摘要 |
PROBLEM TO BE SOLVED: To accurately identify image content of video data. SOLUTION: A video data identification apparatus 70 includes a time-series identification section 11 using HMM, a non time-series feature amount extracting sections 71-1 to 71-M for extracting non time-series feature amount from video data be to processed, and a pattern identification section 72 using NN. An input pattern identification section 72 outputs a scene, as an identification result, corresponding to an (N+M)-dimensional input pattern based on N likelihood values input from the time-series identification section 11 and M non time-series feature amounts input from the non time-series feature amount extracting sections 71-1 to 71-M. This apparatus is applicable to an apparatus for classifying time-series data. COPYRIGHT: (C)2009,JPO&INPIT
|