Event clustering of digital media in personal mobile device based on spatio-temporal clustering

In this paper, we intended to seek a method to retrieval media files on personal mobile devices efficiently. We aimed at finding media-generated event accurately to provide support for annotation and retrieval of multimedia files. In this paper, the event-cluster positioning is realized by spatio-temporal clustering. Time clustering uses a method that maps the time difference to image, and then finds the corresponding boundary points through image processing. We improved the original time clustering algorithm by considering the semantic of “event”, and amended potential errors in time clustering by the adaptive space clustering algorithm. At last, we proposed a new method to find a more abstract activities topic by merging events of certain conditions according to user´s usage habit.