上下半部臉部特徵對快樂與悲傷表情區辨之影響

Abstract

本研究主要是探討人類臉部表情與情緒知覺間的轉換機制，藉由了解臉部不同特徵(眼部及嘴部)對表情辨識的貢獻，來了解視覺系統如何整合臉部特徵的訊息，以建立一個統整的表情知覺。在實驗中，以台灣地區華人情緒資料庫 (Chen, Cho, & Tseng, 2013) 挑選使用快樂和悲傷的情緒臉孔，並在快樂與悲傷的圖像中插入不同融合程度的圖片。之後將影像從鼻尖切成兩半，分為上下兩部分，根據需要再將不同融合程度的上半部或下半部的影像結合。利用2AFC與PSI動態閾限搜尋法測量75%的正確反應閾值。結果顯示，不管在快樂或是悲傷表情中，上半臉固定在任何融合程度，改變下半臉的融合程度，區辨閾值幾乎相同，但在較高的下半臉融合程度時，區辨閾值會有些微上升的現象，且相對於固定下半臉改變上半臉的融合程度的結果，有較低的區辨閾值。另一方面，當固定下半臉改變上半臉的融合程度時，區辨閾值會隨著上半部表情融合程度的升高而有降低的趨勢，但在較高的上半臉融合程度時，區辨閾值有上升的現象。結果可以說明，下半部臉孔對於區辨快樂/悲傷表情時有較大的貢獻。進一歩經由數學模型適配所得到的反應函數可知，在表情判斷的機制中呈現一個非線性的結果。我們認為視覺系統可能存在有兩種機制分別為快樂偵測器與悲傷偵測器來處理不同的表情。

We investigated how the human visual system integrates facial features in order to discriminate subtle changes in faces and perceive facial expression. The stimuli used were morphed faces interpolated between happy and sad expressions of the same six models selected from a facial expression image database (Chen, Cho & Tseng, 2013). The morphed image was then divided into upper and lower parts. In each 2AFC trial, the observer was to compare a reference face with a designated morphing level and a test face with a slight change in morphing level from the reference in either the upper or the lower faces. We measured the discrimination threshold, the minimal morphing level difference for an observer to perceive a change in expression with 75% correct response level. When the upper face was held constant, there was little, if any, difference in discrimination threshold at different lower face morphing levels. The discrimination threshold was much lower than that in the conditions when the lower face was fixed. It suggests that the upper face plays little role in happy-sad discrimination. For each specific lower face, the threshold decreased with the morphing level of the upper reference face at first and then had slight increases. These findings suggest an expansive nonlinearity in the underlying expression discrimination mechanisms.