Characterizaton of Non-linear Auditory Receptive Fields

描述(申请人提供)：听觉神经元的特征是它们的感受场属性，这些属性描述了神经元对刺激的频率和幅度等参数的组织。对于听觉系统较低水平的神经元，感受野的描述相对较好，但较高级别的神经元表现出复杂的非线性，阻碍了系统的量化。然而，高阶感受野的特征对于理解正常和异常的听觉感知过程是基本的，也可能有助于优化假体设备的性能。已经描述了一个吸引人的模型系统，在该系统中，八哥的高阶听觉神经元对嵌入自然信号(歌曲)的声学对象(主题)具有高度的选择性。最近的结果表明，八哥具有显著的序列分析能力，表现出对典型的类似语法的结构的敏感性。在这项拟议的研究中，新的统计技术将与对其歌声识别行为受操作控制的八哥的cmHV神经元的生理记录相结合。在第一个实验中，将开发统计方法，使用贝叶斯推理模型的马尔可夫随机场，并结合非线性时间动力学，学习有效的基元结构基组，并使用分层非线性回归估计基于特征的感受野。在第二个实验中，将使用操作程序来识别八哥在歌曲识别行为中使用的主题的自然特征和主题序列的参数。在第三个实验中，cmHV响应将被表征，重点是分析通过行为测试识别的特征，并使用统计方法来表征非线性特性。这种方法的成功开发将使人们对复杂的声音对象和序列识别行为有定量的神经生理学洞察，同时开发出一种对大脑皮层感觉生理学具有普遍实用价值的方法。

DESCRIPTION (provided by applicant): Auditory neurons are characterized by their receptive field properties, which describe the organization of parameters such as frequency and amplitude of stimuli that the neurons respond to. Receptive fields are relatively better described for neurons at lower levels of the auditory system, but higher-order neurons exhibit complex non-linearities that have resisted systematic quantification. Characterization of higher-order receptive fields, however, is fundamental to understanding normal and abnormal auditory perceptual processes, and also may help optimize performance of prosthetic devices. An attractive model system has been described whereby high-order auditory neurons in starlings become highly selective for acoustic objects ("motifs") embedded in natural signals (songs). Recent results implicate remarkable sequence parsing abilities of starlings that exhibit sensitivity to prototypic grammar-like structures. In the proposed research, novel statistical techniques will be combined with physiological recordings of "cmHV" neurons of starlings whose song recognition behavior is under operant control. In the first experiment, the statistical methodologies will be developed for learning efficient basis sets for motif structure and for estimating feature-based receptive fields with hierarchical non-linear regression, using Markov random fields of Bayesian inference models and incorporating non-linear temporal dynamics. In the second experiment, operant procedures will be used to identify natural features of motifs and parameters of motif sequences that starlings utilize in song recognition behavior. In the third experiment, cmHV responses will be characterized, with emphasis on analysis of the features identified by behavioral testing, and using the statistical methods for characterization of non-linear properties. Successful development of this approach would give quantitative neurophysiological insight into complex acoustic object and sequence recognition behavior while developing an approach of general utility to cortical sensory physiology.