D. Arfib, F. Keiler and U. Zölzer
The central topic of this chapter is the separation of the audio signal into its source signal and a time-varying filter derived from the spectral envelope of the signal. These two features are individually processed before synthesis of an output signal. We introduce three techniques for source-filter separation and show source-filter transformations leading to audio effects such as cross-synthesis, formant changing, spectral interpolation and pitch shifting with formant preservation. The last section covers the extraction of further global features such as pitch or fundamental frequency, amplitude envelope, spectral centroid, and autocorrelation features (voiced/unvoiced detection), which can be estimated by simple time-domain or by advanced time-frequency techniques. The main objective here is the introduction and estimation of these parameters for the control of various other digital audio effects, which are presented throughout this book.
Keywords: Spectrum Analysis, Source Separation, Prediction Methods, Spectral Envelope, Pitch Shifting, Feature Extraction