Paul Mitchell1, Laurel Carney2
1 Department of Biomedical Engineering, University of Rochester, Rochester, USA;Â 2 Departments of Biomedical Engineering and Neuroscience, University of Rochester, Rochester, USA
Background: Recent physiological results have demonstrated that inferior colliculus (IC) neurons in rabbits and gerbils can be sensitive to fast frequency chirps. This sensitivity was shown using harmonic stimuli with chirps within each pitch period caused by the phase spectrum. Chirp sensitivity is not present in state-of-the-art computational models of the IC. We evaluate several models hypothesized to replicate these physiological results. While mechanisms for frequency-modulation (FM) sensitivity have been proposed in previous IC studies, those studies focused on FM specialists, such as bats, rather than generalist mammals. This work has implications for speech coding because fast chirps are contained in vowels. Here, we present a novel computational model of the IC that accurately portrays the influence of chirp sensitivity on responses to common stimuli, while also retaining known IC properties such as amplitude-modulation tuning.
Methods: We use the auditory-nerve model from Zilany et al. (2014) and build upon the same-frequency inhibition and excitation (SFIE) model for IC neurons (Nelson and Carney, 2004). The proposed models are evaluated based on their ability to reproduce physiological responses to chirp stimuli.
Results: The proposed models of IC chirp sensitivity are based on the premise that a neuron selective for chirp direction or velocity receives input from multiple frequency channels. A model that incorporates coincidence detection and delayed inhibition has the best mix of chirp selectivity and preservation of known IC properties. We can vary model configurations, as well as the relative strength, timing, coincidence window size, and duration of inputs.
Conclusion: We tested several models for the neural mechanism of fast chirp sensitivity in the IC. These models allow us to test hypotheses related to chirp sensitivity. This work contributes new insight into the neural representation of speech at the level of the midbrain.
(NIDCD 001641)
Top, SFIE model receiving input from a single frequency, bottom, a cross-frequency model receiving an array of excitatory and inhibitory inputs from multiple frequencies. Right, model responses to two opposite directions of frequency chirp.