Predicting discrimination of formant frequencies in vowels with a computational model of the auditory midbrain

Neural information for encoding and processing temporal information in speech sounds occurs over different time-courses. We are interested in temporal mechanisms for neural coding of both pitch and formant frequencies of voiced sounds such as vowels. In particular, in this study we will describe a strategy for quantifying the ability to discriminate changes in spectral peaks, or formant frequencies, based on the responses of neural models. Previous studies have explored this question based on responses of computational models for the auditory periphery, that is, responses of the population of auditory-nerve (AN) fibers (e.g. [1]-[2]). In this study we quantify formant-frequency discrimination based on the responses of models for auditory midbrain neurons at the level of the inferior colliculus (IC). These neurons are tuned to both audio frequency and to low-frequency amplitude modulations, such as those associated with pitch.