MAMI

Since more and more music consumers have larger and larger music collections, and since buying and downloading (legally) of music from the internet will become more and more the standard way of acquiring music, it is important to develop new search methods for making relevant selections (e.g. Playlists) of a large collection. Obviously, one can store title of the song, name of the composer, etc. but based on advanced signal processing techniques one is presently investigating the development of more natural and flexible search methods. One such technique is query by imitation. In such a method the user imitates a theme from the song by humming, singing or playing it on an instrument. In order to accommodate such a search method, one needs to develop new algorithms that can find the melody sung by the user. Moreover, if one does not want to invest money in collecting score representations (e.g. MIDI scores) of all the songs in the database one wants to query, one also needs methods for retrieving melody lines and rhythmic patterns from the polyphonic audio of the produced music (the song performances). Other methods for selecting music is by specifying a musical style or genre (e.g. pop, rock, new age, rap, etc.) or by specifying that you want to hear something that sounds like a particular song you know. In both cases one needs powerful signal processing methods to extract musical properties from the polyphonic audio before and to store this information in a compact way as meta data in the music database. The MAMI project is exactly trying to develop such techniques.

Partners: 

IPEM