手書き曲線同定法FSCIにおける幾何曲線列認識性能の改善

IPSJ

Bibliographic Information

Other Title
  • Performance Improvement of Geometric Curve Sequence Recognition in the Freehand Curve Identifier FSCI

Search this article

Abstract

我々はすでに，一筆書きの手書き曲線をひと連なりの幾何曲線列として認識する手書き曲線同定法FSCIを提案している．FSCIは，まず「ファジィセグメンテーション法」により描画の停止箇所を検出することで一筆の手書き曲線をいくつかの同定単位に分割し，その後の同定処理においてそれぞれの同定単位を幾何曲線として認識する．しかし，このFSCI前段のファジィセグメンテーション法が後段の同定処理の認識性能に悪影響を与える場合がある．本論文ではこの問題を明らかにするとともに，その解決策として，描画における「停止状態」と「移動状態」に基づいて手書き曲線から同定単位を抽出する新たな手法「ファジィフラグメンテーション法」を提案する．また，評価実験を通して提案手法がFSCIの幾何曲線列認識性能の向上に寄与することを示す．

We have already proposed a freehand curve identifier, FSCI, which identifies a stroke of freehand curve as a sequence of geometric primitive curves. In FSCI, an initial process named “Fuzzy Segmentation” divides a stroke of freehand curve into several identification units based on standstill detection in drawing, so that each of units is identified as a geometric primitive curve by its following process. However, it is revealed that Fuzzy Segmentation occasionally degrades the performance of the following identification process. In this paper, we clarify the problem and, as a solution, we propose a new method named “Fuzzy Fragmentation,” which extracts identification units from a freehand curve based on “Stay States” and “Move States.” Moreover, through experimental results, we show that the proposed method improves the performance of FSCI on geometric curve sequence identification.

Journal

Related Projects

See more

Keywords

Details 詳細情報について

Report a problem

Back to top