听力损失患者习惯性言语与清晰言语的感知自然度及声学特征

文献信息

中文摘要

目的： 本研究考察了清晰言语对感知自然度的影响。我们比较了习惯性言语（HS）、未经训练的清晰言语（UCS）以及经过简短训练后的清晰言语（训练后清晰言语[TCS]），并检验了声学变量在不同条件之间是否存在差异，以及这些变量是否能预测自然度。
方法： 十二名成年女性说话者分别在三种条件（HS、UCS、TCS）下朗读句子。二十八名听力正常的成年人以1-9分量表对自然度进行评分。除自然度外，还获取了语速、停顿次数和停顿时长的声学测量指标，并通过线性混合效应模型（LMMs）比较不同言语条件之间的差异。
结果： LMMs显示，言语条件对四个因变量均有显著影响（p < .001），除一项外，所有两两比较均显著。HS被评为最自然（3.56），且语速最快（3.89个音节/秒），而TCS最不自然（5.65），语速最慢（1.72个音节/秒），停顿次数最多且停顿时长最长。所有变量均显示，说话者在基线测量以及转向UCS和TCS时存在显著的个体间差异。言语自然度评分显示出重要的听者间差异。语速强烈预测感知自然度，停顿次数和停顿时长也与其相关。
结论： 与HS和UCS相比，30分钟的清晰言语训练降低了言语自然度。较慢的语速和增加的停顿与较差的自然度评分相关，尽管这些特征可能无法完全解释TCS自然度的降低。研究结果突显了可懂度与自然度之间的权衡：提高清晰度的策略可能使言语听起来不那么自然。临床方案可能受益于推广那些避免过度减慢或破坏韵律的清晰度增强技术，和/或在训练期间提供实时自然度和可懂度评分作为反馈。

英文摘要

PURPOSE: This study examined the impact of clear speech on perceived naturalness. We compared habitual speech (HS), untrained clear speech (UCS), and clear speech after a brief training (trained clear speech [TCS]) and examined whether acoustic variables differed between conditions and predicted naturalness.
METHOD: Twelve adult female talkers read sentences in three conditions (HS, UCS, TCS). Twenty-eight normal-hearing adults rated naturalness on a 1-9 scale. Along with naturalness, acoustic measures of speech rate, number of pauses, and pause duration were obtained and compared between speaking conditions via linear mixed-effects models (LMMs).
RESULTS: LMMs showed a significant effect of speaking condition on the four dependent variables (p < .001), and all pairwise comparisons, except one, were significant. HS was rated most natural (3.56) and was produced at the fastest rate (3.89 syllables/s), whereas TCS was least natural (5.65) and performed at the slowest rate (1.72 syllables/s), with the most and longest pauses. All variables revealed substantial between-talkers variability in regard to baseline measures and when shifting to UCS and TCS. Speech naturalness ratings showed important between-listeners variability. Speech rate strongly predicted perceived naturalness, and the number of pauses and pause duration were also associated with it.
CONCLUSIONS: A 30-min clear speech training reduced speech naturalness compared to HS and UCS. A slower speech rate and increased pausing were associated with inferior naturalness ratings, although these features may not fully explain the reduced naturalness of TCS. The findings highlight an intelligibility-naturalness trade-off: Strategies that improve clarity can make speech sound less natural. Clinical protocols may benefit from promoting clarity-enhancing techniques that avoid excessive slowing or disruption of prosody and/or from providing real-time naturalness and intelligibility ratings as feedback during training.