A STUDY OF ENTRAINMENT IN VERBAL COMMUNICATION AND ITS APPLICATIONS TO HUMAN INTERFACE

在实际的面对面交流中，我们通过互相关分析评估了说话人的声音和眨眼与听者的眨眼和点头的同步性。我们发现：(A)说话者的声音和他自己的眨眼反应之间存在显著的同步和滞后关系；(B)说话者的声音和听者的眨眼和点头反应之间存在显著的同步和滞后关系。我们还发现，听者的眨眼反应往往发生在点头的时候。在此基础上，提出了一种语音输入的眨眼点头反馈模型，并验证了该模型的有效性。最后，对一个使用数字信号处理器的模型系统进行了测试，该系统接受声控输入并向演讲者提供反馈，目的是开发更方便用户的计算机系统。对信息的注意和理解主要通过模拟点头来表达，点头开始通过二进制声音信号的加权和来估计，而其余的交流信号则使用面部表情Alne来模拟。利用说话人的主观感官评价来衡量使用语音反应系统时语音输入的流畅度，发现该系统提高了语音输入的流畅度。这个模型将适用于各种人机交互，其中视觉反馈是可能的，并很可能导致一个更用户友好的系统。

We evaluated the synchrony between a speaker's voice and eye-blink, and a listener's eye-blink and nodding in actual face-to-face interaction by the cross-correlation analysis. We found significant synchronous, and lagged relationships betweem : (a) a speaker's voice and own his eye-blinking response ; (b) a speaker's voice and a listener's eye-blinking and nodding responses. We also found that the listener's eye blinking response tends to occur during nodding. On the basis of these findings, an eye-blinking and nodding feedback model for speech input was proposed, and the effectiveness was demonstrated. Finally, a model system using a digital signal processor which accepted voice-activated input and provided feedback to the speaker was tested with the aim of developing more user-friendly computer systems. Attentiveness to, and comprehension of, information were mainly expressed by simulated nodding, the nodding onset being estimated by a weighted sum of the binary vocal signal, while the remaining communicative signals were simulated using facial expressions alne. The smoothness of speech input when the voice-reactive system is used was measured using subjective sensory evaluation by the speaker, and the finding was that the smoothness of the speech input was improved by the system. This model would be applicable to a variety of human-computer interactions where visual feedback is possible and may well lead to a more user-friendly system.