He Xiaomin, Raghavan Vinay S, Mesgarani Nima
Department of Electrical Engineering, Columbia University, New York, NY, United States.
Mortimer B. Zuckerman Mind Brain Behavior Institute, Columbia University, New York, NY, United States.
Imaging Neurosci (Camb). 2025 Aug 28;3. doi: 10.1162/IMAG.a.126. eCollection 2025.
Understanding speech in noise depends on several interacting factors, including the signal-to-noise ratio (SNR), speech intelligibility (SI), and attentional engagement. However, how these factors relate to selective neural speech tracking remains unclear. In this study, we recorded EEG and eye-tracking data while participants performed a selective listening task involving a target talker in the presence of a competing masker talker and background noise across a wide range of SNRs. Our results revealed a non-linear relationship, where neural tracking of the target speech first increased with SNR but then paradoxically decreased as SNR continued to improve. To explain this, we quantified SI behaviorally, estimated attentional effort (AE) using gaze velocity, and measured behavioral performance (BP) via a repeated-word detection task. Our analysis showed that neural tracking of the target speech increased with both SI and attentional engagement. However, when intelligibility reached ceiling levels, selective neural speech tracking decreased as AE declined. Statistical modeling indicated that SI and AE were reliable predictors of neural tracking, while SNR showed no independent contribution after accounting for these factors. Our results demonstrate that improved SNR influences selective neural speech tracking primarily by increasing SI and simultaneously reducing AE, which have opposing effects on neural tracking. These findings underscore the importance of jointly considering these factors in studies of speech perception in noise.
在噪声环境中理解语音取决于几个相互作用的因素，包括信噪比（SNR）、语音清晰度（SI）和注意力投入。然而，这些因素与选择性神经语音跟踪之间的关系仍不清楚。在本研究中，我们记录了脑电图（EEG）和眼动追踪数据，同时参与者执行了一项选择性听力任务，该任务涉及在存在竞争掩蔽说话者和背景噪声的情况下，在广泛的信噪比范围内的目标说话者。我们的结果揭示了一种非线性关系，即目标语音的神经跟踪首先随着信噪比的增加而增加，但随后随着信噪比的持续提高而出现反常下降。为了解释这一点，我们通过行为量化了语音清晰度，使用注视速度估计了注意力努力（AE），并通过重复单词检测任务测量了行为表现（BP）。我们的分析表明，目标语音的神经跟踪随着语音清晰度和注意力投入的增加而增加。然而，当清晰度达到上限水平时，随着注意力努力的下降，选择性神经语音跟踪也会减少。统计建模表明，语音清晰度和注意力努力是神经跟踪的可靠预测指标，而在考虑这些因素后，信噪比没有显示出独立的贡献。我们的结果表明，信噪比的提高主要通过增加语音清晰度和同时减少注意力努力来影响选择性神经语音跟踪，而这两者对神经跟踪具有相反的影响。这些发现强调了在噪声环境下语音感知研究中共同考虑这些因素的重要性。
