X. Amatriain, J. Bonada, A. Loscos and X. Serra
This chapter deals with spectral processing where the audio signal is represented by spectral models such as sinusoids plus a residual signal. We introduce two spectral models: Sinusoidal and Sinusoidal plus Residual. With these models we can identify and extract higher-level information of a sound, such as: harmonics, pitch, spectral shape, vibrato, or note boundaries. This analysis step brings the representation closer to our perceptual understanding of a sound. Techniques for analysis, higher-level feature analysis and synthesis are introduced and a variety of new audio effects based on these spectral models are discussed. Effect applications go from nearly lossless time stretching, pitch transposition, vibrato, spectral shape shift, gender change to harmonizer and morphing effects.
Keywords: Modeling, Spectral Models, Signal Representations, Transforms, Fourier Transforms, Feature Extraction