Rehabilitating the world’s worst sounds: Time-frequency reassignment in sound modeling and morphing

Abstract

Sound morphing implies a gradual transformation from one sound to another, or the transfer of features of one sound to another. Morphing has been used to synthesize the voices of aliens and animals and a wide variety of sound effects in films and popular music. Additive sound models provide a flexible medium for performing sound morphing and manipulation. The application of time-frequency reassignment—the computation of precise, signal-dependent loci for time-frequency analysis data from first-order partial derivatives of short-time spectral phase—to the construction of robust, manipulable additive sound models will be discussed, and the advantages afforded by the increased precision of reassigned sound models will be demonstrated. Machines, mammals, and musical instruments will be morphed and manipulated, and an attempt will be made to rehabilitate the very worst sounds in the world.