Browse Prior Art Database

Automatic Clustering of Media Files based on User Usage Patterns

IP.com Disclosure Number: IPCOM000199041D
Publication Date: 2010-Aug-24
Document File: 3 page(s) / 25K

Publishing Venue

The IP.com Prior Art Database

Abstract

This publication describes an automatic method to group or cluster media files based on a user usage pattern which has been collected over a period of time, e.g. in order to create media file playlists (like MP3 playlists) automatically.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 42% of the total text.

Page 1 of 3

Automatic Clustering of Media Files based on User Usage Patterns

This publication describes an automatic method to cluster media files based on a user usage pattern which has been collected over a period of time. As an example one can think of music (MP3) files played on a MP3 player like an iPod. The herein proposed method clusters all media files based on the user usage pattern using data which is already available in all standard MP3 players (like the iPod). The created clustering can then later be used to categorize the media files, e.g. in different playlists automatically, i.e. without user interaction.

There are several solutions to this problem available already, although all of them have several drawbacks compared to the proposed solution. The following list contains the most interesting patents in this area:

US Patent: 7,381,883 "System and methods for providing automatic classification of media entities according to tempo": This algorithm uses a totally different approach and classifies the media files based on the tempo (beats per minute). The drawbacks of this solution are that the user most of the time does not want to classify his media files based on the tempo only but rather on his usual usage patterns and that this approach is applicable for music files only. The herein proposed solution does not take the tempo into account but rather the usage pattern of the user (the media files which have been used, i.e. heard, by the user in a certain time frame).

US Patent: 7,227,073 "Playlist managing apparatus and method": This method needs changes inside the media or MP3 player as well, since information on the user feedback has to be stored inside the media player. The herein proposed solution works already today with every standard MP3 or media player.

US Patent: 7,196,258 "Auto playlist generation with multiple seed songs": This method creates play lists based on a list of seed songs the user has to provide. The herein proposed solution works without seed songs or any interaction with the user.

The herein proposed solution can be differentiated into two phases, a learning phase in which the user of the media file player simply uses the player and provides the information stored inside the player (how often a certain file was played) as input for the learning algorithm and a classification or play list generation phase, in which the classification or play lists are generated using the so far available usage patterns.

During the learning phase the herein proposed method works like the following:

The user uses the media player (e.g. MP3 player). The media player stores a

1.

counter for each media file counting how often a single media file has been played already (all currently available media players already do this).

From time to time (e.g. once a day) the user synchronizes his media player with

2.

his media library (e.g. iTunes) in his personal computer. After this synchronization, a program is started which calculates...