Mandarin Tone Identification in Cochlear Implant Users Using Exaggerated Pitch Contours
Mandarin Tone Identification in Cochlear Implant Users Using Exaggerated Pitch Contours
复制标题
DOI：
10.1097/mao.0000000000000980
复制
发表时间：
2016-04-01
影响因子：
2.1
通讯作者：
Limb, Charles J.
中科院分区：
文献类型：
--
作者：
He, Alice;Deroche, Mickael L.;Limb, Charles J.
目的：探讨夸大普通话基音频率（F0）轮廓的变化是否能提高人工耳蜗（CI）使用者的音调识别能力。方法：对12名听力正常（NH）的听者和11名CI使用者在4个或5个选项的强迫选择范式中识别普通话声调模型的F0轮廓的能力进行测试。研究人员使用了两种刺激：计算机生成的复杂音调和录音。用录音测试了四种轮廓：平坦、上升、下降和下降。第五个轮廓，峰，被添加为复杂的色调。每个轮廓的F0范围自适应变化。使用最大似然技术拟合心理测量函数与性能数据，并以70%的准确率提取阈值。结果：随着F0范围的增大，一些CI用户的色调识别性能有所提高，但并没有达到100%，这表明即使轮廓非常夸张，也会造成轮廓之间的混淆。与NH参与者相比，CI用户需要更大的F0范围来识别音调，在9.3半音和0.4半音之间。CI用户在复杂音调上的表现优于录音，而NH参与者的情况正好相反。混淆矩阵显示，当音调轮廓的F0范围太窄，参与者无法正确反应时，平淡的音调通常是默认选项。结论：这些结果表明CI使用者识别音调轮廓的能力明显受损，但表明使用夸张的音调轮廓可能有助于音调语言感知。
Objective:To determine whether exaggerating the variations in fundamental frequency (F0) contours of Mandarin-based pitch fluctuations could improve tone identification by cochlear implant (CI) users.Methods:Twelve normal-hearing (NH) listeners and 11 CI users were tested for their ability to recognize F0 contours modeled after Mandarin tones, in 4- or 5-alternatives forced-choice paradigms. Two types of stimuli were used: computer-generated complex tones and voice recordings. Four contours were tested with voice recordings: flat, rise, fall, and dip. A fifth contour, peak, was added for complex tones. The F0 range of each contour was varied in an adaptive manner. A maximum-likelihood technique was used to fit a psychometric function to the performance data and extract threshold at 70% accuracy.Results:As F0 range increased, performance in tone identification improved but did not reach 100% for some CI users, suggesting that confusions between contours could always be made even with extremely exaggerated contours. Compared with NH participants, CI users required substantially larger F0 ranges to identify tones, on the order of 9.3 versus 0.4 semitones. CI users achieved better performance for complex tones than for voice recordings, whereas the reverse was true for NH participants. Confusion matrices showed that the flat tone was often a default option when the tone contour's F0 range presented was too narrow for participants to respond correctly.Conclusion:These results demonstrate markedly impaired ability for CI users to identify tonal contours, but suggest that the use of exaggerated pitch contours may be helpful for tonal language perception.