Physiological and Computational-Modeling Studies of Timbre Encoding in the Inferior Colliculus

点击翻译按钮获取中文摘要

Timbre, the quality that allows sounds to be distinguished when they are identical in pitch, level, and duration, is a critical aspect of speech comprehension and music enjoyment. My proposal will fill a gap in neural studies of timbre by testing the hypothesis that capture and off-CF inhibitory mechanisms lead to robust representations of suprathreshold synthetic and natural-instrument timbre in the midbrain. To test my hypothesis, I will record single-unit inferior colliculus (IC) responses from awake Dutch-belted rabbits. I will also develop a new computational IC model based on these physiological responses. Spectral envelopes of harmonic sounds are correlated with the timbral perception of “brightness”. I propose two mechanisms that contribute to spectral-envelope encoding: capture and off-characteristic frequency (CF) inhibition. The first mechanism, capture, refers to the dominance of harmonics near spectral peaks over auditory-nerve fibers tuned near the peaks. Capture is due to saturation of inner hair cells. Capture by a single harmonic reduces the amplitude of low-frequency neural fluctuations in auditory-nerve fibers. Rates of IC neurons are sensitive to low-frequency neural fluctuations as characterized by modulation transfer functions. Ultimately, capture of auditory-nerve responses for fibers tuned near spectral peaks results in IC rate profiles that encode spectral peaks. Preliminary results are partially consistent with spectral peaks of synthetic timbre stimuli capturing peripheral responses, leading to a rate representation of salient spectral features in the midbrain. However, another mechanism that could explain IC representations of timbre is off-CF inhibition, which has been proposed to explain frequency-sweep sensitivity and psychophysical forward masking. A subcortical computational model that features capture, but not off-CF inhibition, was able to predict preliminary responses to synthetic timbre and narrowband tone-in-noise, but could not predict responses to wideband tone-in-noise or natural timbre, indicating the need to update the model. I have developed experiments to test the hypothesis that timbre is robustly encoded in the midbrain via capture and off-CF inhibition. Aim 1.1 will test the hypothesis that responses to wideband tone-in-noise are strongly influenced by off-CF inhibition, and reducing the noise bandwidth increases the influence of capture. In Aim 1.2 I will update a computational IC model by adding off-CF inhibition to test the hypothesis that capture and off-CF inhibition are necessary to explain tone-in-noise stimuli. Aim 2.1 will test the hypothesis that the spectral peak of a shaped harmonic complex, synthetic timbre, is robustly encoded in the IC over a range of suprathreshold sound levels. Aim 2.2 bridges the gap between synthetic and natural timbre by recording responses to real instrument sounds. Responses from Aim 2 will further test the new IC model. This project will provide insight on the IC representation of suprathreshold timbre. This research will lead to novel strategies to restore timbre perception in hearing aids and cochlear implants, which are not designed for timbre perception.