1999 Fiscal Year Final Research Report Summary
An Automatic Lecture Recording System with Recognizing the Situation of the Lecture
| Project/Area Number |
09558030
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 展開研究 |
|---|
| Research Field |
計算機科学
|
|---|
| Research Institution | KYOTO UNIVERSITY |
|---|
Principal Investigator |
IKEDA Katsuo Graduate School of Informatics, Kyoto University. Professor., 情報学研究科, 教授 (30026009)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
YAGI Keisuke Graduate School of Informatics, Kyoto University. Associate Researcher, 総合情報メディアセンター, 助手 (60293940)
AMANO Akira Faculty of Information Sciences, Hiroshima City University. Associate Professor, 情報科学部, 助教授 (60252491)
KAWAHARA Tatsuya Graduate School of Informatics, Kyoto University. Associate Professor, 情報学研究科, 助教授 (00234104)
FUJIKAWA Kenji Graduate School of Informatics, Kyoto University. Associate Researcher, 情報学研究科, 助手 (40293899)
MUKUNOKI Masayuki Graduate School of Informatics, Kyoto University. Associate Researcher, 情報学研究科, 助手 (20283640)
|
|---|
| Project Period (FY) |
1997 – 1999
|
| Keywords | Automatic Lecture Recording / Lecture Video Database / Video Data Processing / Real-time Image Processing / Integration of Multiple Information / Division of Lecture Video / Lecturer Tracking / Recognition of Lecture Situation |
|---|
| Research Abstract |
In order to broadcast or record normal lectures held in universities, we have developed an automatic lecture recording system. The achievement of our research is as follows :
(a) The autonomous robot camera. Once a sample image is given to the robot camera, it moves in a lecture room and takes pictures which are similar to the sample image. The robot acts based on the algorithm of "Hypothesis and Verification" loop. The implementation of the robot is not yet complete but the restricted experiments we have conducted reveal that the algorithm is robust to the change of lighting condition or appearance of obstacles.
(b) The automatic video stream switcher. It can select a single video stream to broadcast to remote students from multiple video streams. We built models of short span scenario in lectures as a sequence of lecture status. The lecture status was expressed by a set of several features extracted from video or audio streams. We also made relation between the scenario and the video stream to be selected. As the result, we could automatically detect the lecture scenario at the time and select the appropriate video stream with our method. We implemented the system with image processing board and realized the real-time selection of video stream.
(c) The semi-automatic lecture video database generator. Using pointer information the lecturer pointed out, the keywords spoken by the lecturer and the contents of the side the lecturer used, the generator adds detailed indexes to lecture video streams. In order to extract the pointer information, we modified an online presentation tool. We also used a natural language processing tool and a speech recognition tool. As the result, the extracted information is not reliable. We used the information complementarily and succeeded to add reliable indexes to video streams.
|
