Computer-based Speech Practice for Dysarthria: A Preliminary Investigation

描述(申请人提供)：构音障碍是一种常见的运动性言语障碍，继后天神经系统疾病，如中风和创伤性脑损伤(TBI)。构音障碍是残疾的一个主要来源，因为它导致沟通困难，并对就业和重新融入家庭和社区构成重大障碍。构音障碍患者需要随着时间的推移进行广泛的言语练习，以便进行运动学习，以提高他们的沟通能力。然而，为构音障碍患者提供恢复功能性言语所需的广泛言语练习机会，以便在工作、家庭和社区环境中有效参与的言语治疗的财政支持有限。因此，开发更有效的方法来支持和补充构音障碍患者有限的言语治疗服务是至关重要的。这项研究进一步促进了NCMRR/NICHD的使命，即通过康复提高残疾人的独立性和幸福感，通过使用补充语音识别(SSR)程序开发一种新的构音障碍语音练习方法。SSR是由研究团队开发的，它利用语音识别技术作为辅助写作程序来管理严重的构音障碍语音。这项研究将为SSR未来发展成为计算机化的语音练习系统提供初步数据。目标1将专注于发展声学处理构音障碍语音样本的技术能力，以提供一个实践模型。具体地说，这个项目将通过用典型的性别匹配、典型的年龄/性别匹配的和合成的性别匹配的样本来替换扭曲或遗漏的辅音音素，来操纵中度到严重构音障碍患者的语音产生。目标2将使用经过处理的语音样本来识别SSR的识别精度。PI假设不同的操作策略将产生类似的识别精度结果，这是由于他们在识别严重失真和非失真语音的能力方面的经验。这一结果将通过确定侧重于哪种操纵策略来产生功能性言语练习目标来推动开发团队未来的方向，从而将SSR发展成为一种言语练习工具。最后，Aim 3将记录15个试验治疗过程中言语表现的初步变化，以确定SSR在长期言语实践中的潜在作用。PIs假设，中度到重度构音障碍的个体将表现出目标练习单词的言语可理解性增加，并且他们的言语产出在使用SSR时将发生声学变化，与他们的习惯产出相比。这项研究将采用三水平研究设计，包括摩尔(使用受试者内重复测量方差的群体分析)、中间水平和分子(个别详细案例研究)水平的分析。本项目目标的实现将为开发一种计算机练习程序奠定基础，该程序可自动进行语音刺激操作，以进行功能性语音练习。

DESCRIPTION (provided by applicant): Dysarthria is a common motor speech disorder following acquired neurologic conditions such as stroke and traumatic brain injury (TBI). Dysarthria is a major source of disability as it causes difficulty with communication and poses significant barriers to employment and re-integration into the home and community. Individuals with dysarthria require extensive speech practice over time in order for motor learning to occur to improve their ability to communicate. However, there is limited financial support for speech therapy to provide individuals with dysarthria the extensive speech practice opportunities required to regain functional speech for productive participation in work, home and community settings. Thus, it is critical that more effective approaches be developed to support and supplement the limited speech therapy services individuals with dysarthria receive. This study furthers the mission of the NCMRR/NICHD to enhance the independence and well-being of persons with disability through rehabilitation by developing a novel approach to speech practice for dysarthria using the Supplemented Speech Recognition (SSR) program. The SSR has been developed by the research team and utilizes speech recognition technology that can manage severely dysarthric speech as an assistive writing program. This investigation will provide the preliminary data to support the future development of the SSR into a computerized speech practice system. Aim 1 will focus on developing the technical ability to acoustically manipulate dysarthric speech samples to provide a practice model. Specifically, this project will manipulate the speech productions of individuals with moderate to severe dysarthria by replacing distorted or omitted consonant phonemes with typical gender-matched, typical age/gender-matched, and synthetic gender-matched samples. Aim 2 will identify the recognition accuracy of the SSR with manipulated speech samples. The PIs hypothesize that the different manipulation strategies will yield comparable recognition accuracy results due to their experience with its ability to recognize severely distorted as well as non-distorted speech. The result of this will drive the future direction of the development team by identifying which manipulation strategy to focus on to produce functional speech practice targets to develop the SSR into a speech practice tool. Finally, Aim 3 will document the preliminary changes in speech performance across 15 trial therapy sessions to identify the potential role of SSR in long-term speech practice. The PIs hypothesize that individuals with moderate to severe dysarthria will demonstrate increases in speech intelligibility of target practice words and that their speech productions will change acoustically while using the SSR compared to their habitual productions. This study will utilize a three-level research design that includes molar (group analysis using a within subjects repeated measures ANOVA), intermediate, and molecular (individual detailed case studies) levels of analysis. Achievement of the aims of this project will lay the groundwork for the development of a computer practice program that automates the speech stimuli manipulations for functional speech practice.