Music Retrieval Using Onomatopoeic Query

We propose a music retrieval method that uses onomatopoeia query to enable easy and intuitive retrieval of music even for users without musical knowledge. When we use a music retrieval system, we usually need some information on the music, such as artist names or lyrics. However, if we do not know such information, it is difficult to retrieve the music, especially instrumental music. In the proposed method, we create an index of changes in note length and pitch in all of the music to be retrieved in advance. When searching, the proposed system extracts the change in note length and pitch from the user's query. The extracted information is used to calculate the similarity of a musical piece and the query by using a dynamic programming matching algorithm. The proposed system ranks a musical piece on the basis of similarity. The music format of the retrieval target is MIDI. In this paper, the result of retrieval experiments with the proposed method is shown.