A Methodology for the Segmentation and Identification of Music Works

The identification of unknown recordings is a challenging problem that has several applications. In this paper, we focus on the identification of alternative releases of a given music work. To this end, a statistical model of the possible performances of a given score is built from the recording of a single performance. The methodology is based on the automatic segmentation of audio recordings, exploiting a technique that has been proposed for text segmentation. The segmentation is followed by the automatic extraction of a set of relevant audio features from each segment. Identification is then carried out using an application of hidden Markov models. The approach has been tested with a collection of orchestral music, showing good results in the identification of acoustic performances.