Video Semantic Keyword Extraction Based on Multi-Channel Information

Digital videos have became important multimedia information carrier. However, videos have features of large number, wide variety, lengthy content and complex structure, which make video retrieval, video query and video classification difficult. Video keyword extraction can not only quickly grasp the subject content of video, but also use key words to classify video and establish video database index. Due to the unstructured and multi-channel feature of video, it is not enough to use only the keyword extraction method for text. This paper proposes a semantic based algorithm to extract keywords from video from multi-channel information, which combines the information of visual channel and audio channel, and extracts keywords from video based on semantics. Experiment results show that the algorithm proposed in this paper is very effective for both video and plain text.