University of Houston

Lecture videos are extremely useful and a great source of learning these days. At University of Houston, these videos are widely used by students throughout the department. Since, most videos are generally very long, it is difficult for the students to directly access the required topic of interest within a lecture video. ICS (Indexed, Captioned and Searchable) videos project provides student direct access to the required topic within video lecture by providing index points that represents the topic. These index points are generated using text from the images that are extracted using OCR (Optical Character Recognition) technology. Index points are achieved by the help of indexing algorithm that determines topic change which is based on text similarity.

Here we present topic based lecture video segmentation using speech text/captions. The purpose of this thesis is to utilize the spoken text of a lecture video in order to achieve index points using underlying text based indexing algorithm. In order to achieve this goal a set of twenty five lecture videos was taken from different departments at University of Houston and Coursera website. The captions were achieved with the help of YouTube Speech Recognition System. The performances and limitations of OCR text, uncorrected/original speech text and corrected speech text based indexing were analyzed. The results indicate that slide text based indexing generally gives 4% better results than spoken text based indexing. The corrected speech text/caption provides better indexing results (11%) where OCR text fails to perform and the results closely matched the ground truth. The error analysis done on speech texts and slide texts prove that poor OCR text and poor caption quality are one of the main reasons that hamper the indexing accuracy.