Erwin Wu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Erwin Wu has a total of 49 co-authors. Here are the top co-authors by number of shared publications:
- Hideki Koike
School of Computing, Institute of Science Tokyo, Tokyo, Japan
28 shared publications - Chen-Chieh Liao
School of Computing, Institute of Science Tokyo, Tokyo, Japan
16 shared publications - Yichen Peng
School of Computing, Institute of Science Tokyo, Tokyo, Japan
16 shared publications - Ruofan Liu
Department of Computer Science, Institute of Science Tokyo, Tokyo, Japan and Stanford University, Stanford, USA
6 shared publications - Toshihiro Hirano
Institute of Science Tokyo, Tokyo, Japan
6 shared publications - Masaki Kuribayashi
Waseda University, Tokyo, Japan
5 shared publications
Latest Publications
- Open Access
- 16 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Dynamic Blur Guidance for Look-Ahead Gaze in VR Ski Training
- 16 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Open AccessDynamic Blur Guidance for Look-Ahead Gaze in VR Ski Training
- Open Access
- 2 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-13
SnowEcho: End-to-End Blind Skiing Assistance System Using Spatial Audio for Simulated Dynamic Slope
- 2 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-13
Open AccessSnowEcho: End-to-End Blind Skiing Assistance System Using Spatial Audio for Simulated Dynamic Slope
- 10 September 2026
- Published by Springer Nature
- p. 115-133
Motion Style Slider: Endpoint-Supervised Continuous Style Control for Human Motion Diffusion
- 10 September 2026
- Published by Springer Nature
- p. 115-133
Motion Style Slider: Endpoint-Supervised Continuous Style Control for Human Motion Diffusion
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-17
SoleCoach: Sole Pressure and IMU-based MLLMs for Skill Coaching
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-17
Open AccessSoleCoach: Sole Pressure and IMU-based MLLMs for Skill Coaching
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-20
Sensing Your Vocals: Exploring the Activity of Vocal Cord Muscles for Pitch Assessment Using Electromyography and Ultrasonography
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-20
Open AccessSensing Your Vocals: Exploring the Activity of Vocal Cord Muscles for Pitch Assessment Using Electromyography and Ultrasonography
- Open Access
- 3 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 32 (5) , 3093-3103
MuscleGolfAR: Embodied versus Detached Visualization for Motions and Inferred Muscle Activations in Augmented Reality
Precise kinematic and dynamic representations are equally critical for fine-grained motor skills such as golf, yet the latter remains underexplored. This work introduces MuscleGolfAR, an augmented reality (AR) training system integrating swing motions and muscle activities. To support cost-effective inference of muscle activations, we construct a multimodal dataset encompassing posture, electromyography (EMG), and plantar pressure. The system displays inferred EMG through embodied (first-person perspective) and detached (third-person perspective) visualizations. User studies are subsequently conducted to evaluate the impact of these strategies on training effectiveness. Results reveal that embodied visualizations enhance ownership over augmented feedback, while detached visualizations facilitate a holistic comprehension of the whole body. Furthermore, individuals' preferences for these strategies correlate with their practice habits and skill proficiencies, offering new insights for the design of AR-based motor skill training systems.
- 3 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 32 (5) , 3093-3103
Open AccessMuscleGolfAR: Embodied versus Detached Visualization for Motions and Inferred Muscle Activations in Augmented Reality
Precise kinematic and dynamic representations are equally critical for fine-grained motor skills such as golf, yet the latter remains underexplored. This work introduces MuscleGolfAR, an augmented reality (AR) training system integrating swing motions and muscle activities. To support cost-effective inference of muscle activations, we construct a multimodal dataset encompassing posture, electromyography (EMG), and plantar pressure. The system displays inferred EMG through embodied (first-person perspective) and detached (third-person perspective) visualizations. User studies are subsequently conducted to evaluate the impact of these strategies on training effectiveness. Results reveal that embodied visualizations enhance ownership over augmented feedback, while detached visualizations facilitate a holistic comprehension of the whole body. Furthermore, individuals' preferences for these strategies correlate with their practice habits and skill proficiencies, offering new insights for the design of AR-based motor skill training systems.
- Open Access
- 16 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 691-697
Seeing and Saying Movement: Combining Skeleton Visualization and LLM Feedback for Motor Skill Coaching
We present a near real-time kickboxing coaching system that integrates (i) side-by-side skeleton visualization of an expert and a learner, (ii) per-joint directional difference highlighting, and (iii) a hybrid comment module combining rule-based messages with large language model (LLM) output. Rather than claiming robust learning effects, this short paper frames the contribution as a design exploration of how tightly coupled visual and language-based feedback can support novice interpretation of movement errors. In a small exploratory user study (n=6; brief practice per condition), we compare three feedback modalities—avatar-only (AO), rule-based feedback (RB), and hybrid LLM feedback (HL)—and report descriptive joint-level patterns and subjective user perceptions. Although objective improvements were limited at this scale, the results suggest a movement-dependent fit between modalities (global visualization for punches; localized comments for kicks) and a consistent preference for HL. We discuss design trade-offs and limitations, including temporal alignment choices and the assumption of a single expert reference.
- 16 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 691-697
Open AccessSeeing and Saying Movement: Combining Skeleton Visualization and LLM Feedback for Motor Skill Coaching
We present a near real-time kickboxing coaching system that integrates (i) side-by-side skeleton visualization of an expert and a learner, (ii) per-joint directional difference highlighting, and (iii) a hybrid comment module combining rule-based messages with large language model (LLM) output. Rather than claiming robust learning effects, this short paper frames the contribution as a design exploration of how tightly coupled visual and language-based feedback can support novice interpretation of movement errors. In a small exploratory user study (n=6; brief practice per condition), we compare three feedback modalities—avatar-only (AO), rule-based feedback (RB), and hybrid LLM feedback (HL)—and report descriptive joint-level patterns and subjective user perceptions. Although objective improvements were limited at this scale, the results suggest a movement-dependent fit between modalities (global visualization for punches; localized comments for kicks) and a consistent preference for HL. We discuss design trade-offs and limitations, including temporal alignment choices and the assumption of a single expert reference.
- Open Access
- 16 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 672-682
Real-time Ski Coaching using Pose Estimation from Sole Pressure and MLLMs Based on Reference-Pose Errors
We present a real-time skiing coaching system that uses insole pressure sensors and MLLMs to provide actionable feedback. Coaching facilitates skill acquisition by offering expert insights and actionable guidance. However, existing MLLM-based approaches often emphasize content generation while neglecting essential implementation factors, such as the timing and granularity of feedback. To address this, our system analyzes errors between an ideal posture and users’ postures to determine appropriate feedback timing and to select targeted feedback points. By relying exclusively on foot pressure data, our system ensures portability and is well-suited for skiing, a sport characterized by large-scale displacement. A user study comparing different feedback timings and content shows that real-time coaching helps users more accurately identify their errors and improves their awareness of left–right balance. These findings highlight the importance of timing and focus when designing real-time coaching systems.
- 16 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 672-682
Open AccessReal-time Ski Coaching using Pose Estimation from Sole Pressure and MLLMs Based on Reference-Pose Errors
We present a real-time skiing coaching system that uses insole pressure sensors and MLLMs to provide actionable feedback. Coaching facilitates skill acquisition by offering expert insights and actionable guidance. However, existing MLLM-based approaches often emphasize content generation while neglecting essential implementation factors, such as the timing and granularity of feedback. To address this, our system analyzes errors between an ideal posture and users’ postures to determine appropriate feedback timing and to select targeted feedback points. By relying exclusively on foot pressure data, our system ensures portability and is well-suited for skiing, a sport characterized by large-scale displacement. A user study comparing different feedback timings and content shows that real-time coaching helps users more accurately identify their errors and improves their awareness of left–right balance. These findings highlight the importance of timing and focus when designing real-time coaching systems.
- Open Access
- 14 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Spatial Audio for Ski Simulation with Visually Impaired Users in Dynamic Environments
- 14 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Open AccessSpatial Audio for Ski Simulation with Visually Impaired Users in Dynamic Environments
- Open Access
- 14 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-3
A Demonstration of ShiftingGolf: Gross Motor Skill Correction Using Redirection in VR
- 14 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-3
Open AccessA Demonstration of ShiftingGolf: Gross Motor Skill Correction Using Redirection in VR