A study on interaction between production and perception in speech communication

本研究采用听觉反馈范式，通过肌电图（EMG）信号、发音运动和频谱分析来测量语音产生器官的同步波动，研究语音感知和语音产生之间的相互作用，其中发音数据由电磁发音图（EMA）系统获得。选择汉语元音对[i]-[y]和日语元音对[e]-[a]、[e]-[i]和[e]-[u]作为实验对象。当说话者保持第一个元音时，通过操纵前三个共振峰，反馈声音从每对中的第一个元音随机改变到第二个元音。谱分析表明，元音的第一和第二共振峰存在明显的补偿。肌电图和EMA信号的分析也显示肌肉重新激活和舌头运动来补偿扰动。补偿响应的延迟时间约为150毫秒，从扰动开始的最大补偿时间约为290毫秒。根据测量，似乎在大多数情况下，说话者试图通过实时监控来补偿由听觉扰动引起的“误差”，并且听觉反馈通常在语音产生过程中同时发生。

This study employed an auditory feedback paradigm with perturbed fed-back speech to investigate interaction between speech perception and production by measuring simultaneous fluctuations of speech production organs using the electromyographic (EMG) signals, articulatory movements, as well as spectral analyses, where the articulatory data were obtained by the electromagnetic articulographic (EMA) system. Chinese vowels pair [i]-[y] and Japanese vowels pairs [e]-[a], [e]-[i] and [e]-[u] were chosen as the experimental objects. When the speaker is maintaining the first vowel, the feedback sound is randomly changed from the first vowel to the second one in each pair by manipulating the first three formants. Spectral analysis showed that a clear compensation was seen in the first and second formants of the vowels. Analyses of EMG and EMA signals also showed muscle reactivation and tongue movements to compensate for the perturbations. Latency of the compensating response is about 150 ms to start and about 290 ms for maximum compensation from the onset of the perturbation. According to the measurements, it seems that in most cases the speaker attempts to compensate for the "error" caused by the auditory perturbation by a real-time monitoring, and the auditory feedback takes place simultaneously often during speech production.