摘要 |
PROBLEM TO BE SOLVED: To carry out event-based clustering of data files. SOLUTION: In the method, the temporal event clustering of a collection of data files is carried out by extracting one or more elements selected from a group of time stamp, GPS data, file size, file type, file name, text annotations, audio annotations, low level features from a meta-data tag headers or contents of each of the data files, generating one or more similarity matrices based on comparing the extracted one or more elements of at least one data file with the extracted element from the data file or one or more neighboring data files, determining one or more initial boundary partitions by applying one or more kernel functions of different extent L to the similarity matrices, using one or more algorithms to select one or more boundaries from the determined boundaries, and organizing the data files based on the selected boundaries. COPYRIGHT: (C)2006,JPO&NCIPI
|