Augmenting the human arm surface via projection mapping can have a great impact on our daily lives with regards to entertainment, human-computer interaction, and education. However, conventional methods ignore skin deformation and have a high latency from motion to projection, which degrades the user experience. In this paper, we propose a projection mapping system that can solve such problems. First, we newly combine a state-of-the-art parametric deformable surface model with an efficient regression-based accuracy compensation method of skin deformation. The compensation method modifies the texture coordinate to achieve high-speed and highly accurate image generation for projection using joint-tracking results. Second, we develop a high-speed system that reduces latency from motion to projection within 10 ms. Compared to the conventional methods, this system provides more realistic experiences.