权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Maximizing Speech Recognition Under Adverse Listening Conditions

在不利的聆听条件下最大限度地提高语音识别能力

基本信息

批准号：
8678897
负责人：
Daniel Fogerty
金额：
$ 10.61万
依托单位：
UNIVERSITY OF SOUTH CAROLINA AT COLUMBIA
依托单位国家：
美国
项目类别：
财政年份：
2012
资助国家：
美国
起止时间：
2012-07-01 至 2016-06-30
项目状态：
已结题

项目摘要

DESCRIPTION (provided by applicant): Current progress in the programming and customization of assistive listening devices is limited due to an inadequate understanding of the context-dependent weighting of acoustic cues, the interference of these cues under adverse listening conditions, and the processing limitations imposed on these cues by aging and hearing impairment. This proposal targets this gap by identifying how older normal-hearing and hearing-impaired listeners use auditory cues during aided presentations for understanding speech in noise. The primary objectives of this project are to identify (1) the auditory cues that maximize speech understanding under adverse listening conditions, (2) how older listeners perceptually weight those cues, and (3) what auditory properties of the masking speech limit speech understanding the most. The central aim is to examine and identify the most informative auditory cues when only partial speech information is available in noisy environments for older listeners when audibility is restored. The central hypothesis is that in quiet, temporal envelope cues contribute most to speech intelligibility and are available for older listeners with aided hearing. However, in noise and competing speech the contribution of these acoustic cues are more limited, with the contribution of other cues becoming more important, such as the temporal fine structure, the processing of which may be limited by age or cochlear pathology. Specific Aim #1 addresses speech in continuous, interrupting, and speech-babble noise. Different listener groups enable the investigation of age, cochlear pathology, and amplification contributions to auditory perceptual weights. Correlational analysis will explore the relationship between perceptual weights, speech in noise performance, and cognitive abilities. Specific Aim #2 investigates temporal properties of the target and competing speech and how they interact. These experiments explore how temporal properties of the target talker facilitate speech understanding and properties of the competing talker interfere. Informational masking is also explored via time-reversed competition. The long-term goal of this project is to define acoustic parameters for enhanced programming of assistive hearing technology and identify individual weighting strategies to assist in future customization of these devices to capitalize on existing capabilities of the device and the listener. The significant contribution of this project is in identifying the speech cues that will be most informative for these listeners in different noisy conditions. The approach of this project is innovative. It uses novel signal processing strategies to independently vary complex temporal properties of speech via noisy signal extraction. Furthermore, it extends the 'glimpsing' theory of automatic speech recognition to the human understanding of speech from partial information. These innovations allow for the direct investigation of auditory temporal cue use during a competing talker paradigm, quite arguably the most difficult listening condition for older listeners.

描述（由申请人提供）：由于对声学提示的上下文相关加权、不利听力条件下这些提示的干扰以及老化和听力损伤对这些提示施加的处理限制的理解不足，目前在辅助听力设备的编程和定制方面的进展有限。这项建议针对这一差距，确定老年人的听力正常和听力受损的听众如何使用听觉线索在辅助介绍理解语音噪声。这个项目的主要目标是确定（1）在不利的听力条件下最大限度地提高语音理解的听觉线索，（2）老年听众如何感知这些线索的权重，以及（3）掩蔽语音的听觉特性最大限度地限制语音理解。其中心目的是检查和识别最翔实的听觉线索时，只有部分语音信息是在嘈杂的环境中为老年听众时，恢复可听。中心的假设是，在安静的，时间包络线索有助于语音清晰度，并可用于老年人的听众与辅助听力。然而，在噪声和竞争语音中，这些声学线索的贡献更加有限，其他线索的贡献变得更加重要，例如时间精细结构，其处理可能受到年龄或耳蜗病理学的限制。具体目标#1解决了连续、中断和语音多路重合噪声中的语音。不同的听众群体，使调查的年龄，耳蜗病理，和放大的听觉感知权重的贡献。相关性分析将探讨感知权重、噪声中语音性能和认知能力之间的关系。具体目标#2研究目标和竞争语音的时间特性以及它们如何相互作用。这些实验探讨了目标说话者的时间特性如何促进语音理解，而竞争说话者的时间特性如何干扰语音理解。信息掩蔽也通过时间反演的竞争。该项目的长期目标是为辅助听力技术的增强编程定义声学参数，并确定个人加权策略，以帮助这些设备的未来定制，以利用设备和收听者的现有功能。这个项目的重大贡献是在识别语音线索，将是最翔实的这些听众在不同的嘈杂条件。这个项目的方法是创新的。它使用新的信号处理策略，通过噪声信号提取独立地改变语音的复杂时间特性。此外，它还将自动语音识别的“瞥见”理论扩展到人类从部分信息中理解语音。这些创新允许直接调查的听觉时间线索的使用在竞争的谈话者范例，相当可以说是最困难的听力条件为老年听众。