DESC9115 2014 WRITTEN REVIEW 2: PITCH SHIFTING WITH FORMANTS OF THE ORIGINAL SIGNAL Digital Audio Systems
Rachel Dowdy · 2014
Pitch shifting is a common audio processing tool which can be implemented for a variety of audio applications such as transposition, creating harmony tracks and chorusing to be used in either live or studio applications [1]. When a sound is pitch shifted, its inherent characteristics (formants) are also shifted and the character of the sound changes, deviating from the intention and original sound identity. The aim is to create an effect which maintains the formants of the original signal and imposes this spectral envelope on the pitch shifted signal. Pitch shifting is the inverse process of time stretching, and both are related to sample rate conversion by the imposed scaling factor which determines the amount of shift [2]. These processes can be accomplished in either the frequency domain or the time domain; pitch shifting in the time domain is called time domain harmonic scaling (TDHS) or pitch synchronous overlap and add (PSOLA), where pitch shifting in the frequency domain is referred to as a phase vocoder [3]. Regardless of the domain, the integral component of pitch shifting is interpolation which is the process of reconstructing the original waveform at desired locations; these may be inbetween the previously sampled data points in order to alter the sample rate conversion to either shift the pitch up or down [2]. The basic process of a phase vocoder begins by time scaling the signal by multiplying the signal by a constant giving the degree of pitch shift. To multiply by less than one indicates the signal is pitch shifted down, conversely multiplying by more than one indicates the signal is being pitch shifted up. This signal is then resampled (which is the interpolation process) by the following equation: