DETECT AND RECOGNIZE THE POSSIBLE SPOKESPERSON WITH CAMERA, IN A TOBOT-HUMAN COMMUNICATION SYSTEM

カテゴリ: 部門大会

論文No: GS11-3

グループ名: 【C】平成21年電気学会電子・情報・システム部門大会講演論文集

発行日: 2009/09/03

タイトル(英語): DETECT AND RECOGNIZE THE POSSIBLE SPOKESPERSON WITH CAMERA, IN A TOBOT-HUMAN COMMUNICATION SYSTEM

著者名: 馬嘉 (徳島大学),鈴木 基之(徳島大学),任福継 (徳島大学)

著者名(英語): jia Ma(The University of Tokushima),motoyuki Suzuki(The University of Tokushima),fuji Ren(The University of Tokushima)

キーワード: 動態追跡|顔認識|顔検出|人体検出|人間-機械通信システム相関関係フィルター|Motion analysis|Face recognition|Face detection|Human body detection|Man-machine communication systemCorrelation filter

要約(日本語): In order to provide a better man-machine communication system, we hope the computer or the robot can recognize the speaker, who is speaking to the robot. However, in the realistic scene, the factors of the complex environment, such as the noise, may course some difficulties for the robot to recognize the commands. So imitating the performance of human, we try to use the eyes of the robot to solve the problem. In this research, we propose a new spokesperson detection method and found a new system for recognizing the speaker and his expression, based on image processing. The robot might find and identified all the possible speakers around it, by detecting the human body and its movement, with a omni-directional camera. It might also find out the spokesperson who is speaking, by judging whether the possible spokesperson`s mouth are acting, with two eye cameras. We can also recognize that who is the spokesperson by comparing the face video from the eye cameras with the video data in our face expression video database. Using eye cameras and a omni-directional camera, we developed the system and some experiments will be conducted to verify this new system.

PDFファイルサイズ: 3,866 Kバイト