Biomedical Engineering, Parks College of Engineering, Aviation and Technology, Saint Louis University, 3507 Lindell Blvd, St Louis, MO, 63103, USA.
J Assoc Res Otolaryngol. 2021 Jun;22(3):349-363. doi: 10.1007/s10162-021-00787-2. Epub 2021 Apr 13.
This study explored the physiological response of the human brain to degraded speech syllables. The degradation was introduced using noise vocoding and/or background noise. The goal was to identify physiological features of auditory-evoked potentials (AEPs) that may explain speech intelligibility. Ten human subjects with normal hearing participated in syllable-detection tasks, while their AEPs were recorded with 32-channel electroencephalography. Subjects were presented with six syllables in the form of consonant-vowel-consonant or vowel-consonant-vowel. Noise vocoding with 22 or 4 frequency channels was applied to the syllables. When examining the peak heights in the AEPs (P1, N1, and P2), vocoding alone showed no consistent effect. P1 was not consistently reduced by background noise, N1 was sometimes reduced by noise, and P2 was almost always highly reduced. Two other physiological metrics were examined: (1) classification accuracy of the syllables based on AEPs, which indicated whether AEPs were distinguishable for different syllables, and (2) cross-condition correlation of AEPs (r) between the clean and degraded speech, which indicated the brain's ability to extract speech-related features and suppress response to noise. Both metrics decreased with degraded speech quality. We further tested if the two metrics can explain cross-subject variations in their behavioral performance. A significant correlation existed for r, as well as classification based on early AEPs, in the fronto-central areas. Because r indicates similarities between clean and degraded speech, our finding suggests that high speech intelligibility may be a result of the brain's ability to ignore noise in the sound carrier and/or background.
本研究探索了人类大脑对退化语音音节的生理反应。通过噪声声码化和/或背景噪声引入退化。目标是确定听觉诱发电位（AEPs）的生理特征，这些特征可能解释言语可懂度。10 名听力正常的人类受试者参与了音节检测任务，同时记录了他们的 32 通道脑电图的 AEPs。受试者以辅音-元音-辅音或元音-辅音-元音的形式呈现了六个音节。对音节应用了 22 或 4 个频率通道的噪声声码化。在检查 AEPs 中的峰值高度（P1、N1 和 P2）时，单独的声码化没有一致的效果。背景噪声没有一致地降低 P1，噪声有时会降低 N1，而 P2 几乎总是高度降低。还检查了另外两个生理指标：（1）基于 AEPs 的音节分类准确率，这表明 AEPs 是否可以区分不同的音节，以及（2）AEPs 之间的条件间相关性（r）在干净和退化语音之间，这表明大脑提取语音相关特征和抑制对噪声响应的能力。这两个指标都随语音质量的退化而降低。我们进一步测试了这两个指标是否可以解释其行为表现的跨受试者变化。在额中央区域，r 以及基于早期 AEPs 的分类都存在显著相关性。由于 r 表示干净和退化语音之间的相似性，我们的发现表明，高言语可懂度可能是大脑忽略声音载体和/或背景噪声的能力的结果。
