Complex sound discrimination: Predictions of the EWAIF model

Abstract

The envelope‐weighted average of instantaneous frequency (EWAIF) model of auditory perception has been successfully applied to simple, two‐component signals, simultaneously amplitude‐ and frequency‐modulated tones, and even to the complex signals used in the early “profile analysis” work. The EWAIF model predicts performance from a calculation of the envelope‐weighted average of the instantaneous frequency fluctuations inherent in almost every complex periodic sound. When the model is in error, it generally predicts better performance than our listeners achieve. A revised version of the EWAIF model incorporates a temporal processing window. The revised model was tested by comparing its predictions with listeners' performance in a frequency‐change discrimination task. The experiment requires the listener to distinguish a smooth frequency glide from a discrete, multistep transition over the same trajectory. The listeners' ability to distinguish the glide from the multistep transition, in a 2IFC task, decreased to chance as the number of steps increased. The EWAIF model performance follows that of the listeners, given the appropriate choice of temporal window parameters. [Work supported by AFOSR and NIH.]