<正> This paper addresses the problem of recognizing the elements of video semantics by investigating the presence and temporal behavior of low-level features in video sequences. Examples of these features are color, texture, shape, frequency components, audio and speech characteristics. The extracted elements of video semantics can then be used to organize the content of a video database in such a way that the overall large search space and consequently the total interaction (browsing, query) time between the user and the database is as much reduced as possible. We concentrate in this paper on the elements of video semantics that can be extracted by working only with visual low-level features of a video. For this purpose we outline and discuss the methodologies for analyzing the content of a movie and broadcast news programs, that were developed in the course of the EU-ACTS research project SMASH (Storage for Multimedia Application Systems in the Home).