RI: Medium: Collaborative Research: Understanding Individual-Level Speech Variability: From Novel Articulatory Data to Robust Speaker Recognition

RI：媒介：协作研究：了解个体层面的语音变异性：从新颖的发音数据到强大的说话人识别

• 批准号: 1514544
• 负责人: Shrikanth Narayanan
• 金额: $ 119.95万
• 依托单位: University of Southern California
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-09-01 至 2020-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1514544&HistoricalAwards=false
• 关键词: RI; Medium; Collaborative; Research; Understanding

Speech is a unique human capability. The vocal tract is the universal human instrument played with great dexterity and skill in the production of speech to convey rich linguistic and paralinguistic information. The project will enable fundamental understanding of how individuals differ in their speech articulation due to differences in shape and size of their physical vocal instrument. Knowledge of how people differ in their speech production can help create improved automatic speaker recognition, technologies important for national security. The project can inform design of technologies for robust speech-based access for all members of the population, including children, the elderly, and non-native speakers of a language. Results from the project can also assist in better understanding and treating disorders (e.g., cleft lip/palate), illness (e.g., head and neck cancer, apnea) or injury where human speech articulation is affected. The novel imaging data from 200 individuals, and associated tools, annotations and interpretations created by the interdisciplinary team will be shared broadly with the scientific community. The project will provide a unique research training opportunity for students in integrated speech science and technology. The overarching goal of this project is to advance scientific understanding of how vocal tract morphology and speech articulation interact and explain the variant and invariant aspects of speech signal properties across talkers. Of particular scientific interest is the nature of articulatory strategies adopted by individuals in the presence of structural differences across them to achieve phonetic equivalence. Equally of interest are what aspects of, and how, vocal tract morphological differences are reflected in the acoustic speech signal, and if those differences can be estimated from speech acoustics. A crucial part of this goal is to create forward and inverse computational models that relate vocal tract details to speech acoustics toward shedding light on individual speaker differences and informing design of robust speaker recognition technologies. This project goes beyond state-of-the-art methods by focusing on direct investigation of the dynamic human vocal tract using novel imaging techniques and computational modeling to illuminate inter-speaker variability in vocal tract structure, as well as the strategies by which linguistic articulation is implemented. Using novel Magnetic Resonance Imaging with superior spatial resolution of the entire moving vocal tract that we helped develop (dynamic realtime 2D with excellent temporal resolution and accelerated volumetric 3D), the project will gather and quantify spatio-temporal details of speech production from 160 native American English covering the major dialectal regions of North America and 40 non-native speakers. The experimental, theoretical, and methodological approaches investigating the interplay between structure (shape and size) and function (dynamics of vocal-tract shaping and its acoustic consequences) can lead to new theoretical advances with improved phonetic characterizations of linguistic units that are general across speakers. It also offers the ability to explain individual specific speech patterns that can improve both understanding the scientific underpinning and creating robust automatic speaker recognition technology, enabling to determine not only that two talkers are different by the adoption of novel speaker dependent features, but also how and why they differ, by analyzing biologically-inspired details of structure and articulation.

说话是人类独有的能力。声道是人类普遍使用的乐器，在语言的产生和传递丰富的语言和副语言信息中发挥着极大的灵活性和技巧。该项目将使人们能够从根本上了解个体如何因其身体发声工具的形状和大小的差异而导致其语音清晰度的差异。了解人们在说话时的差异有助于改进自动说话人识别技术，这项技术对国家安全至关重要。该项目可以为所有人口（包括儿童、老人和非母语人士）提供基于语音的强大访问技术设计信息。该项目的结果还可以帮助更好地理解和治疗疾病（如唇裂/腭裂）、疾病（如头颈癌、呼吸暂停）或人类语言表达受到影响的伤害。来自200个人的新成像数据，以及跨学科团队创建的相关工具，注释和解释将与科学界广泛共享。该项目将为综合语音科学与技术的学生提供一个独特的研究训练机会。该项目的总体目标是促进对声道形态和语音发音如何相互作用的科学理解，并解释说话者语音信号特性的变异和不变方面。特别令人感兴趣的是，在存在结构差异的情况下，个体为实现语音对等而采用的发音策略的性质。同样令人感兴趣的是，声道形态差异的哪些方面以及如何反映在声学语音信号中，以及这些差异是否可以从语音声学中估计出来。该目标的一个关键部分是创建将声道细节与语音声学联系起来的正向和反向计算模型，以揭示个体说话者的差异，并为设计稳健的说话者识别技术提供信息。该项目超越了最先进的方法，专注于动态人类声道的直接调查，使用新颖的成像技术和计算模型来阐明声道结构中的说话人间变异性，以及实现语言发音的策略。使用我们帮助开发的具有优越空间分辨率的整个运动声道的新型磁共振成像（具有优异时间分辨率的动态实时2D和加速体积3D），该项目将收集和量化来自160个北美主要方言地区的美国本土英语和40个非母语人士的语音产生的时空细节。研究结构（形状和大小）和功能（声道形成的动力学及其声学后果）之间相互作用的实验、理论和方法方法，可以通过改进语言单位的语音特征，在说话者之间普遍存在，从而导致新的理论进展。它还提供了解释个人特定语音模式的能力，可以提高对科学基础的理解，并创建强大的自动说话人识别技术，不仅可以通过采用新颖的说话人依赖特征来确定两个说话人的不同之处，还可以通过分析生物学启发的结构和发音细节来确定他们如何以及为什么不同。