| Sumario: | This monograph reports an investigation into the use of case based reasoning for expressivity aware tempo transformation of audio-recorded performances of melodies. This specific task is illustrative of a wider application domain that has emerged during the past decade: content-based multimedia processing. Large scale availability of image, video, and audio information in digital format requires new ways of managing and transforming information. The work presented in the monograph is an example of such content-based transformation. The monograph investigates the problem of how a musical performance played at a particular tempo can be rendered automatically at another tempo, while preserving naturally sounding expressivity. The work presented raises a number of challenging topics. First there is the question of data modeling. It is an open issue how expressivity information can be extracted from the performance and appropriately represented, and what aspects of the melody should be explicitly described for content-based manipulation of performed music. Secondly, from the case based reasoning perspective tempo-transformation of performed melodies is an interesting problem domain, since the problem and solution data are composite structures of temporal nature, and the domain expertise (expressively performing music) is almost entirely a tacit skill. Thirdly, since problem solving in case based reasoning is based on the reuse of previous problems, similarity measures for melodies and performances play a central role. This creates an overlap with the field of music information retrieval. Lastly, this research raises the question of how the quality of transformed performances can be evaluated. The evaluation of models for expressive music performance is an important unsettled issue that deserves broad attention.
|
|---|
