Jun-Hong Shen
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
Jun-Hong Shen has a total of 57 co-authors. Here are the top co-authors by number of shared publications:
- Ching-Ta Lu
Department of Communications Engineering, Feng Chia University, Taichung City 407102, Taiwan
8 shared publications Ching-Ta Lu
Asia University
3 shared publications- Ming-Shen Jian
National Formosa University
3 shared publications Ye-In Chang
National Sun Yat-Sen University
3 shared publications- Ling-Ling Wang
Asia University
2 shared publications Mu-Yen Chen
National Taichung University of Science and Technology
2 shared publications
Latest Publications
- Open Access
- 17 July 2026
- Published by Elsevier in SSRN Electronic Journal
When the doctor is not real: How authenticity disclosure shapes trust, eeriness, and sharing intentions across age groups
Highly realistic AI-generated presenters are increasingly common in online health communication, yet little is known about how viewers respond when the synthetic origin of such content is disclosed after viewing. Drawing on a dual-stimulus extension of the Stimulus-Organism-Response model integrated with expectancy violation and uncanny valley perspectives, this single-stimulus mixed-methods study examined how 70 viewers in three age groups (18–29, 30–54, 55+; older group n = 10, treated as exploratory) judged, trusted, and reacted to a fact-checked AI-generated health education video before and after authenticity disclosure; 35 participants additionally completed semi-structured interviews. Asked retrospectively immediately after the disclosure, 52.9% reported having identified the video as AI-generated before being told; reported identification was most frequent among young adults and was associated most strongly with AI experience and AI literacy. Middle-aged adults reported the highest pre-disclosure source and message credibility. After a neutral disclosure, perceived deception and trust erosion remained moderate and did not differ by age group or prior identification status; instead, higher pre-disclosure credibility was associated with weaker negative reactions. Eeriness was stronger among viewers who had correctly identified the video as synthetic and among young adults. Information acceptance exceeded sharing intention; middle-aged adults were the most accepting, and AI literacy was negatively associated with sharing intention. Interviews revealed a conditional-acceptance pattern contingent on content risk, institutional endorsement, and verifiability. These findings, exploratory with respect to age, inform the design of AI disclosure and labeling for health media.
- 17 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessWhen the doctor is not real: How authenticity disclosure shapes trust, eeriness, and sharing intentions across age groups
Highly realistic AI-generated presenters are increasingly common in online health communication, yet little is known about how viewers respond when the synthetic origin of such content is disclosed after viewing. Drawing on a dual-stimulus extension of the Stimulus-Organism-Response model integrated with expectancy violation and uncanny valley perspectives, this single-stimulus mixed-methods study examined how 70 viewers in three age groups (18–29, 30–54, 55+; older group n = 10, treated as exploratory) judged, trusted, and reacted to a fact-checked AI-generated health education video before and after authenticity disclosure; 35 participants additionally completed semi-structured interviews. Asked retrospectively immediately after the disclosure, 52.9% reported having identified the video as AI-generated before being told; reported identification was most frequent among young adults and was associated most strongly with AI experience and AI literacy. Middle-aged adults reported the highest pre-disclosure source and message credibility. After a neutral disclosure, perceived deception and trust erosion remained moderate and did not differ by age group or prior identification status; instead, higher pre-disclosure credibility was associated with weaker negative reactions. Eeriness was stronger among viewers who had correctly identified the video as synthetic and among young adults. Information acceptance exceeded sharing intention; middle-aged adults were the most accepting, and AI literacy was negatively associated with sharing intention. Interviews revealed a conditional-acceptance pattern contingent on content risk, institutional endorsement, and verifiability. These findings, exploratory with respect to age, inform the design of AI disclosure and labeling for health media.
- 14 November 2025
- Published by Taylor & Francis in Enterprise Information Systems
- Vol. 20 (5-6)
A novel attention-based multimodal framework for automatically detecting cryptocurrency fake news
- 14 November 2025
- Published by Taylor & Francis in Enterprise Information Systems
- Vol. 20 (5-6)
A novel attention-based multimodal framework for automatically detecting cryptocurrency fake news
- 21 January 2024
- Published by Springer Nature
- p. 152-159
An Efficient Edge-Based Index for Processing Collective Spatial Keyword Query on Road Networks
- 21 January 2024
- Published by Springer Nature
- p. 152-159
An Efficient Edge-Based Index for Processing Collective Spatial Keyword Query on Road Networks
- 21 November 2023
- Published by Springer Nature in Multimedia Tools and Applications
- p. 1-23
A speech denoising demonstration system using multi-model deep-learning neural networks
- 21 November 2023
- Published by Springer Nature in Multimedia Tools and Applications
- p. 1-23
A speech denoising demonstration system using multi-model deep-learning neural networks
- 1 January 2022
- Published by EAGE Publications bv
- Vol. 2022 (1) , 1-5
Self-Potential Survey at the Yuhuang Hydrothermal Field on the Southwest Indian Ridge
- 1 January 2022
- Published by EAGE Publications bv
- Vol. 2022 (1) , 1-5
Self-Potential Survey at the Yuhuang Hydrothermal Field on the Southwest Indian Ridge
- Open Access
Theoretically Principled Deep RL Acceleration via Nearest Neighbor Function Approximation
Recently, deep reinforcement learning (RL) has achieved remarkable empirical success by integrating deep neural networks into RL frameworks. However, these algorithms often require a large number of training samples and admit little theoretical understanding. To mitigate these issues, we propose a theoretically principled nearest neighbor (NN) function approximator that can improve the value networks in deep RL methods. Inspired by human similarity judgments, the NN approximator estimates the action values using rollouts on past observations and can provably obtain a small regret bound that depends only on the intrinsic complexity of the environment. We present (1) Nearest Neighbor Actor-Critic (NNAC), an online policy gradient algorithm that demonstrates the practicality of combining function approximation with deep RL, and (2) a plug-and-play NN update module that aids the training of existing deep RL methods. Experiments on classical control and MuJoCo locomotion tasks show that the NN-accelerated agents achieve higher sample efficiency and stability than the baseline agents. Based on its theoretical benefits, we believe that the NN approximator can be further applied to other complex domains to speed-up learning.
Open AccessTheoretically Principled Deep RL Acceleration via Nearest Neighbor Function Approximation
Recently, deep reinforcement learning (RL) has achieved remarkable empirical success by integrating deep neural networks into RL frameworks. However, these algorithms often require a large number of training samples and admit little theoretical understanding. To mitigate these issues, we propose a theoretically principled nearest neighbor (NN) function approximator that can improve the value networks in deep RL methods. Inspired by human similarity judgments, the NN approximator estimates the action values using rollouts on past observations and can provably obtain a small regret bound that depends only on the intrinsic complexity of the environment. We present (1) Nearest Neighbor Actor-Critic (NNAC), an online policy gradient algorithm that demonstrates the practicality of combining function approximation with deep RL, and (2) a plug-and-play NN update module that aids the training of existing deep RL methods. Experiments on classical control and MuJoCo locomotion tasks show that the NN-accelerated agents achieve higher sample efficiency and stability than the baseline agents. Based on its theoretical benefits, we believe that the NN approximator can be further applied to other complex domains to speed-up learning.
- Open Access
- 8 March 2021
- Published by MDPI AG in Sustainability
- Vol. 13 (5) , 2934
A Study on the Influence of Number/Distribution of Sensing Points of the Smart Insoles on the Center of Pressure Estimation for the Internet of Things Applications
The past decade has seen the emergence of numerous new wearable devices, including many that have been widely adopted by both physicians and consumers. In this paper, we discuss the design and application of smart insoles to measure gait and plantar pressure. Herein, we investigate the potential applications of insoles with fewer sensing spots and the consequent reduction in the amount of data acquired from both feet. The main purpose is to discuss the influence of the layout of these pressure sensing points of the insole design on the center of pressure (COP) calculation. The insole used in this study has 89 pressure sensing spots, and we used data from 36, 29, 20, and 11 sensing points in simplified calculation types. Among these four simplified calculation types, Type 1 exhibited the best accuracy of the COP calculation, and Type 4 obtained the worst results. Type 2 and Type 3 exhibited inferior accuracy of the COP calculation, but they still sufficed for applications that did not require high accuracy. Aside from the factor of the number of sensing spots used in the calculation, we also demonstrated that the location of selected sensors could influence the accuracy of COP calculation in the analyses by using the different combinations of metatarsal areas and other areas (heel, central, lateral toes, and hallux). The results of this research could be a reference for making a simplified form of pressure sensing Internet-of-Health Things (IoHT) insole with a reduced product cost.
- 8 March 2021
- Published by MDPI AG in Sustainability
- Vol. 13 (5) , 2934
Open AccessA Study on the Influence of Number/Distribution of Sensing Points of the Smart Insoles on the Center of Pressure Estimation for the Internet of Things Applications
The past decade has seen the emergence of numerous new wearable devices, including many that have been widely adopted by both physicians and consumers. In this paper, we discuss the design and application of smart insoles to measure gait and plantar pressure. Herein, we investigate the potential applications of insoles with fewer sensing spots and the consequent reduction in the amount of data acquired from both feet. The main purpose is to discuss the influence of the layout of these pressure sensing points of the insole design on the center of pressure (COP) calculation. The insole used in this study has 89 pressure sensing spots, and we used data from 36, 29, 20, and 11 sensing points in simplified calculation types. Among these four simplified calculation types, Type 1 exhibited the best accuracy of the COP calculation, and Type 4 obtained the worst results. Type 2 and Type 3 exhibited inferior accuracy of the COP calculation, but they still sufficed for applications that did not require high accuracy. Aside from the factor of the number of sensing spots used in the calculation, we also demonstrated that the location of selected sensors could influence the accuracy of COP calculation in the analyses by using the different combinations of metatarsal areas and other areas (heel, central, lateral toes, and hallux). The results of this research could be a reference for making a simplified form of pressure sensing Internet-of-Health Things (IoHT) insole with a reduced product cost.
- 5 September 2020
- Published by Springer Nature in Mobile Networks and Applications
- Vol. 27 (1) , 208-218
Monitoring Spatial Keyword Queries Based on Resident Domains of Mobile Objects in IoT Environments
- 5 September 2020
- Published by Springer Nature in Mobile Networks and Applications
- Vol. 27 (1) , 208-218
Monitoring Spatial Keyword Queries Based on Resident Domains of Mobile Objects in IoT Environments
- 26 February 2020
- Published by Springer Nature
- p. 362-371
An Adjustable-Tree Method for Processing Reverse Nearest Neighbor Moving Queries
- 26 February 2020
- Published by Springer Nature
- p. 362-371
An Adjustable-Tree Method for Processing Reverse Nearest Neighbor Moving Queries
- 19 October 2019
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 76 (4) , 2597-2621
Mining image frequent patterns based on a frequent pattern list in image databases
- 19 October 2019
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 76 (4) , 2597-2621
Mining image frequent patterns based on a frequent pattern list in image databases