Video managing apparatus, method, and program

PROBLEM TO BE SOLVED: To perform a video recognition or a video retrieval also in consideration of time information with a sufficient accuracy based on the video and a relation with additional information added to the video.SOLUTION: By a topic trajectory model learning section 1, an accumulation state variable aggregation that is an aggregation of an accumulation state variable which indicates distribution of an accumulation potential variable acquired by using an accumulation potential variable derivation function on the basis of an accumulation video section feature aggregation and an accumulation additional information feature aggregation, and a topic trajectory model which is made up of the accumulation potential variable derivation function are built. By a video recognition section 2, an input state variable aggregation or input state variable distribution is extracted based on an input potential variable aggregation that is an aggregation of an input potential variable acquired by using an input potential variable derivation function on the basis of an input video section feature aggregation and on the accumulation state variable aggregation, and an output additional information feature aggregation that is an aggregation of an output additional information feature associated with the input image section feature is estimated based on a new correction input potential variable aggregation corrected based on the input state variable aggregation or the input state variable distribution and the input video section feature aggregation.