| 摘要 |
A method and a system for recommending a music according to a user model, and a method for updating conditionally the user model are provided to learn a preference model from a device usage pattern of a user and to make a user preference reflect a change of a user's favorite music inclination. A system for recommending a music according to a user model includes a condition detector(300), a user model unit(302), a user model update unit(304), a user model selection unit(306), a music search unit(308), an event detector(310), a monitoring unit(312), and a music database(314). The condition detector detects external conditions for making a user preference model reflect conditions like time, weather, holiday, AM, PM and others and provides the external conditions for the user model selection unit and the user model update unit. The user model unit models kinds of musics which the user likes or not, and stores them. The user model update unit updates the user model from musics played at the music play device, or receives information on a usage pattern of the music play device from the monitoring unit and the event detector and make the received information reflect the user model. The user model selection unit selects a set of property vectors on a preferred or non-preferred music from plural user models stored at the user model unit. The music search unit receives the set of property vectors from the user model selector, calculates user preference, provides the music for the user in an ascending or descending preference value order according as the music is preferred or not. The event detector provides an event like music play time and repeated play of a specific music, occurring at the music play device, for the user model update unit. The monitoring unit monitors the usage pattern of the music play device like an MP3 player or a PC. The music database stores plural music files and can be located in or out of the music recommending system. |
