Effects of lowpass and highpass filtering on the intelligibility of speech based on temporal fine structure or envelope cues
Effects of lowpass and highpass filtering on the intelligibility of speech based on temporal fine structure or envelope cues
复制标题
DOI：
10.1016/j.heares.2009.12.002
复制
发表时间：
2010-02-01
期刊：
影响因子：
2.8
通讯作者：
Lorenzi, Christian
中科院分区：
文献类型：
--
作者：
Ardoint, Marine;Lorenzi, Christian
本研究旨在探讨言语中的时间包络（E）和精细结构（TFS）线索是否传递不同的语音信息。由男性和女性发言者发出的音节（i）处理，以保留在16个频带内的E或TFS，（ii）低通或高通滤波在不同的截止频率，和（iii）提出的识别7个听众。使用S形函数拟合心理测量函数，并用于确定交叉频率（低通和高通滤波产生等效性能的截止频率）和心理测量函数每个点的梯度（相对于截止频率的性能变化）。交叉频率和梯度没有显着不同的扬声器。E和TFS语音之间的交叉频率没有显著差异（类似于1.5 kHz）。在不同的滤波条件下，E和TFS语音的频率有显著差异。当刺激高通滤波高于2.5 kHz，性能显着高于机会水平和梯度显着不同，从0为E语音。这些研究结果表明，E和TFS传达重要的，但不同的语音线索之间的1和2 kHz。与TFS不同，E传递的信息高达6 kHz，这与E和TFS的神经锁相特性一致。(C)2009爱思唯尔有限公司版权所有。
This study aimed to assess whether or not temporal envelope (E) and fine structure (TFS) cues in speech convey distinct phonetic information. Syllables uttered by a male and female speaker were (i) processed to retain either E or TFS within 16 frequency bands, (ii) lowpass or highpass filtered at different cut-off frequencies, and (iii) presented for identification to seven listeners. Psychometric functions were fitted using a sigmoid function, and used to determine crossover frequencies (cut-off frequencies at which lowpass and highpass filtering yielded equivalent performance), and gradients at each point of the psychometric functions (change in performance with respect to cut-off frequency). Crossover frequencies and gradients were not significantly different across speakers. Crossover frequencies were not significantly different between E and TFS speech (similar to 1.5 kHz). Gradients were significantly different between E and TFS speech in various filtering conditions. When stimuli were highpass filtered above 2.5 kHz, performance was significantly above chance level and gradients were significantly different from 0 for E speech only. These findings suggest that E and TFS convey important but distinct phonetic cues between 1 and 2 kHz. Unlike TFS, E conveys information up to 6 kHz, consistent with the characteristics of neural phase locking to E and TFS. (C) 2009 Elsevier B.V. All rights reserved.