A study on fluctuation of auditory information based on acoustic information deviations and its perception

本研究探讨了声学特征在时间上的偏差如何影响人们对音质(1)，尤其是唱歌时的声音，以及对声源方向(2)的感知。对于(1)，本研究首先通过从F0轮廓中去除每个F0波动，使用修改后的F0重新合成歌唱声音，并将这些声音呈现给受试者，来研究这些F0波动对歌唱声音感知的影响程度。结果表明，所有F0波动对声乐感知的影响都很大，其中超调的影响最大。这项研究还调查了哪些特征与“歌唱”有关。结果表明，“歌唱”与“振动”、“清脆”、“清晰”等心理声学特征密切相关，“振动”受以下特征影响；(a) F0和振幅包络中的4 - 6hz调制，(b)共振峰的频率和幅度波动具有相同的调制频率，(c)共振峰偏差的间隔和相位与F0偏差的间隔和相位相对应。对于(2)，本研究利用计算模型从夹带和同步的角度论证了单个神经元增强时间信息的机制。仿真的结果是，具有多输入配置的模型在低CFs下产生增强的同步和近乎完美的夹带。输入终端数量与输出尖峰的输入事件数量之间的关系是定量推导的，以保持增强的时间信息，而不是来自anf的输入的类似初级行为。结果表明，在AVCN的单个神经元上可能存在多个输入的放电机制。

This research discusses how acoustic feature deviations in time influence on perception of voice quality (1), especially in singing-voice, and of sound source direction (2).For (1), this study first investigate how much these F0 fluctuations influence singing-voice perception, by removing each F0 fluctuation from the F0 contours, by re-synthesizing the singing-voices using the modified F0s, and by presenting these sounds to subjects. The results show that the effects of all F0 fluctuations on singing-voice perception are large, and effect of overshoot is the largest. This study also investigates what features are related to "singingness." The results indicate "singingness" is strongly associated with "vibration," "ringing" and "clearness" as psychoacoustic features and "vibration" is affected by the following characteristics ; (a) 4 -6 Hz modulation in the F0 and amplitude envelope, (b) formants are fluctuated in frequency and amplitude with the same modulation frequency, and (c) intervals and phases of formant deviation are corresponding to those of F0 deviation.For (2), this study demonstrates the mechanism through which a single neuron enhanced temporal information from the viewpoint of entrainment as well as synchronization using a computational model.As a result of simulation, a model with multiple-input configurations yields enhanced synchronization and nearly perfect entrainment at low CFs.The relation between the number of input terminals and the number of input events to output a spike is derived quantitatively to maintain enhanced temporal information against the primary-like behavior of inputs from ANFs.The results suggest the possibility of a firing mechanism that had multiple inputs on a single neuron in the AVCN.