Expressivity-Preserving Tempo Transformation for Music - A Case-Based Approach

The research described in this paper focuses on global tempo transformations of monophonic audio recordings of saxophone jazz performances. More concretely, we have investigated the problem of how a performance played at a particular tempo can be automatically rendered at another tempo while preserving its expressivity. To do so we have develppoped a case-based reasoning system called TempoExpress. The results we have obtained have been extensively compared against a standard technique called uniform time stretching (UTS), and show that our approach is superior to UTS. 1 The Problem of Generating Expressive Music It has been long established that when humans perform music from score, the result is never a literal, mechanical rendering of the score (the so-called nominal performance). As far as performance deviations are intentional (that is, they originate from cognitive and affective sources as opposed to e.g. motor sources), they are commonly thought of as conveyingmusical expressivity, which forms an important aspect of music. Twomain functions of musical expressivity are generally recognized. Firstly, expressivity is used to clarify the musical structure (in the broad sense of the word: this includes for example metrical structure [Sloboda, 1983], but also the phrasing of a musical piece [Gabrielsson, 1987], and harmonic structure [Palmer, 1996]). Secondly, expressivity is used as a way of communicating, or accentuating affective content [Juslin, 2001; Gabrielsson, 1995]. An important issue when performing music is the effect of tempo on expressivity. It has been argued that temporal aspects of performance scale uniformly when tempo changes [Repp, 1994]. That is, the durations of all performed notes maintain their relative proportions. This hypothesis is called relational invariance (of timing under tempo changes). However, counter-evidence for this hypothesis has been provided [Desain and Honing, 1994; Timmers et al., 2002], and a recent study shows that listeners are able to determine above chance-level whether audio-recordings of jazz and classical performances are uniformly time stretched or original recordings, based solely on expressive aspects of the performances [Honing, 2006]. Our approach also experimentally refutes the relational C. Freksa, M. Kohlhase, and K. Schill (Eds.): KI 2006, LNAI 4314, pp. 1–6, 2007. c © Springer-Verlag Berlin Heidelberg 2007 2 R.L. de Mántaras, M. Grachten, and J.-L. Arcos invariance hypothesis by comparing the automatic transformations generated by TempoExpress against uniform time stretching.