A pitch model based on rate, place, and envelope information

Abstract

A functional model of the auditory processing of pitch is presented. The model draws on Licklider's duplex theory of pitch [Experientia, 7, 128-143] and further extends his approach by adding an envelope processing stage. The incoming signal is sent through a gammatone filter bank to simulate the behavior of the basilar membrane. Within each auditory band the autocorrelation function is calculated and a pitch estimate is made based on the delay between the main peak and the adjacent side peak. A weighting for the salience of each pitch cue is then calculated based on the place theory using the distance of the estimated fundamental frequency from the band's center frequency. Finally, the envelope is determined for each band and an additional pitch estimate is made based on the autocorrelation function of the envelope. These three calculations—which provide rate, place, and envelope information—are combined into a global pitch estimate, explaining the pitch percept associated with a wide range of stimuli. The model can be easily integrated into larger auditory processing frameworks in which envelope information is important (e.g., monaural source segregation and binaural localization). [Work supported by NSF BCS-1539276.]