Nonlinear dynamics of complex sound processing in auditory cortex

描述（由申请人提供）：所提出的实验定义了皮层神经元如何在包括背景噪声的自然聆听环境中支持复杂声音（例如语音和其他通信声音）的稳健感知。噪声中的信号（SIN）处理主要是从心理物理学角度进行研究的。支持正常听力表现出的对噪音的显着耐受性的神经机制尚不清楚。我们专注于低频包络信息的编码，因为它对于可理解的语音至关重要，而提高听力障碍者的语音可懂度是一个重要的目标。 重要的临床目标。更广泛地说，声音包络的动态特征（例如共同起始点、偏移和调制特征）驱动听觉场景分割。这些特征在皮质神经元的反应动力学中也得到了很好的体现。然而，事实证明，开发一个理解皮质包络处理的通用框架很困难，因为刺激包络和神经反应模式之间的关系通常既复杂又基本上是非线性的。我们假设皮层反应的非线性动力学赋予它们时间精度，这对于 SIN 处理的鲁棒性至关重要。为了检验这一假设，我们将采用一种新颖的非线性建模框架来估计使用 16 通道线性探针从清醒行为松鼠猴的核心听觉场记录的神经元的频谱时间感受野 (STRF)。我们将评估非线性 STRF 模型（包括简化（例如线性）和修改形式）的能力，以描述皮层对具有简单、参数变化包络的声音的动态响应（目标 1）。我们将比较模型与真实神经元在编码嵌入噪声中的复杂发声方面的性能（目标 2），并测试在最佳贝叶斯群体解码方法的背景下对这些信号进行“去噪”的候选神经机制。最后，我们将通过记录动物在执行单独任务时受到相同复杂刺激的记录来评估注意力过滤对 SIN 处理的影响，其中只有一项任务需要注意详细的包络特征（即调制频率变化检测与声音偏移检测），同时导出 STRF 模型以供后续比较（目标 3）。这些实验将为候选神经机制提供有价值的见解，这些机制支持听觉场景分割的自下而上和自上而下的方面，并支持基于严格的定量模型的方法来表征复杂声音的皮层表示中的层状变换。

DESCRIPTION (provided by applicant): The proposed experiments define how cortical neurons support robust perception of complex sounds, such as speech and other communication sounds, in natural listening environments that include background noise. Signal in noise (SIN) processing has primarily been studied psychophysically; the neural mechanisms that support the remarkable tolerance to noise exhibited by normal hearing are not well understood. We focus on the encoding of low frequency envelope information because it is crucial for intelligible speech and improving speech intelligibility for the hearing impaired is an important clinical goal. More broadly, dynamic features of sound envelopes, such as common onsets, offsets, and modulation characteristics, drive auditory scene segmentation. These features are also particularly well represented in the response dynamics of cortical neurons. However, it has proven difficult to develop a general framework for understanding cortical envelope processing because the relationship between the stimulus envelope and the neural response pattern is typically both complex and substantially nonlinear. We hypothesize that the nonlinear dynamics of cortical responses endow them with a temporal precision that is essential to the robustness of SIN processing. To test this hypothesis, we will employ a novel nonlinear modeling framework to estimate spectrotemporal receptive fields (STRFs) of neurons recorded from the core auditory fields of awake behaving squirrel monkeys using 16- channel linear probes. We will evaluate the ability of nonlinear STRF models - including reduced (e.g., linear) and modified forms - to describe the dynamics of cortical responses to sounds with simple, parametrically varied envelopes (Aim 1). We will compare the performances of the models against real neurons in encoding complex vocalizations embedded in noise (Aim 2), and test candidate neural mechanisms for 'denoising' those signals in the context of optimal Bayesian population decoding methods. Finally, we will assess the effect of attentional filtering on SIN processing by recording from animals presented with identical complex stimuli while engaged in separate tasks, only one of which requires attention to detailed envelope features (i.e., modulation frequency change detection versus sound offset detection), while simultaneously deriving STRF models for subsequent comparison (Aim 3). These experiments will provide valuable insight into candidate neural mechanisms that support both bottom-up and top-down aspects of auditory scene segmentation, and support rigorous quantitative model-based approaches to characterizing laminar transformations in the cortical representation of complex sounds.