The Interaction of Bottom-up and Top-down Information in Human Auditory Learning and Object Recognition

人类听觉学习和物体识别中自下而上和自上而下信息的相互作用

• 批准号: 0749986
• 负责人: Maximilian Riesenhuber
• 金额: $ 58.34万
• 依托单位: Georgetown University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2008
• 资助国家: 美国
• 起止时间: 2008-04-01 至 2012-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0749986&HistoricalAwards=false
• 关键词: Interaction; Bottom; up; Top; down

Object recognition is a crucial cognitive task for all sensory modalities. A particular challenge in object recognition is that physically dissimilar stimuli can have the same label (e.g., the same word spoken by different people, or the same face viewed under different lighting conditions), while physically similar stimuli may be labeled differently (e.g., hearing "b" or "p" based on small differences on when the vocal folds start vibrating). In audition, various animal species, but especially humans, have developed an elaborate system for communication based on the discrimination of fine acoustic differences between complex sounds. Speech perception is probably the most remarkable achievement in this domain and one that may influence the overall architecture of auditory cortex. The underlying neural mechanisms are poorly understood, however. With support from the National Science Foundation, Maximilian Riesenhuber and Josef P. Rauschecker of Georgetown University will address this issue by integrating behavioral results with functional magnetic resonance imaging (fMRI) measures of brain activity. The team will apply the insights gained in understanding the neural bases of visual object recognition to auditory object recognition. One study will investigate the neural mechanisms underlying the categorization and discrimination of speech sounds as well as the transformation of bottom-up acoustic information into categorical phonetic information along the cortical auditory hierarchy. The second study will investigate neuronal plasticity in auditory object recognition by training humans on an auditory categorization task involving modified monkey communication calls. By using novel yet natural auditory communication sounds, the investigators will be able to study the interaction of acoustic information and category labeling during learning and during recognition. The findings will be relevant for the understanding of higher-level auditory processing but also for a general understanding of the interactions between sensory and task-specific information in the brain as well as the commonalities and differences in the mechanisms supporting recognition of speech and non-speech sounds in the brain.Understanding the general principles of sensory processing in the brain, and in particular identifying the underlying neural mechanisms across sensory modalities, has implications for education (e.g., second language learning) and in engineering (e.g., the development of neurally-inspired speech recognition systems). The research project will also help train the next generation of scientists, at the graduate and undergraduate level, with a particular focus on underrepresented minorities. At the graduate level, the project directly involves graduate students from Georgetown University's Interdisciplinary PhD Program in Neuroscience, both at the thesis level and for introductory lab rotations. The project will further impact undergraduate education at Georgetown and beyond, with the opportunity for undergraduates to participate in the project as part of their cognitive science curriculum, or during summer internships. The summer outreach program for undergraduate students has a strong focus on minority recruitment through a partnership with Howard University in Washington, DC.

物体识别是所有感觉模态的重要认知任务。对象识别中的一个特定挑战是物理上不同的刺激可以具有相同的标签（例如，不同的人说的同一个词，或者在不同的照明条件下看到的同一张脸），而物理上相似的刺激可能会被不同地标记（例如，基于声带何时开始振动的微小差异听到“B”或“p”）。在听觉方面，各种动物，尤其是人类，已经发展出一套复杂的交流系统，其基础是辨别复杂声音之间细微的声学差异。 言语知觉可能是这一领域最显著的成就，也可能影响听觉皮层的整体结构。然而，对潜在的神经机制知之甚少。在美国国家科学基金会的支持下，乔治敦大学的马克西米利安·里森胡贝尔和约瑟夫·P·劳斯切克将通过将行为结果与大脑活动的功能性磁共振成像（fMRI）测量相结合来解决这个问题。该团队将把在理解视觉物体识别的神经基础中获得的见解应用于听觉物体识别。其中一项研究将探讨语音分类和辨别的神经机制，以及自下而上的声学信息沿着皮层听觉层次转化为分类语音信息的神经机制。第二项研究将通过训练人类进行听觉分类任务，包括修改猴子的通信呼叫，来研究听觉对象识别中的神经元可塑性。通过使用新颖而自然的听觉交流声音，研究人员将能够研究学习和识别过程中声学信息和类别标签的相互作用。这些发现将有助于理解更高层次的听觉处理，也有助于全面了解大脑中感官和特定任务信息之间的相互作用，以及支持大脑中语音和非语音识别的机制的共性和差异。了解大脑中感官处理的一般原理，特别是识别跨感觉模态的潜在神经机制，对教育具有意义（例如，第二语言学习）和工程学（例如，神经启发的语音识别系统的开发）。该研究项目还将帮助培养下一代研究生和本科生科学家，特别关注代表性不足的少数群体。在研究生阶段，该项目直接涉及乔治敦大学神经科学跨学科博士课程的研究生，无论是在论文水平还是在介绍性实验室轮换。该项目将进一步影响乔治敦大学及其他地区的本科教育，本科生有机会参与该项目，作为其认知科学课程的一部分，或在暑期实习期间。针对本科生的暑期外展计划通过与华盛顿的霍华德大学的合作，非常关注少数族裔的招聘。