家庭用ゲーム機の入力デバイスを用いた階層型ニューラルネットワークによるジェスチャ認識  [in Japanese] Gesture Recognition Based on Multilayer Neural Network by Using Input Device of Home Gaming Console  [in Japanese]

Access this Article

Search this Article

Abstract

ノンバーバルコミュニケーションであるジェスチャにおいて,人間の意思や感情は,手などの位置よりもむしろ身体に加えられた力に顕著に現れると考えられる.本論文では,運動中に働く力は加速度によって検出できることから,三軸加速度センサを用いた階層型ニューラルネットワークによるジェスチャ認識を試みた.家庭用ゲーム機Wiiに付属のWiiリモコンを入力デバイスとしたジェスチャ認識実験の結果,階層型ニューラルネットワークの入力値に単なる正規化された加速度の順列を用いることで,特定演者だけでなく不特定演者に対しても高い認識率を示し,汎用性のあるジェスチャ認識エンジンを構築できることが確認できた.We consider that the human intention and feelings conspicuously appear at the force added to a body rather than the positions of hands in the gesture that is nonverbal communication. In this paper, we tried gesture recognition based on multilayer neural network by using acceleration sensor because the force to act during motion can be detected by acceleration. As a result of our gesture recognition experiment in which the Wii remote attached to the home gaming console Wii was used as an input device, we have confirmed the following. By using the permutation of the simple normalized acceleration for input values of multilayer neural network, we have obtained high recognition rate in the case of not only specific performers but also unspecified performers. Therefore, we could develop versatile gesture recognition engine.

We consider that the human intention and feelings conspicuously appear at the force added to a body rather than the positions of hands in the gesture that is nonverbal communication. In this paper, we tried gesture recognition based on multilayer neural network by using acceleration sensor because the force to act during motion can be detected by acceleration. As a result of our gesture recognition experiment in which the Wii remote attached to the home gaming console Wii was used as an input device, we have confirmed the following. By using the permutation of the simple normalized acceleration for input values of multilayer neural network, we have obtained high recognition rate in the case of not only specific performers but also unspecified performers. Therefore, we could develop versatile gesture recognition engine.

Journal

  • 情報処理学会論文誌

    情報処理学会論文誌 51(1), 199-203, 2010-01-15

    情報処理学会

Cited by:  1

Keywords

Codes

  • NII Article ID (NAID)
    110007970628
  • NII NACSIS-CAT ID (NCID)
    AN00116647
  • Text Lang
    JPN
  • Article Type
    Journal Article
  • ISSN
    1882-7764
  • NDL Article ID
    024198909
  • NDL Call No.
    YH247-743
  • Data Source
    CJPref  NDL  NII-ELS  IPSJ 
Page Top