Optimization and prediction for fast and robust AAC

 在美国，超过250万人患有严重的言语障碍，影响了他们的沟通，其中包括25%的人报告说他们的自然语言无法被理解。无法有效沟通会对个人的生活质量产生严重影响，扰乱基本功能，如 从事有意义的工作，保持与家人和朋友的关系，甚至提醒照顾者紧急需要。这些人必须依赖某种形式的辅助和替代交流。不幸的是，很少有研究已经解决了AAC接口的优化，通过定量评价。因此，用户通常受限于缓慢且不灵活的语音输出，这不足以支持有效的通信。该研究的长期目标是开发和定量评估新的AAC技术，优化语音输出的速度和灵活性。如果成功的话，这项工作可以大大有利于那些严重瘫痪的人在实现足够的沟通功能方面遇到最大的困难。通信速率将通过两种方式减少用户选择目标所需的时间来提高。一是将组织好音素标的编排，提高选准率。三种不同的方法将进行评估：发音，发音功能，并通过一个迭代的，定量的过程，减少通常出现在一起的音素之间的距离。我们将测试的假设，训练后，定量安排的音素接口将提供更高的通信速率比音素接口安排的发音或发音功能，通过评估通信速率在健康的用户作为一个模型的严重瘫痪的用户。第二，不管目标排列如何，在序列中接下来可能被选择的目标将被动态放大，从而增加目标的大小并减少进行选择所花费的时间。我们将通过评估健康用户和严重瘫痪用户的通信速率来测试预测音素接口在训练期间和训练之后将提供比非预测音素接口更高的通信速率的假设。这些研究将促进AAC设备通信接口的开发，最终为AAC用户提供可靠和快速的通信，包括那些严重瘫痪的人，他们最需要改善通信。

 DESCRIPTION (provided by applicant): More than 2.5 million individuals in the U.S. have severe speech disorders that impair their communication, including the 25% of those who that report that their natural speech cannot be understood. The inability to communicate effectively can have drastic effects on an individual's quality of life, disrupting essential functions such as holding meaningful employment, maintaining relationships with family and friends, or even alerting caregivers to urgent needs. These individuals must rely on some form of augmentative and alternative communication (AAC). Unfortunately, little research has addressed the optimization of the AAC interface through quantitative evaluation. Consequently, users are often limited to slow and inflexible speech output that is insufficient to support effective communication. The long-term goal of the research proposed is to develop and quantitatively evaluate new AAC technology that optimizes speed and flexibility of speech output. If successful, this work could significantly benefit those severely paralyzed individuals experiencing the greatest difficulty in achieving adequate communicative function. Communication rates will be improved by reducing the time it takes users to select a target in two ways. First, the arrangement of phonemic targets will be organized to improve selection rates. Three different approaches will be evaluated: alphabetically, by articulatory features, and through an iterative, quantitative process that reduces the distance between phonemes that commonly occur together. We will test the hypothesis that after training, a quantitatively arranged phonemic interface will provide higher communication rates than phonemic interfaces arranged alphabetically or by articulatory features by evaluating communication rates in healthy users acting as a model of severely paralyzed users. Second, regardless of target arrangement, targets that are likely to be selected next in the sequence will be dynamically enlarged, thus increasing the size of the target and reducing the time it takes to make a selection. We will test the hypothesis that a predictive phonemic interface will provide higher communication rates than a non-predictive phonemic interface during and after training, by evaluating communication rates in both healthy users and severely paralyzed users. These studies will lead to advancements in the development of communication interfaces for AAC devices, eventually providing reliable and fast communication to AAC users, including those severely paralyzed individuals who demonstrate the greatest need for improved communication.