A Constant Percentage Bandwidth Transform for Acoustic Signal Processing

This paper describes a constant percentage bandwidth transform for acoustic signal processing. Such a transform is shown to emulate behavior found in the human auditory system. A synthesis transformation is developed which, when cascaded with the absence of spectral modification. The effects of spectral domain modification are described. Principles governing discrete implementation of the transform pair are discussed, and relationships are formalized which specify minimal sample requirements for the spectral domain. The Constant-Q spectral magnitude and phase functions are discussed, and three main methods are evaluated whereby the spectral phase may be unwrapped. Finally, the use of the transform pair is discussed in the solution of the perception-related problem of time scale compression and expansion of speech.