Mechanisms of noise robust representation of speech in primary auditory cortex

Abstract

Humans and animals can reliably perceive behaviorally relevant sounds in noisy and reverberant environments, yet the neural mechanisms behind this phenomenon are largely unknown. To understand how neural circuits represent degraded auditory stimuli with additive and reverberant distortions, we compared single-neuron responses in ferret primary auditory cortex to speech and vocalizations in four conditions: clean, additive white and pink (1/f) noise, and reverberation. Despite substantial distortion, responses of neurons to the vocalization signal remained stable, maintaining the same statistical distribution in all conditions. Stimulus spectrograms reconstructed from population responses to the distorted stimuli resembled more the original clean than the distorted signals. To explore mechanisms contributing to this robustness, we simulated neural responses using several spectrotemporal receptive field models that incorporated either a static nonlinearity or subtractive synaptic depression and multiplicative gain normalization. The static model failed to suppress the distortions. A dynamic model incorporating feedforward synaptic depression could account for the reduction of additive noise, but only the combined model with feedback gain normalization was able to predict the effects across both additive and reverberant conditions. Thus, both mechanisms can contribute to the abilities of humans and animals to extract relevant sounds in diverse noisy environments.hearing | cortical | population code | phonemes V ocal communication in the real world often takes place in complex, noisy acoustic environments. Although substantial effort is required to perceive speech in extremely noisy conditions, accurate perception in moderately noisy and reverberant environments is relatively effortless (1), presumably because of the presence of general filtering mechanisms in the auditory pathway (2). These mechanisms likely influence the representation and perception of both speech and other natural sounds with similarly rich spectrotemporal structure, such as species-specific vocalizations (3-5). Despite the central role this robustness must play in animal and human hearing, little is known about the underlying neural mechanisms and whether the brain maintains invariant representations of these stimuli across variable soundscapes causing acoustic distortions of the original signals.Several theoretical and experimental studies have postulated that the distribution of linear spectrotemporal tuning of neurons found in the auditory pathway could support enhanced representation of temporal and spectral modulations matched to those prevalent in natural stimuli (6, 7). Others have attributed this effect to nonlinear response properties of neurons (8) and adaptation with various timescales (9). In this study, we tested the noise robustness of auditory cortical neurons by recording responses in ferret primary auditory cortex (A1) to natural vocalizations that were distorted by additive white and pink (1/f) noise or by convolutive reverber...

Neural Noise Hypothesis of Developmental Dyslexia

“…Results from modeling in vivo responses from gerbil primary auditory cortex in response to phonemes in noise suggest that the dynamic nonlinear control of neural thresholds, requiring both synaptic depression and gain adjustments, is critical for reducing the effects of perceptual noise [86]. In particular, synaptic depression was found to be necessary for suppressing additive perceptual noise of the type used in studies in noise exclusion in RD [85].…”

“…For example, a previous study [86] used simple filter models to investigate the mechanisms involved in auditory noise exclusion, a deficit in RD [85]. Importantly, these models perform differently under different noise conditions, allowing them to be empirically distinguished in vivo (in gerbils).…”

“…Processes in light text are not discussed in detail in the main text. Numbers by arrows refer to supporting references in the main text ( [5,9,10,11,27,39,44,45,49,55,58,72,75,80,82,84,85,86,88,89,90]). In a balanced excitation-inhibition regimen, stimuli evoke excitatory synaptic conductances (green curve) followed by comparable inhibitory conductances (red curve) within a few milliseconds.…”

Neural Segregation of Concurrent Speech: Effects of Background Noise and Reverberation on Auditory Scene Analysis in the Ventral Cochlear Nucleus

“…The differential effects of noise and reverberation on speech representations in the early stages of brain-stem neural processing are in clear contrast to noise-and reverberation-invariant representations of speech in auditory cortex (Mesgarani et al 2014). Perhaps one clue to the neural underpinnings of robust speech understanding in challenging acoustic environments is compensation for the effects of reverberation on coding of temporal-envelope modulation in the inferior colliculus, based on sensitivity to inter-aural correlation (Slama and Delgutte 2015).…”

Robust Voice Activity Detection Algorithm Based on Feature of Frequency Modulation of Harmonics and Its DSP Implementation

“…Based on the neurophysiological recordings, a computational auditory model has been proposed accordingly [25] and used to derive a noise-robustness representation of speech [26]. On the other hand, psychoacoustic experiments also demonstrate that the joint spectro-temporal modulations are highly related to speech intelligibility [27] and speech comprehension [11].…”