This paper briefly describes a system called SaxEx, capable of generating expressive musical performances based on examples. We have done several recordings of a tenor sax expressively playing Jazz ballads. These recordings are analyzed, using spectral modelling techniques, to extract information related to five expressive parameters. The results of this analysis, together with the score, constitute the set of examples (cases) of the case-based component SaxEx. From these examples, plus background musical knowledge based on Narmour’s implication/realization theory of musical perception and Lerdahl and Jackendoffs generative theory of tonal music understanding (GTTM), SaxEx is able to infer a set of expressive transformations to apply to any given input sound file containing an inexpressive musical phrase of another ballad. Finally, SaxEx uses its spectral synthesis capabilities to actually apply the inferred transformations to the input sound file resulting in an expressive sound file at the output.
[1]
Ramón López de Mántaras,et al.
Perspectives: A Declarative Bias Mechanism for Case Retrieval
,
1997,
ICCBR.
[2]
Josep Lluis Arcos,et al.
The Noos Representation Language
,
1997
.
[3]
Xavier Serra,et al.
Musical Sound Modeling with Sinusoids plus Noise
,
1997
.
[4]
Stephen Travis Pope,et al.
Musical Signal Processing
,
1997
.
[5]
Stephan M. Schwanauer,et al.
Machine Models of Music
,
1993
.
[6]
Xavier Serra,et al.
SaxEx: a case-based reasoning system for generating expressive musical performances
,
1998,
ICMC.
[7]
Ray Jackendoff,et al.
An overview of hierarchical structure in music
,
1983
.
[8]
Eugene Narmour,et al.
The Analysis and Cognition of Basic Melodic Structures: The Implication-Realization Model
,
1990
.