FRAMEWORK FOR MULTIMEDIA LECTURE SYSTEM USING REAL-TIME VIDEO OBJECT EVENT DETECTION AND SCRIPTING

摘要

We propose a novel framework for a multimedia lecture system. In this framework, we have applied several video processing techniques to a real-time lecture video stream. The video data is then processed, indexed and combined with other multimedia based on the type of event occurred which is governed by set of predefined scripts. We describe in detail the techniques applied in order to achieve this framework, including object tracking using mean shift and object recognition using PCA-SIFT as well as performance enhancement using SIMD instructions. We illustrate our preliminary results from the prototype which demonstrates that while this system provides sophisticated video processing, it can be achievable using inexpensive hardware.