CHS: Small: Game for Cleft Speech Therapy

口腔面裂（即唇裂、唇腭裂和孤立性腭裂，以及罕见的正中、外侧[横向]和斜面裂）是最常见的先天性畸形之一。在美国，平均每天大约有20个婴儿出生时患有唇腭裂，每年大约有7500个。腭裂患者无法使用正常机制阻止气流通过鼻腔；因此，腭裂语音包含了从鼻子中漏出空气的声音，称为“鼻漏”。在控制语言学习的反复试验的世界里，孩子使用他/她唯一可用的工具来防止空气从鼻子里逸出；他/她在声门或喉部的水平上把它控制住。这种机制被正常的声音用来发“go”中的硬“g”音。孩子用这个“声门顿音”来代替他/她不能正常发出的各种声音。因此，腭裂语音成为以声门停顿和不适当的鼻漏为特征的声音集合；这些异常的发音模式通常被称为补偿性发音障碍（CAD），它们严重影响语音的可理解性。由于这个原因，矫正手术通常在10-12个月大的时候进行，目的是在孩子开始练习语言的时候提供一个更正常的解剖框架。然而，修复后的上颚继续受到腭裂典型的低于正常肌肉量和正常术后瘢痕组织僵硬的不同程度的损害。手术修复后大约一年，大多数儿童的腭功能自然改善到足以选择性地防止鼻漏的程度。矫正CAD的手术后语言治疗从两岁开始，通常持续多年。矫正唇腭裂对孩子未来独立生活和充分参与社会的能力非常重要。尽管有文献记载的好处，但对于语言病理学家来说，在早期成功的可能性最高的时候训练孩子正确的语言产生是一个挑战，因为年幼的孩子通常不太合作，有时不能完全理解他们被要求做什么，并且通常不愿意做没有回报的语言作业，通常是在缺乏经验的父母的指导下，他们无法评估细微的进步（或缺乏进步）。PI在这项研究中的目标是了解帮助腭裂矫正儿童产生正常语言的最佳策略，并通过儿童可以在家中使用的游戏来促进这一过程，而父母的帮助最少，同时允许将有关儿童进展的数据实时传递给语言病理学家。项目成果将特别使服务不足人口的儿童受益。一个唇裂语音语料库将有利于研究唇裂语音检测的语音识别算法，而新的语音引擎算法将有利于整个语言治疗。这项工作还将促进UCSC和UCD在以人为中心的健康和健康生活游戏方面的新研究重点的发展。该项目的一个主要科学贡献将是更深入地了解腭裂儿童的决定性特征以及这些特征与腭裂言语的语音和语音规则原因的关系。技术贡献将是程序生成的游戏和语音引擎，以支持矫正腭裂儿童的家庭和独立管理的语言治疗，以及一种算法，使纵向语音数据管理和分析能够在每个参与者以及跨参与者的干预历史中实时进行。方法学方面的贡献将包括，在设计语音识别系统以识别特定的兔唇发音错误时，测量错误率，以及一种涉及计算机科学家、工程师、发展心理学家、语言和语言病理学家、整形外科医生、矫正兔唇儿童及其父母的语言治疗游戏的参与式设计方法。

Orofacial clefts (i.e., cleft lip, cleft lip and palate, and isolated cleft palate, as well as the rare median, lateral [transversal], and oblique facial clefts) are among the most common congenital anomalies. Approximately 20 infants are born in the United States with orofacial clefts on an average day, or 7500 every year. With a cleft palate, one is unable to stop airflow through the nose using normal mechanisms; cleft palate speech therefore contains sounds with air leaking out of the nose, referred to as "nasal escape." In the world of trial-and-error that governs speech learning, the child uses the only tool s/he has available to keep air from escaping out of the nose; s/he holds it back at the level of the glottis or larynx. This mechanism is used by the normal voice to make a hard 'g' as in 'go.' The child uses this "glottal stop" as a substitute for a variety of sounds that s/he cannot create normally. Cleft palate speech thus becomes a collection of sounds characterized by glottal stops and inappropriate nasal escape; these anomalous articulation patterns are usually referred to as compensatory articulation disorder (CAD), and they severely affect speech intelligibility. For this reason, corrective surgery is commonly performed around 10-12 months of age, with the goal of providing a more normal anatomical framework by the time the child begins practicing speech. The repaired palate continues, however, to be variably impaired by the less-than-normal muscle bulk typical of cleft palates and by the stiffness of normal post-surgical scar tissue. Over perhaps one year following surgical repair, palatal function spontaneously improves to the point where in the majority of children it is adequate to selectively prevent nasal escape. Speech therapy after surgery to correct CAD begins at the age of two years and often continues for many years. Correcting cleft speech is important for the child's future ability to live independently and to participate fully in society. Despite the documented benefits, it is a challenge for speech pathologists to train children in proper speech production at an early age when the likelihood of success is highest, because young children are typically less cooperative, sometimes do not fully comprehend what they are being asked to do, and are often unwilling to do unrewarding speech homework, typically under the guidance of inexperienced parents who are unable to assess subtle progress (or lack thereof). The PI's goal in this research is to understand the best strategy for helping children with corrected cleft palate produce normalized speech, and to facilitate this process through games that children can use at home with minimal help from parents while allowing data relating to the child's progress to be delivered to speech pathologists in real time. Project outcomes will especially benefit children from underserved populations. A cleft speech corpus will benefit researchers working on speech recognition algorithms for cleft speech detection, and new speech engine algorithms will benefit speech therapy at large. The work will also spur development of a new research focus at UCSC and UCD in human-centered games for health and healthy living.A major scientific contribution of this project will be a deeper understanding of the determining characteristics of children with cleft palate and how these relate to the phonetic and phonological rule causes of cleft speech. The technological contributions will be the games and speech engines that are procedurally generated to support in-home and independently administered speech therapy for children with corrected cleft palate, and an algorithm that enables longitudinal voice data curation and analysis to be carried out in real time over the intervention history for every participant as well as across participants. Methodological contribution will include a measure of error rates when a speech recognition system is designed to pick up cleft-specific mispronunciations, and a method for conducting participatory design of games for speech therapy involving computer scientists, engineers, developmental psychologists, speech and language pathologists, plastic surgeons, and children with corrected cleft palate and their parents.