Extracting Expressive Tempo Curves from Music Recordings

Musicians give a piece of music their personal touch by continuously varying tempo, dynamics, and articulation. Instead of playing mechanically they speed up at some places and slow down at others in order to shape the piece of music. Similarly, they continuously change the sound intensity and stress certain notes. The automated analysis of different interpretations, also referred to as performance analysis, has become an active research field [2, 4, 5, 7, 9, 8]. Here, one goal is to find commonalities between different interpretations, which allow for the derivation of general performance rules. A kind of orthogonal goal is to capture what is characteristic for the style of a particular musician. Before one can actually analyze a specific performance, one needs the information about when and how the notes of the underlying piece of music are played, see Fig. 1. Such information typically comprises parameters that make explicit the exact timing and intensity of the various note events occurring in the performance. Most of the current algorithms for automated performance analysis rely on accurate annotations of the music material by means of such parameters. Here, the annotation process, in particular in the case of music recordings, is often done manually, which is prohibitive in view of large audio collections. In this paper, we present a fully automatic approach for extracting temporal information from music recordings. This information is given in the form of tempo curves that reveal the relative tempo difference between an actual performance and some reference representation (in the form of a MIDI file) of the underlying musical piece. We conclude this paper with a short summary on previous work in the field of performance analysis and give an outlook on future work.