Using Hidden Markov Model for Web-based Synchronized Multimedia Lecture Structure Analysis and Summarization

• Main Authors: Yi Hsin-Pond, 易昕鵬
• Other Authors: Chen Herng-Yow
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/59894548164684177545

碩士 === 國立暨南國際大學 === 資訊工程學系 === 91 === With the advance of multimedia and Internet technologies, how to retrieve the data effectively in hidden and complex data structure is becoming an important problem now. Tools for structure analysis and summarization are needed. In recent years, several technologies have been developed for structure analysis and summarization. Most of these technologies make stochastic process of lecture structure by using the audio/video materials. In this thesis, we present a new approach to analyzing the lecture structure based on Hidden Markov Model (HMM) in Web-Based Synchronized Multimedia Lecture (WSML) system. The HMM is a statistical model which analyze the structure information by utilizing the temporal behaviors. We use the navigation events which are one of the special features in the WSML system to achieve our approach. We pre-defined the three types of multimedia lecture: Recitation type, explanation type and unknown type of the above. The lecture structure can be found with this new approach and user can retrieve and browse the data more effectively.