Auditory Scene Analysis with Complex Sounds

复杂声音的听觉场景分析

• 批准号: 9759900
• 负责人: Josh H McDermott
• 金额: $ 37.42万
• 依托单位: MASSACHUSETTS INSTITUTE OF TECHNOLOGY
• 依托单位国家: 美国
• 财政年份: 2016
• 资助国家: 美国
• 起止时间: 2016-09-01 至 2021-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9759900
• 关键词: Acoustics; Address; Age; Algorithms; Area; Attention; Auditory; Auditory system; Behavioral; Binaural; Brain; Characteristics; Cochlear Implants; Communication; Competence; Complex; Cues; Dependence; Development; Devices; Disease; Ear; Engineering; Exhibits; Frequencies; Frustration; Goals; Grouping; Hearing; Hearing Aids; Human; Implant; Knowledge; Measures; Methods; Modeling; Music; Noise; Performance; Process; Property; Prosthesis; Psychoacoustics; Reporting; Research; Restaurants; Role; Signal Transduction; Social isolation; Source; Speech; Speech Intelligibility; Stimulus; Stream; Structure; System; Technology; Testing; Time; Voice; Work; auditory comprehension; automated speech recognition; design; developmental disease; experimental study; hearing impairment; improved; insight; interest; normal hearing; novel; remediation; segregation; sound; speech recognition; speech synthesis; success; vocal cord

PROJECT SUMMARY / ABSTRACT Perhaps the most pervasive problem faced by listeners with hearing impairment or cochlear implants is the difficulty of recognizing speech and other sounds in the presence of competing sound sources, as when conversing at a restaurant. This difficulty in “sound segregation” – hearing a particular sound of interest when it is embedded in a mixture of other sounds – often leads to frustration and social isolation, and is not adequately addressed by current hearing aids and implants. Sound segregation difficulties are also commonly reported in developmental auditory disorders. The long-term goal of the proposed research is to reveal the basis of sound segregation and to provide insights that will facilitate improved prosthetic devices and remediation strategies, as well as more effective machine systems for processing sounds, e.g. for automatic speech recognition. The development of more effective devices, technologies, and therapies is currently limited by an incomplete understanding of the factors that underlie sound segregation by normal-hearing listeners. In particular, little is known about sound segregation with complex naturalistic sounds, in part because much of the research in this area has been conducted using simple signals that are impoverished relative to the sounds listeners normally encounter. We propose to enrich the understanding of sound segregation with three sets of experiments that use novel sound synthesis methods to manipulate properties of natural speech and other sounds and test their role in segregation with behavioral experiments in human listeners. Aim 1 manipulates the classically proposed grouping cue provided by harmonic frequency relations and investigates the mechanisms underlying their effect. Aim 2 investigates the role of prior knowledge of voice and speech structure on segregation, and should help to explain why some voices are easier or harder to segregate than others. Aim 3 investigates the role of attentive tracking in the segregation of sounds from mixtures, and will explore the factors that facilitate tracking or cause it to fail. The results will reveal the mechanisms underlying sound segregation by the healthy auditory system, and will provide insights into the factors that limit auditory comprehension in the presence of multiple sound sources, hopefully suggesting new strategies for signal enhancement, prosthetic devices, and behavioral remediation.

项目概要/摘要 也许有听力障碍或植入人工耳蜗的听众面临的最普遍的问题是 在存在竞争声源的情况下，难以识别语音和其他声音，例如 在餐馆里交谈时，会遇到“声音隔离”的困难——听到特定的感兴趣的声音。 嵌入其他声音的混合中——常常导致沮丧和社会孤立，并且没有充分体现 目前的助听器和植入物所解决的声音分离问题也经常被报道。 该研究的长期目标是揭示声音的基础。 隔离并提供有助于改进假肢装置和补救策略的见解， 以及更有效的声音处理机器系统，例如自动语音识别。 目前，更有效的设备、技术和疗法的开发受到不完整的限制 尤其是，对于听力正常的听众对声音隔离的因素了解甚少。 人们知道复杂的自然声音的声音分离，部分原因是这方面的大量研究 区域是使用简单的相对信号进行的，这些信号通常对声音听众来说是贫乏的 我们建议通过三组实验来丰富对声音分离的理解。 使用新颖的声音合成方法来操纵自然语音和其他声音的属性并测试它们 目标 1 操纵了经典提出的方法。 由谐波频率关系提供的分组线索并研究其背后的机制 目标 2 研究语音和语音结构的先验知识对分离的作用，并且应该。 有助于解释为什么某些声音比其他声音更容易或更难隔离。目标 3 调查了以下角色的作用。 仔细跟踪从混合物中分离声音，并将探索促进跟踪的因素 或导致其失败，结果将揭示健康听觉的声音隔离机制。 系统，并将深入了解在存在多个声音的情况下限制听觉理解的因素 声源，希望能够提出信号增强、假肢装置和 行为矫正。