Touching Movement: 3D Tactile Poses for Supporting Blind People in Learning Body Movements
Kengo Tanaka, Xiyue Wang, Hironobu Takagi, Yoichi Ochiai, Chieko Asakawa · 2026 · Proceedings of the Twentieth International Conference on Tangible, Embedded, and Embodied Interaction (TEI '26) · doi:10.1145/3731459.3773325
Summary
Tanaka and colleagues investigate whether 3D-printed tactile models of whole-body poses can help blind learners acquire physical movements that are normally taught through visual demonstration. Verbal descriptions are notoriously inconsistent and struggle to convey nuance, while physical guidance by instructors raises concerns about personal space and consent, and 2D tactile graphics compress depth and limb relationships in ways that can be hard to reconstruct mentally. Refreshable tactile displays can animate motion but lack the resolution and screen size to represent simultaneous multi-limb movement. The authors build a semi-automated pipeline that estimates 3D body posture from a single RGB image using the Humans in 4D tool, refines the mesh in Blender, and prints the result with a Formlabs Form 3L stereolithography printer. They run a three-session participatory design study with a 66-year-old blind designer and researcher (P0) to refine model scale (settling on a 10.5 cm standing figure), base design, and tactile reference markers that indicate foot position and direction of movement across a sequence. They then run two controlled user studies with ten blind participants: Study 1 compares 3D models, 2D swell-paper tactile graphics, and step-by-step audio instruction on six yoga poses at two difficulty levels; Study 2 compares 3D models and 2D tactile graphics on two four-pose calisthenic sequences inspired by Japan's Radio Calisthenics programme. Objective measures include completion time, reproduction accuracy scored against an eight-point joint-position rubric, and the number of clarification questions asked. Subjective ratings are collected on a seven-point Likert scale across enjoyment, easiness, effectiveness, time pressure, motivation, and overall preference.
Key findings
For static yoga poses, 3D models were significantly faster than 2D tactile graphics (Level 1 48 s vs 155 s, Level 2 72 s vs 135 s) and achieved the highest reproduction scores at both difficulty levels. Participants using 3D models asked an average of 0.5 to 0.9 clarification questions, compared with 1.8 to 2.1 for audio and 3.7 to 4.1 for 2D tactile graphics; 80% of participants needed no clarification at all with 3D models. For dynamic calisthenic sequences, 3D models were again significantly faster (266 s vs 427 s), and 60% of participants needed no clarification versus 0% for 2D tactile graphics. Subjective ratings favoured 3D models on all six Likert items for calisthenics and on easiness, effectiveness, time pressure, motivation and overall preference for yoga (all p < .0167 after Bonferroni correction). Qualitative feedback emphasised that tactile confirmation of body shape gave participants a sense of capability ('It made me feel like I could do it'), that free rotation of the model helped align with their own body, and that reference markers on the base reliably communicated foot movement direction. Reported limitations included difficulty inferring transitional poses between printed keyframes and the memory load of longer sequences; participants suggested pairing 3D models with optional step-by-step audio or real-time corrective feedback.
Relevance
For accessibility practitioners this paper is a concrete case study in how 3D printing has moved from novelty to a credible alternative format for kinesthetic and spatial content, particularly for blind learners whom 2D tactile graphics and audio descriptions under-serve. The semi-automated pipeline (pose estimation from a single image to printable mesh) is significant because it makes small-batch, pose-specific tactile materials feasible outside specialist braille production workflows. The design details are immediately reusable: a standing-figure height around 10.5 cm, a rigid floor base, and a small reference marker for foot position repeatedly surfaced as the factors that made models interpretable. Limitations include the small sample (ten participants, one co-designer), the focus on a Japanese cultural context (Radio Calisthenics), and the absence of longitudinal retention data. The paper fits naturally alongside adaptive PE, yoga, and dance research for blind and low-vision learners, and suggests that multimodal combinations — tactile model plus step-by-step audio plus pose-tracking feedback — are the most promising direction for independent home practice.
Tags: tactile graphics · 3D printing · blindness and low vision · visual impairment · physical activity · accessible education · pose estimation · participatory design · multimodal interaction · embodied interaction