For Full-Text PDF, please login, if you are a member of IEICE,|
or go to Pay Per View on menu list, if you are a nonmember of IEICE.
Lecture Context Recognition Based on Statistical Feature of Lecturer Action for Automatic Video Recording
Takafumi MARUTANI Yoshitaka SUGIMOTO Koh KAKUSHO Michihiko MINOH
D - Abstracts of IEICE TRANSACTIONS on Information and Systems (Japanese Edition)
Publication Date: 2007/10/01
Online ISSN: 1881-0225
Print ISSN: 1880-4535
Type of Manuscript: PAPER
automatic shooting video, recognizing lecture situation, statistical properties, Hidden Markov Model,
Full Text(in Japanese): PDF(732.2KB)
>>Buy this Article
In this paper, we discuss the problem of recognizing situations of lectures for shooting lecture videos automatically. We classify the situations of the lectures based on the subjects to be gazed at for understanding the lectures in a lecture room. However it is not easy to recognize the lecture situations by conventional approaches based on the features of the sensory data obtained in a lecture room because the same sensory data occurs in the different lecture situations. We employ statistical properties for the occurrence of sensory data in each lecture situation together with the transition probability between lecture situations. This process is realized by Hidden Markov Model that represents those statistical properties.