Audio thumbnail generation of Irish traditional music

Abstract

An approach is presented which generates an audio thumbnail of Irish Traditional music. An audio thumbnail is considered to be the most representative segment of the music. For popular music, the chorus is considered to be an ideal audio thumbnail, however in Irish Traditional music there is no chorus. An Irish Traditional tune consists of two or more short structural segments called parts. Parts are repeated to extend the tune, and the tune itself is also repeated once or more in its entirety. To further extend a performance, tunes are concatenated to form a set of tunes. As a result, there is plenty of repetition within this music type. The presented approach utilises an existing approach which calculates the structure of Irish Traditional Music. The structural information is used to extract a single rendition of each distinctive part. The resulting parts are concatenated to form the audio thumbnail.

Music Thumbnailing for Radio Podcasts: A Listener Evaluation

“…This use case presents an interesting challenge in deciding which section (or sections) of a piece of music should be included in a 30-second clip. Previous work on music thumbnailing typically focuses on either finding a continuous representative section of music [1][2][3][4][5][6][7] or by concatenating multiple song segments together [8][9][10][11]. However, in radio shows it is very common for the presenter to talk over the start and end of a song.…”

“…Kelly et al [11] used chroma based similarity matrices but also applied beat-synchronous segment boundaries to compile multi-section thumbnails from Irish folk songs. This task is somewhat similar to ours, where sections of a song need to be edited together into a summary.…”

Automatic Music Summarization. A "Thumbnail" Approach

“…No prior semantic information about the musical piece is needed (a pitch, a rythm, a key, etc.). Such a summary is often called "music thumbnail" (Bartsch, Wakefield, 2001;Kelly et al, 2010). The algorithm consists of the extraction of parameters characterizing spectrum of a mu-sical piece, choosing the similarity measure that will be used for calculating the similarity of the musical signal frames, constructing the self-similarity matrix and applying the aggregated similarity measure to find a starting point of a thumbnail.…”