Predicting masked thresholds based on physiological models of auditory nerve and inferior colliculus population responses

Abstract

Acoustic cues for detection of tones in background noise include envelope, fine-structure, and energy. These cues are transformed in the process of neural coding and are represented in fundamentally different ways by different types of neurons. Here, to study the detection of a tone masked by noise population responses were derived from model auditory-nerve (AN) fibers with low or high spontaneous rates (HSR or LSR) and inferior colliculus (IC) neurons [M. S. A. Zilany, I. C. Bruce, and L. H. Carney, J. Acoust. Soc. Am. 135, 283-286 (2014) and J. Mao, A. Vosoughi, and L. H. Carney, J. Acoust. Soc. Am. 134, 396-406 (2013)]. Initial analyses focused on average response rates. Percent correct on two-interval forced-choice tasks was estimated based on the distance between responses to each stimulus interval and a template based on noise-alone responses. Although HSR AN fibers had saturated rates, the addition of a tone reduced low-frequency fluctuations of the time-varying rate, a change encoded by modulation-sensitive model IC cells. This HSR-based code was robust to the roving-level paradigm. While the non-saturated LSR AN fibers responded to added tones with an increase in average rate, in the roving-level paradigm these fibers did not provide reliable information to the model IC cell.