Phase vocoder and beyond

Abstract

For a broad range of sound transformations, quality is measured according to the common expectation about the result: if a male’s voice has to be changed in a female’s one, there exists a common reference for the perceptive evaluation of the result; the same holds if an instrumental sound has to be made longer, or shorter. Following the argument in Röbel, “Between Physics and Perception: Signal Models for High Level Audio Processing”, a fundamental requirement for these transformation algorithms is their need of signal models that are strongly linked to perceptually relevant physical properties of the sound source. This paper is a short survey about the phase vocoder technique, together with its extensions and improvements relying on appropriate sound models, which have led to high level audio processing algorithms.