Multi-modal Instructional Video Semantic Analysis And Implementation

As a kind of multimedia information carrier that integrates model data including image, audio and text, video is widely used in all areas of social life. Teaching videos are playing an important role in modern teaching activities by combining with the internet platform as a kind of important field videos. In recent years, content-based video semantic analysis has become a hot topic; semantic analysis on teaching videos is of important practical significance for enhancing the efficiency of retrieving and browsing teaching content, as well as improving users’ learning experience.By studying on key technology of video semantic analysis, and combining with specific characteristics of teaching videos, the paper presents a kind of multi-modal teaching video semantic analysis system framework based on image modal and text modal. First, a kind of video shot segmentation method based on cumulative histogram is proposed to divide teaching video into shot set, after which key frames are extracted. Second, face detection method is used to divide key frames into teacher-frame and courseware-frame, and a kind of title text extraction method is adopted to obtain title information from courseware-frame. Last, a kind of shot merge algorithm based on the above findings is given to obtain semantic teaching unit.The paper finally designs and implements a teaching video on demand system based on B/S architecture, on which users are able to retrieval and browse teaching videos using title keywords of semantic teaching unit. Experimental results show that the proposed method has a satisfactory effect and is of high use value.