Domain Adaptation Improves Digital Pen Handwriting for Children.

Digital pens with kinematic sensors allow users to write on any surface, capturing handwriting digitally and enhancing human-computer interaction. This technology holds considerable promise as an educational aid, particularly for teaching writing in classrooms. A significant challenge arises from the distinct sensor signals produced by adults versus children, even for similar handwriting traces, due to differences in writing speed and confidence. To overcome this, the research explores a domain adaptation approach. The goal is to construct a unified intermediate feature representation that can effectively bridge the gap between adult and child handwriting signals, thereby facilitating more accurate trajectory reconstruction for both groups. The study compares this domain adaptation method against training models from scratch and fine-tuning existing models. The results highlight the benefits of domain adaptation in leveraging existing knowledge to apply the technology effectively across different user contexts, particularly for children learning to write.