Adaption of String Matching Algorithms for Identification of Near-Duplicate Music Documents

The number of copyright registrations for music documents is increasing each year. Computer-based systems may help to detect near-duplicate music documents and plagiarisms. The main part of the existing systems for the comparison of symbolic music are based on string matching algorithms and represent music as sequences of notes. Nevertheless, adaptation to the musical context raises specific problems and a direct adaptation does not lead to an accurate detection algorithm: indeed, very different sequences can represent very similar musical pieces. We are developing an improved system which mainly considers melody but takes also into account elements of music theory in order to detect musically important differences between sequences. In this paper, we present the improvements proposed by our system in the context of the near-duplicate music document detection. Several experiments with famous music copyright infringement cases are proposed. In both monophonic and polyphonic context, the system allows the detection of plagiarisms.