仿生人形機器人頭部設計與表情辨識

Abstract

本論文研究目標為設計一具有類似人臉外型的機器人頭部，且可以呈現不同的臉部表情，並利用視覺功能進行表情辨識。 在表情呈現部分，本研究利用釣魚線與釣魚網模擬人臉線性與非線性肌肉結構，並且設計眼淚機構使機器人頭具有哭的表情並可以掉下眼淚。在表情辨識部分，本研究利用少量的嘴巴與眉毛參數進行表情辨識，研究結果顯示可以利用少量的參數達成高辨識率。最後結合表情呈現與表情辨識，設計一些狀況讓人與機器人間產生互動。

Our research goal was to develop a face robot that has a human appearance and can display facial expressions. We also constructed facial expression recognition with vision in order to create human－robot interaction. To display facial expressions, we used fishing line and fishing net to produce a muscular structure similar to the linear and nonlinear muscles of the human face. We also designed a mechanism for producing tears and made the face robot cry like a human being. For facial expression recognition, we used a few mouth and eyebrow descriptors by which to identify facial expressions. The results proved that a few descriptors could be used to achieve a high recognition ability. Finally, we combined facial expression display with facial expression recognition, and designed some conditions for human－robot interactions. We thus produced a closer and more natural relationship between humans and robots.