Computing Distances Between Japanese Kanji from Stroke Order and Pen Movements
Year: 2026
Type: article
Abstract: KanjiVG provides vector drawings of Japanese kanji in which individual strokes are stored separately and in their prescribed order. This information makes it possible to compare characters not only as static images, but also as sequences of drawing movements. This article introduces a straightforward way to calculate distances between kanji using their SVG paths. Each stroke is represented by a small set of points, and the movements between successive strokes are recorded. A sequence-alignment procedure then compares the resulting representations, allowing characters with different stroke counts to be compared. The method is illustrated with a short JavaScript implementation applied to six characters from a fixed KanjiVG release. In this small set, each character's nearest neighbor is its intended visually similar partner (未/末, 木/本, 日/目), and the optimal alignments identify the strokes that distinguish the members of each pair. The procedure is intended as an accessible starting point for exploring stroke-based character similarity, rather than as a validated model of human visual perception.
Author: Mantelli, Alessandro
Institution: Ca' Foscari University of Venice
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7225175604