Factors that facilitate perception of degraded speech and speech in noise

在日常生活中，人们经常在干扰谈话者或其他噪音的背景下听到讲话。如果背景声音足够强烈，目标语音会被有效地降级，变得难以理解，甚至最复杂的自动语音识别系统也会在很大程度上失败。然而，正常听力的人类听众在类似的背景噪声水平下理解语音具有明显的（欺骗性的）容易。人类善于利用自己的知识和经验来帮助他们解释模糊的语音信号。特别是，有意义的语义上下文有助于年轻人和老年人理解口语句子，特别是在困难的听力条件下。虽然这一研究有着悠久的历史，但我们对语境效应的理解--它们何时出现以及它们影响理解的机制--是有限的。这里提出的工作将探讨如何约束信息（即，有意义的上下文）在语音难以理解时有助于理解，或者因为它以某种方式降级（例如，因为对话者有很重的口音或语音是通过耳蜗植入物转换的）或因为它被背景声音部分地掩蔽。这项工作将从根本上有助于言语知觉的语境效应的理论解释。这一点很重要，因为在噪音中理解语音是老年人最常见的抱怨之一，迄今为止的研究主要集中在解释这个问题（与年龄相关的听力和认知变化），就其本质而言，这是不容易补救的。我实验室的研究是非传统的，重点是促进成功感知语音的因素，这些因素可能会减轻听力损失的影响。其中最重要的是有效地利用有意义的上下文。随着人们年龄的增长和声音输入处理的保真度的恶化，这种基于知识的语音理解线索变得越来越重要。更全面地了解促进言语理解的因素有助于优化它们并促进它们的使用。

In everyday life, speech is often heard against a background of interfering talkers or other noise. If the background sound is sufficiently intense, target speech is effectively degraded and becomes hard to understand, to the point that even the most sophisticated automatic speech recognition systems will largely fail. Nevertheless, normally hearing human listeners presented with similar levels of background noise understand speech with apparent (deceptive) ease. Humans are adept at using their knowledge and experience to help them interpret an ambiguous speech signal. In particular, meaningful semantic context facilitates comprehension of spoken sentences by young and older adults, especially in difficult listening conditions. Although this research has a long history, our understanding of contextual effects -- when they arise and the mechanisms by which they influence comprehension -- are limited. The work proposed here will explore how constraining information (i.e., meaningful context) facilitates comprehension when speech is difficult to understand, either because it is degraded in some way (e.g., because the interlocutor has a strong accent or the speech is transduced through a cochlear implant) or because it is partially masked by background sounds. This work will contribute fundamentally to a theoretical account of contextual effects in speech perception. This is important because understanding speech in noise is one of the most common complaints of older individuals, and research to date has focused on explanations of the problem (age-related changes in hearing and cognition) that, by their nature, are not easily remediable. Research in my lab is unconventional in its focus on factors that facilitate successful perception of speech and which may act to mitigate the effects of hearing loss. One of the most important of these is the effective use of meaningful context. As people age and the fidelity with which acoustic input is processed deteriorates, such knowledge-based cues to speech comprehension become increasingly important. A more complete understanding of the factors that can facilitate speech comprehension is instrumental to optimizing them and promoting their use.