Semantic Enriched Lecture Video Retrieval System Using Feature Mixture and Hybrid Classification

Abstract

The advancement in the web technologies has increased the lecture video contents tremendously. The lecture video retrieval for the e-learning process is a challenging task since the videos are unstructured and have a large size. Since many video lectures have less information, the video retrieval system needs to be built with the enhanced features to improve the efficiency of the retrieval process. In this paper, a semantic enriched lecture video retrieval system has been proposed. The key frames from the video are extracted through the pre-processing. The proposed model uses the feature mixture database with the more relevant features such as text, semantic word, and the Local Gabor Pattern (LGP) vectors. The video retrieval from the feature mixture database is done by using the hybrid K-Nearest Neighbour Naive Bayes (KNB) classifier. This classifier uses the techniques of both the Naive Bayes (NB) classifier and the K-Nearest Neighbour (K-NN) classifier. The performance metrics such as precision, recall and the f-measure analyze the efficiency of the proposed model. Simulation is done by giving the text query and the video query to the video database. The simulation results show that the proposed model has better precision and recall value of 1.0 and 0.7500 respectively. The f-measure of the proposed model has a better value of 0.8519 than the existing K-NN system.