Multimedia based Information Retrieval Approach based on ASR and OCR and Video Recommendation System

Lecture videos and e-learning are upcoming effective learning resources. Getting the appropriate lecture video out of all video archives available on the internet is not an easy task. Analyzing the video title, description and other static metadata contents is not sufficient to find the relevance of a video. This paper presents an approach for lecture video analysis based on the content of the video. We apply video segmentation to retrieve the frames from given video at specific time interval. Then we retrieve keywords from the frames using OCR technology. At the same time ASR technique extract textual metadata from audio track of the video which is easily separable from video. On the basis of retrieved keywords web links, image links and YouTube links are provided. Users can access the related videos using provided links and they can give rating for viewed video according to the relevance of that video. This rating is helpful for further attempts of finding the relevance of video. Recommendation system is the major part of this research which is implemented using cosine similarity and Pearson correlation score.