Collaborative Research: Enhancing Speech Science Training through Collaboration: Investigating Perception of a Variable Speech Signal

该奖项全部或部分由《2021年美国救援计划法案》（公法117-2）资助。近年来，自动语音识别系统已经变得更加准确和广泛使用。然而，这些系统仍然不能像人类听众那样处理语音的变化。此外，语音技术对代表性不足的群体往往效果不佳。该项目将通过研究人类听众如何理解可变语音，以及通过增加语言科学领域中代表性不足的群体的能力和参与来解决这些问题。这项研究将对人类语言感知如何工作的科学知识做出突出贡献。这些贡献将与改进自动语音识别系统直接相关，自动语音识别系统在技术、医疗保健和教育领域变得无处不在。该项目还将大大加强北卡罗来纳农工大学（美国历史上最大的黑人大学）在语音科学方面的研究能力，并将支持北卡罗来纳农工大学和合作机构宾夕法尼亚州立大学增加对代表性不足群体学生的培训和参与。该项目的研究和培训成果将有助于提高语音科学和技术的效率和公平性。语音科学中一个关键但尚未解决的问题是人类听者如何在高度可变的语音信号下实现鲁棒的语音感知。语言感知理论在一些基本假设上存在分歧：例如，一些理论认为语言感知主要取决于对声音模式的识别，而另一些理论则认为听者感知到引起语音的声道运动。该项目将通过研究协同发音补偿（CfC）现象来测试和改进这些相互竞争的理论，该现象发现听者对语音片段的感知受到周围语音特性的影响。几个实验将被用来检验CfC在言语可变性条件下（例如，说话者可变性、语速可变性和竞争性言语）是如何工作的，以及CfC是如何受学习影响的。此外，该项目将通过以下三个方面建设北卡罗来纳理工学院的能力：(a)为语音科学的学生提供研究培训，(b)促进北卡罗来纳理工学院和宾夕法尼亚州立大学研究人员之间的合作，以及(c)增加教师发展的机会。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2). In recent years, automated speech recognition systems have become much more accurate and widely used. However, these systems still cannot handle speech variability as well as human listeners can. Furthermore, speech technology tends to work less effectively for underrepresented groups. This project will address these problems by investigating how human listeners understand variable speech, and by increasing the capacity and involvement of underrepresented groups in the field of speech science. The research will make prominent contributions to scientific knowledge of how human speech perception works. These contributions will be directly relevant to improving automatic speech recognition systems, which are becoming ubiquitous in technology, healthcare, and education. The project will also significantly strengthen research capacity in speech science at North Carolina A&T State University, which is the Nation’s largest Historically Black University, and will support increased training and involvement of students from underrepresented groups at both NC A&T and at Penn State University, the collaborating institution. The research and training outcomes of the project will help to make speech science and technology both more effective and more equitable.A critical yet unsolved question in speech science is how human listeners achieve robust speech perception despite a highly variable speech signal. Theories of speech perception disagree on basic assumptions: for example, some theories suggest that speech perception depends mainly on recognizing acoustic patterns, while other theories suggest that listeners perceive the vocal tract movements that cause the speech sounds. This project will test and improve these competing theories by investigating the phenomenon of compensation for coarticulation (CfC), which is the finding that listeners’ perception of speech segments is affected by the properties of surrounding speech. Several experiments will be used to examine how CfC works in conditions of speech variability (e.g., talker variability, rate variability, and competing speech), and how CfC is affected by learning. In addition, the project will build capacity at NC A&T along three dimensions by (a) providing research training for students in speech science, (b) fostering collaborations between researchers at A&T and Penn State, and (c) enhancing opportunities for faculty development.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.