Sound Source Segregation

声源隔离

• 批准号: 7728345
• 负责人: MARK A BEE
• 金额: $ 33.56万
• 依托单位: UNIVERSITY OF MINNESOTA
• 依托单位国家: 美国
• 财政年份: 2009
• 资助国家: 美国
• 起止时间: 2009-08-01 至 2014-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/7728345
• 关键词: Acoustics; Address; Algorithms; Amphibia; Animal Model; Animals; Auditory; Auditory system; Auricular prosthesis; Basilar Papilla; Behavior; Biological; Biological Models; Biomimetics; Breeding; Characteristics; Cochlear Implants; Cognitive; Communication; Complement; Computers; Coupled; Cues; Data; Discrimination; Employee Strikes; Environment; Evolution; Female; Frequencies; Future; Goals; Hearing; Hearing Aids; Human; Individual; Knowledge; Labyrinth; Lead; Life; Literature; Location; Logic; Masks; Mediating; Methods; Modeling; Natural Selections; Neurosciences; Noise; Organism; Partner in relationship; Pattern Recognition; Perception; Physiologic pulse; Problem Solving; Process; Property; Pulse Rates; Rana; Recording of previous events; Relative (related person); Research; Role; Sensory Receptors; Shapes; Signal Transduction; Simulate; Social Environment; Solutions; Source; Stream; System; Testing; Thinking; Time; Trees; Ursidae Family; Vertebrates; Work; analog; base; experience; improved; innovation; insight; interest; male; middle ear; neuromechanism; neurophysiology; novel; pressure; research study; segregation; social; sound; speech recognition; vocalization

DESCRIPTION (provided by applicant): The broad aim of this research is to better understand the mechanisms of sound source perception. A critical aspect of perceiving distinct sound sources in multi-source environments is the auditory system's ability to separate sounds of interest from other overlapping sounds and background noise. Hearing aids and cochlear implants often provide their users little benefit when it comes to segregating multiple sound sources. Similarly, computer algorithms for automated speech recognition (ASR) also have persistent difficulty segregating multiple sources. A major goal of auditory neuroscience is to uncover the neural mechanisms that perform sound source segregation. A better understanding of these mechanisms, in turn, can lead to biologically-inspired improvements in hearing prosthetics and ASR algorithms. Though seldom stated explicitly in hearing research, the logic of this biomimetic approach is based on evolutionary thinking: the aim is to understand how natural selection has already solved problems of sound source segregation in living organisms. Naturally, most of this work is done using mammalian models because their auditory systems are most similar to those of humans. The project proposed here employs similar logic based on the premise that evolution is well known for finding diverse solutions to common problems in different animal lineages. The long-term goal of the proposed research is to increase knowledge about the mechanisms of sound source segregation by integrating perceptual and neurophysiological experiments in a lower vertebrate model (frogs) with a unique auditory system and an evolutionary history of solving difficult problems of source segregation. This project uses well-established methods to investigate how female tree frogs segregate the mating calls of individual males from the overlapping signals of other males and the general din of noise in a large breeding chorus. The problems that frogs encounter (and solve) when communicating in noisy social aggregations share many similarities with the human cocktail party problem. Three specific aims will investigate the spectral, temporal, and spatial cues that promote sound source segregation. Aim 1 (spatial release from masking) will investigate how the frog auditory system exploits spatial separation between signals and noise to achieve a release from auditory masking. Aim 2 (masking release in modulated noise) will investigate a form of masking release that depends on a listener's ability to exploit temporal fluctuations in background noise levels. Aim 3 (auditory stream segregation) will investigate the perceptual segregation of two overlapping calls. By investigating these aims in frogs, this project is expected to generate insights into the potential diversity of neural mechanism by which evolution has solved problems of source segregation. Hearing prosthetics and computer algorithms for automated speech recognition perform poorly in environments with multiple competing sound sources. A better understanding of how evolution has solved this type of sound source segregation problem in a diversity of animal models could lead to further biologically-inspired technological advances. Results from this study, and future related projects that will integrate behavior with neurophysiological methods, are expected to generate new and deeper insights into the neurosensory mechanisms of sound source segregation in a lower vertebrate model system that evolved to vocally communicate in noisy, multi-source environments.

描述（由申请人提供）：本研究的主要目标是更好地理解声源感知的机制。在多源环境中感知不同声源的一个关键方面是听觉系统将感兴趣的声音与其他重叠声音和背景噪声分开的能力。在隔离多个声源方面，助听器和人工耳蜗通常无法给用户带来什么好处。同样，自动语音识别 (ASR) 的计算机算法也一直难以隔离多个来源。听觉神经科学的一个主要目标是揭示执行声源分离的神经机制。反过来，更好地理解这些机制可以促进听力假体和 ASR 算法的生物学改进。尽管在听力研究中很少明确说明，但这种仿生方法的逻辑基于进化思维：目的是了解自然选择如何解决生物体中的声源隔离问题。当然，大部分工作都是使用哺乳动物模型完成的，因为它们的听觉系统与人类的听觉系统最相似。这里提出的项目采用了类似的逻辑，其前提是进化因为不同动物谱系的常见问题找到不同的解决方案而闻名。该研究的长期目标是通过将感知和神经生理学实验整合到具有独特听觉系统和解决声源分离难题的进化历史的低等脊椎动物模型（青蛙）中，增加对声源分离机制的了解。 。该项目使用成熟的方法来研究雌性树蛙如何将单个雄性树蛙的交配叫声与其他雄性树蛙的重叠信号以及大型繁殖合唱团中的噪音分开。青蛙在喧闹的社交群体中交流时遇到（和解决）的问题与人类鸡尾酒会问题有许多相似之处。三个具体目标将研究促进声源分离的光谱、时间和空间线索。目标 1（掩蔽的空间释放）将研究青蛙听觉系统如何利用信号和噪声之间的空间分离来实现听觉掩蔽的释放。目标 2（调制噪声中的掩蔽释放）将研究一种掩蔽释放形式，该形式取决于听众利用背景噪声水平的时间波动的能力。目标 3（听觉流分离）将研究两个重叠呼叫的感知分离。通过研究青蛙的这些目标，该项目预计将深入了解神经机制的潜在多样性，进化通过该机制解决了源隔离问题。用于自动语音识别的听力假体和计算机算法在具有多个竞争声源的环境中表现不佳。更好地理解进化如何在多种动物模型中解决此类声源分离问题可能会带来进一步的生物学启发技术进步。这项研究的结果以及未来将行为与神经生理学方法相结合的相关项目，预计将对低等脊椎动物模型系统中声源分离的神经感觉机制产生新的、更深入的见解，该系统进化为在嘈杂的多源声音中进行交流。环境。