point scale (difference 0.34, 95% CI 0.11-0.57; P=.004). The effect size was small (Cohen dz=0.16). No significant change was observed in overall communication quality, with a mean of 7.06 points pretest and 6.97 points posttest (difference –0.09 points, 95% CI –0.28 to 0.10; P=.35). A significant decrease from pre‑ to posttest was observed in the specific items “Collects information about the patient’s location” (P<.001) and “Ensures that the triage decision is understandable and feasible” (P=.002). None of the remaining secondary outcomes showed significant changes. Conclusions: The prototype yielded a modest improvement in medical quality within the scenario‑based test environment. Although overall communication quality remained unchanged, aspects of the interaction were negatively affected. Artificial intelligence–based question recommendations combined with structured documentation may serve as useful functionalities within a decision support system, but each functionality requires further testing and development before such technology can be implemented in the triage of unselected, real‑world calls.
- Open Access
- 4 November 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Impact of a Prototype Combining Recommender Functionality With Structured Documentation on Operator Performance in Calls to Medical Communication Centers: Quasi-Experimental Feasibility Study (Preprint)
BACKGROUND Management of contacts to medical communication centers relies heavily on clinical judgment, contextual understanding, and communication skills. Decision support systems, intended to complement medical expertise, may, due to their rigidity, impede effective caller interaction and may, together with the obligatory documentation of calls, contribute to a workflow that draws attention away from the communication. Recommender systems have demonstrated potential in supporting decision-making across various domains by nudging individuals toward better choices without undermining autonomy. We built a prototype that combined artificial intelligence–based question recommendations with structured documentation (hereafter: the prototype) and conducted a feasibility study to test its influence on operators’ performance. OBJECTIVE This study aimed to examine whether the prototype influenced the operators’ performance during telephone triage. We hypothesized that the prototype would affect medical quality without affecting communication quality. METHODS A quasi-experimental pre- and posttest feasibility study was conducted in a simulated setting. Twenty-five operators were voluntarily recruited from 5 Norwegian medical communication centers, in which 22 operators contributed to both the pretest (before the prototype) and the posttest (with the prototype). The operators handled the same 15 medical cases presented by simulated callers, with a 5-month interval between the 2 sessions. The question recommender was trained on other data and then fine-tuned on the 15 scenarios used. Audio recordings of the calls were rated using the tool Assessment of Quality in Telephone Triage. Pre- and posttest values were compared, with overall medical and communication quality as the primary outcomes. Secondary outcomes included specific items related to medical content and communication, accuracy of triage, patient safety, call duration, and efficiency. RESULTS A total of 320 paired calls were analyzed. Overall medical quality improved significantly with use of the prototype, from a mean of 6.83 points pretest to 7.16 points posttest rated on a 10-point scale (difference 0.34, 95% CI 0.11-0.57; P=.004). The effect size was small (Cohen dz=0.16). No significant change was observed in overall communication quality, with a mean of 7.06 points pretest and 6.97 points posttest (difference –0.09 points, 95% CI –0.28 to 0.10; P=.35). A significant decrease from pre‑ to posttest was observed in the specific items “Collects information about the patient’s location” (P<.001) and “Ensures that the triage decision is understandable and feasible” (P=.002). None of the remaining secondary outcomes showed significant changes. CONCLUSIONS The prototype yielded a modest improvement in medical quality within the scenario‑based test environment. Although overall communication quality remained unchanged, aspects of the interaction were negatively affected. Artificial intelligence–based question recommendations combined with structured documentation may serve as useful functionalities within a decision support system, but each functionality requires further testing and development before such technology can be implemented in the triage of unselected, real‑world calls. CLINICALTRIAL
- 4 November 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessImpact of a Prototype Combining Recommender Functionality With Structured Documentation on Operator Performance in Calls to Medical Communication Centers: Quasi-Experimental Feasibility Study (Preprint)
BACKGROUND Management of contacts to medical communication centers relies heavily on clinical judgment, contextual understanding, and communication skills. Decision support systems, intended to complement medical expertise, may, due to their rigidity, impede effective caller interaction and may, together with the obligatory documentation of calls, contribute to a workflow that draws attention away from the communication. Recommender systems have demonstrated potential in supporting decision-making across various domains by nudging individuals toward better choices without undermining autonomy. We built a prototype that combined artificial intelligence–based question recommendations with structured documentation (hereafter: the prototype) and conducted a feasibility study to test its influence on operators’ performance. OBJECTIVE This study aimed to examine whether the prototype influenced the operators’ performance during telephone triage. We hypothesized that the prototype would affect medical quality without affecting communication quality. METHODS A quasi-experimental pre- and posttest feasibility study was conducted in a simulated setting. Twenty-five operators were voluntarily recruited from 5 Norwegian medical communication centers, in which 22 operators contributed to both the pretest (before the prototype) and the posttest (with the prototype). The operators handled the same 15 medical cases presented by simulated callers, with a 5-month interval between the 2 sessions. The question recommender was trained on other data and then fine-tuned on the 15 scenarios used. Audio recordings of the calls were rated using the tool Assessment of Quality in Telephone Triage. Pre- and posttest values were compared, with overall medical and communication quality as the primary outcomes. Secondary outcomes included specific items related to medical content and communication, accuracy of triage, patient safety, call duration, and efficiency. RESULTS A total of 320 paired calls were analyzed. Overall medical quality improved significantly with use of the prototype, from a mean of 6.83 points pretest to 7.16 points posttest rated on a 10-point scale (difference 0.34, 95% CI 0.11-0.57; P=.004). The effect size was small (Cohen dz=0.16). No significant change was observed in overall communication quality, with a mean of 7.06 points pretest and 6.97 points posttest (difference –0.09 points, 95% CI –0.28 to 0.10; P=.35). A significant decrease from pre‑ to posttest was observed in the specific items “Collects information about the patient’s location” (P<.001) and “Ensures that the triage decision is understandable and feasible” (P=.002). None of the remaining secondary outcomes showed significant changes. CONCLUSIONS The prototype yielded a modest improvement in medical quality within the scenario‑based test environment. Although overall communication quality remained unchanged, aspects of the interaction were negatively affected. Artificial intelligence–based question recommendations combined with structured documentation may serve as useful functionalities within a decision support system, but each functionality requires further testing and development before such technology can be implemented in the triage of unselected, real‑world calls. CLINICALTRIAL
- Open Access
- 4 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 31423-31441
EffTEE: Efficient Image Classification and Object Detection on Mobile Devices Using Trusted Execution Environments
Deep neural networks (DNNs) have proven highly effective in image classification and object detection, with critical applications such as autonomous driving, security, Unmanned Aerial Vehicle (UAV) obstacle avoidance, and robot vision. However, the sensitivity of both input data and DNN architectures has increased the demand for secure execution on mobile devices. To address these concerns, hardware-based Trusted Execution Environments (TEEs), such as ARM TrustZone, are employed to ensure secure DNN operations. Despite their potential, deploying full-scale DNNs within mobile TEEs is challenging due to stringent resource and performance constraints. This work presents a novel security framework to enable the secure and efficient execution of DNNs within mobile TEEs. Specifically, the framework employs an effective dynamic suppression method that identifies and prunes unimportant neurons, resulting in a more compact and computationally efficient network model. Foundational neuron restructuring optimizes memory usage by deallocating storage for pruned neurons. Additionally, a dynamic slicing method combined with composite layer integration and loading strategies partitions pruned DNN models into multiple segments designed to fit within TEE memory constraints. Experiments conducted across various DNN models demonstrate statistically significant performance improvements. On the ImageNet dataset, EffTEE achieved an average Weight Compression Ratio (CR) of 82:3% ± 1:2%, a FLOPs CR of 85:7% ± 1:0%, and a 33:2% ± 1:4% reduction in inference time while maintaining97:2% accuracy. On CIFAR-10, EffTEE reduced inference time by 29:6%±1:2% with accuracy retention within ±0:8% of the original model. For object detection on the MS COCO dataset, EffTEE reduced inference time by 35:1%±1:5%, achieving consistent improvements in Frames Per Second (FPS). Overall, EffTEE achieves a significant 2 to 60 times reduction in inference time while maintaining accuracy comparable to existing methods for securing entire DNNs within mobile TEEs. Statistical validation, with a mean accuracy retention of 97:2% ± 0:8% and a p-value of <0.05, confirms the framework's robustness. These results highlight EffTEE's capability for robust and efficient DNN deployment in resource-constrained environments, balancing security, performance, and accuracy.
- 4 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 31423-31441
Open AccessEffTEE: Efficient Image Classification and Object Detection on Mobile Devices Using Trusted Execution Environments
Deep neural networks (DNNs) have proven highly effective in image classification and object detection, with critical applications such as autonomous driving, security, Unmanned Aerial Vehicle (UAV) obstacle avoidance, and robot vision. However, the sensitivity of both input data and DNN architectures has increased the demand for secure execution on mobile devices. To address these concerns, hardware-based Trusted Execution Environments (TEEs), such as ARM TrustZone, are employed to ensure secure DNN operations. Despite their potential, deploying full-scale DNNs within mobile TEEs is challenging due to stringent resource and performance constraints. This work presents a novel security framework to enable the secure and efficient execution of DNNs within mobile TEEs. Specifically, the framework employs an effective dynamic suppression method that identifies and prunes unimportant neurons, resulting in a more compact and computationally efficient network model. Foundational neuron restructuring optimizes memory usage by deallocating storage for pruned neurons. Additionally, a dynamic slicing method combined with composite layer integration and loading strategies partitions pruned DNN models into multiple segments designed to fit within TEE memory constraints. Experiments conducted across various DNN models demonstrate statistically significant performance improvements. On the ImageNet dataset, EffTEE achieved an average Weight Compression Ratio (CR) of 82:3% ± 1:2%, a FLOPs CR of 85:7% ± 1:0%, and a 33:2% ± 1:4% reduction in inference time while maintaining97:2% accuracy. On CIFAR-10, EffTEE reduced inference time by 29:6%±1:2% with accuracy retention within ±0:8% of the original model. For object detection on the MS COCO dataset, EffTEE reduced inference time by 35:1%±1:5%, achieving consistent improvements in Frames Per Second (FPS). Overall, EffTEE achieves a significant 2 to 60 times reduction in inference time while maintaining accuracy comparable to existing methods for securing entire DNNs within mobile TEEs. Statistical validation, with a mean accuracy retention of 97:2% ± 0:8% and a p-value of <0.05, confirms the framework's robustness. These results highlight EffTEE's capability for robust and efficient DNN deployment in resource-constrained environments, balancing security, performance, and accuracy.
- Open Access
- 20 December 2024
- Published by Qeios Ltd
Review of: "Grounding-IQA: Multimodal Language Grounding Model for Image Quality Assessment"
The paper introduces a new paradigm termed "Grounding-IQA," which aims to enhance Image Quality Assessment (IQA) methods by integrating multimodal referring and grounding capabilities. Traditional IQA approaches focus largely on holistic score predic...
- 20 December 2024
- Published by Qeios Ltd
Open AccessReview of: "Grounding-IQA: Multimodal Language Grounding Model for Image Quality Assessment"
The paper introduces a new paradigm termed "Grounding-IQA," which aims to enhance Image Quality Assessment (IQA) methods by integrating multimodal referring and grounding capabilities. Traditional IQA approaches focus largely on holistic score predic...
- 26 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Explainable Transformer Model for Short-Term Wind Forecast
Short-term wind field forecasting, particularly wind speed prediction, is a critical component in wind farm control. Traditional statistical techniques, such as autoregressive integrated moving average, and machine learning approaches, such as recurrent neural networks, have been widely used for short-term forecasting, but these methods are often difficult to interpret. In this study, we propose a transformer model that can capture the internal dependencies within previous wind measurements to predict wind speeds in the upcoming hours. Measurement data from the FINO1 research platform collected in 2007 and 2008, including wind speeds, direction, and air temperature at three altitudes (40m, 60m, and 80m), have been used to train and evaluate the model. With the help of the attention mechanism in the transformer model, we can interpret the impact of the previous wind profile on future predictions, thereby enabling the explanation of deep learning models for wind forecasting. Our experimental results on the FINO1 measurement data demonstrate exceptional performance of the proposed model in short-term wind forecasting.
- 26 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Explainable Transformer Model for Short-Term Wind Forecast
Short-term wind field forecasting, particularly wind speed prediction, is a critical component in wind farm control. Traditional statistical techniques, such as autoregressive integrated moving average, and machine learning approaches, such as recurrent neural networks, have been widely used for short-term forecasting, but these methods are often difficult to interpret. In this study, we propose a transformer model that can capture the internal dependencies within previous wind measurements to predict wind speeds in the upcoming hours. Measurement data from the FINO1 research platform collected in 2007 and 2008, including wind speeds, direction, and air temperature at three altitudes (40m, 60m, and 80m), have been used to train and evaluate the model. With the help of the attention mechanism in the transformer model, we can interpret the impact of the previous wind profile on future predictions, thereby enabling the explanation of deep learning models for wind forecasting. Our experimental results on the FINO1 measurement data demonstrate exceptional performance of the proposed model in short-term wind forecasting.
- 2 March 2023
- Published by Springer Nature
- p. 455-473
Visual Mechanisms Inspired Efficient Transformers for Image and Video Quality Assessment
- 2 March 2023
- Published by Springer Nature
- p. 455-473
Visual Mechanisms Inspired Efficient Transformers for Image and Video Quality Assessment
- Open Access
Visual Mechanisms Inspired Efficient Transformers for Image and Video Quality Assessment
Visual (image, video) quality assessments can be modelled by visual features in different domains, e.g., spatial, frequency, and temporal domains. Perceptual mechanisms in the human visual system (HVS) play a crucial role in generation of quality perception. This paper proposes a general framework for no-reference visual quality assessment using efficient windowed transformer architectures. A lightweight module for multi-stage channel attention is integrated into Swin (shifted window) Transformer. Such module can represent appropriate perceptual mechanisms in image quality assessment (IQA) to build an accurate IQA model. Meanwhile, representative features for image quality perception in the spatial and frequency domains can also be derived from the IQA model, which are then fed into another windowed transformer architecture for video quality assessment (VQA). The VQA model efficiently reuses attention information across local windows to tackle the issue of expensive time and memory complexities of original transformer. Experimental results on both large-scale IQA and VQA databases demonstrate that the proposed quality assessment models outperform other state-of-the-art models by large margins. The complete source code will be published on Github.
Open AccessVisual Mechanisms Inspired Efficient Transformers for Image and Video Quality Assessment
Visual (image, video) quality assessments can be modelled by visual features in different domains, e.g., spatial, frequency, and temporal domains. Perceptual mechanisms in the human visual system (HVS) play a crucial role in generation of quality perception. This paper proposes a general framework for no-reference visual quality assessment using efficient windowed transformer architectures. A lightweight module for multi-stage channel attention is integrated into Swin (shifted window) Transformer. Such module can represent appropriate perceptual mechanisms in image quality assessment (IQA) to build an accurate IQA model. Meanwhile, representative features for image quality perception in the spatial and frequency domains can also be derived from the IQA model, which are then fed into another windowed transformer architecture for video quality assessment (VQA). The VQA model efficiently reuses attention information across local windows to tackle the issue of expensive time and memory complexities of original transformer. Experimental results on both large-scale IQA and VQA databases demonstrate that the proposed quality assessment models outperform other state-of-the-art models by large margins. The complete source code will be published on Github.
- Open Access
A set of web-based public decision support tools for integrated planning and management in aquaculture
- Open Access
- 1 January 2022
- Published by Elsevier in Journal of Visual Communication and Image Representation
- Vol. 82
Attention integrated hierarchical networks for no-reference image quality assessment
- 1 January 2022
- Published by Elsevier in Journal of Visual Communication and Image Representation
- Vol. 82
Open AccessAttention integrated hierarchical networks for no-reference image quality assessment