Feature Extraction and State Decomposition for Image-based Japanese Sign Language Words Recognition Using HMM

DOI Web Site
  • Matsuo Tadashi
    Research Organization of Science and Engineering, Ritsumeikan University
  • Yamada Yutaka
    Graduate School of Science and Engineering, Ritsumeikan University
  • Shirai Yoshiaki
    School of Information Science and Engineering, Ritsumeikan University
  • Shimada Nobutaka
    School of Information Science and Engineering, Ritsumeikan University

Bibliographic Information

Other Title
  • HMMを利用した画像処理による手話単語の認識のための特徴抽出および状態分割
  • HMM オ リヨウ シタ ガゾウ ショリ ニ ヨル シュワ タンゴ ノ ニンシキ ノ タメ ノ トクチョウ チュウシュツ オヨビ ジョウタイ ブンカツ

Search this article

Abstract

<p>This paper proposes a method of extracting efficient features for recognizing Japanese sign language words and recognizing them with Hidden Markov Model. Hand and face regions are extracted and tracked by image processing technique. Features of hand motion and shape are calculated by the extracted regions. Each HMM model is trained by Baum-Welch algorithm. The number of states is determined based on segmentation of the hand motion. Viterbi algorithm is employed for recognition. Experimental results show the validity of the proposed method. </p>

Journal

Related Projects

See more

Keywords

Details 詳細情報について

Report a problem

Back to top