Identifying Communicative Signals in Dysarthric Speech

描述(申请人提供)：构音障碍，一种由大脑或周围神经损伤导致的言语障碍，导致缓慢、不准确和不协调的言语运动。对于许多人来说，演讲不是一种可行的交流方式。这对生活质量造成了毁灭性的后果。尽管构音障碍严重，但许多人在与熟悉的交流伙伴互动时继续使用发声，这表明构音障碍的语音信号中嵌入了可靠的信息。识别这些一致的声音信号有可能为许多人提供另一种交流渠道。以前的工作假设显著信息主要在语音段中编码。我们最近的研究表明，韵律特征，如基频(感知为基音)、强度(感知为响度)和音节时长也承载着重要的交际信息。我们假设，患有严重构音障碍的说话者对韵律有足够的控制，因为这些特征在较慢的时间尺度上比音素等分段单位变化得更慢，因此可能更容易产生。我们已经在孤立的元音和标记问句对比的短语中展示了对韵律的控制。为了超越以前的工作，我们将研究一个更复杂的任务，即更接近自然语音的局部音节级别的韵律控制。15名有严重构音障碍的说话者和15名无构音障碍的人将产生四个音节的短语，重音放在特定的音节上以调整短语的意思。为了确定有严重构音障碍的人是否能准确地发出重读信号，一组60名听者将判断每个录音中哪个音节重读(具体目标1)。将进行统计分析，以检查每个发言者的听者准确性，并比较不同发言者群体的结果。为了了解构音障碍的说话者如何标记对比重音，我们将提取和分析几个声学特征，包括基频、强度、持续时间、停顿和频谱倾斜(特定目标2)。将对每个特征进行单独的方差分析，以量化其在发出重音信号方面的重要性，并确定说话者之间的特征差异。最后，我们将比较严重构音障碍说话者和非构音障碍说话者的重音模式，以更好地了解与构音障碍信号应激相关的声学相关和补偿机制(特定目标3)。如果构音障碍的说话者能够可靠地利用韵律线索，我们的发现将指导关注韵律作用的干预策略的发展，以提高自然交流的有效性，并指导利用韵律作为交流信号的新型交流装置的设计。我们的最终目标是了解导致演讲者表现力提高的因素，并利用这一知识帮助个人更自然、更有效地进行交流，使他们能够更充分地参与教育、职业和社会活动。

DESCRIPTION (provided by applicant): Dysarthria, a speech disorder resulting from damage to the brain or peripheral nerves leads to slow, imprecise, and poorly coordinated speech movements. For many individuals, speech is not a viable communication option. This has devastating consequences on quality of life. Despite severe dysarthria, many individuals continue to use vocalizations when interacting with familiar communication partners suggesting that reliable information is embedded in the dysarthric speech signal. Identifying these consistent vocal signals has the potential to afford many individuals with an alternative channel of communication. Previous work has assumed that salient information is primarily encoded in speech sound segments. Our recent investigations have shown that prosodic features such as fundamental frequency (perceived as pitch), intensity (perceived as loudness), and syllable duration also carry significant communicative information. We hypothesize that speakers with severe dysarthria have sufficient control of prosody, given that these features vary over slower time scales than segmental units such as phonemes and thus may be easier to produce. We have demonstrated control of prosody in isolated vowels and in phrases to mark question-statement contrasts. To extend beyond previous work we will study a more complex task of local syllable-level control of prosody that better approximates natural speech. Fifteen speakers with severe dysarthria and fifteen non-impaired controls will produce four-syllable phrases with stress placed on specified syllables to modulate the meaning of the phrase. To determine whether speakers with severe dysarthria can accurately signal stress, a group of 60 listeners will judge which syllable was stressed in each recording (Specific Aim 1). Statistical analyses will be performed to examine listener accuracy for each speaker and to compare results across speaker groups. To understand how speakers with dysarthria mark contrastive stress, we will extract and analyze several acoustic features including fundamental frequency, intensity, duration, pause and spectral tilt (Specific Aim 2). Separate analyses of variance will performed on each feature to quantify its importance in signaling stress and to determine feature-wise differences between speakers. Last, we will compare stress patterning in speakers with severe dysarthria and non-impaired speakers to gain a better understanding of acoustic correlates and compensatory mechanisms associated with signaling stress in dysarthria (Specific Aim 3). If speakers with dysarthria can reliably exploit prosodic cues, our findings will guide the development of intervention strategies that focus on the role of prosody for improving the effectiveness of natural communication, and the design of novel communication devices that leverage prosody as a communicative signal. Our ultimate goal is to understand the factors that lead to improved speaker expressiveness, and to leverage this knowledge to help individuals communicate more naturally and effectively, enabling them to engage more fully in educational, vocational, and social activities.