Auditory processing of complex signals using the multichannel EWAIF

Abstract

The envelope weighted average instantaneous frequency (EWAIF) has been used to model listener performance in complex signal discrimination [L. L. Feth, J. Acoust. Soc. Am. Suppl. 1 88, S48 (1990)]. Previous work with EWAIF has involved narrow‐band signals that are confined entirely to a single critical band; consequently, it is sufficient to compute a single EWAIF value from the signal. Here, the application of EWAIF to wideband signals is extended. The signal is first passed through the gamma‐tone filter bank of the Patterson‐Holdsworth auditory sensation processing model [R. Patterson, J. Acoust. Soc. Am. Suppl. 1 88, S26 (1990)] and an EWAIF is computed from the output of each channel. This leads to a vector of EWAIF values, which are used in two ways: (1) the EWAIF values from the different channels are combined back into a composite EWAIF value, which is then used to model listener performance in complex sound perception such as profile analysis, comodulation masking release, and modulation masking; and (2) the vector of EWAIF values is used as a feature vector in a vowel recognition task. Preliminary results will be presented and further modifications to the model discussed. [Work supported by a grant from AFOSR.]