從步態的性別辨識探討訊息來源

Gender identification from a point light display improves with age

目的：本研究用動態光點的方式，以年齡與人體發展的基礎，藉由步態辨識性別的方法，來探究人類辨識性別的訊息來源，並用以佐證訊息的恆定性。方法：實驗中以幼兒、兒童、成年的動作者，同時拍攝正面及側面的影片三次，且在拍攝時透過距離調整使呈現的影像人物大小相似，事後挑出影片處理結果較好的一次，經重覆一次並隨機排列後，請觀察者五十名做性別辨識。結果：一、無論正面或側面，動態影片的辨識率顯著高於靜態圖片的展現。二、不同面向與動作者年齡層達顯著效果，成年者的識別率顯著優於兒童與幼兒，且正面的識別率顯著優於側面。三、不論正面或側面，辨識得分與辨識一致性的相關係數皆達顯著。雖然，肩髖比例的結果未如預期，但成年男生的步幅與步頻大於女生且接近統計上顯著差異，推測可能是用以辨識性別的主要訊息；此外，就整體而言，男生在行走時兩手肘距離身體較遠，身體也較前傾，且確實達到性別上的顯著差異，而其他例如兩膝距離越大、手擺動的幅度越大、髖關節往前突出的距離越多，都會影響參與者將影片辨識為男生。結論：不同觀察面向皆可以辨識出性別，佐證了訊息的恆定性，而性別辨識的訊息來源除了身體肢段比例、步幅與步頻之外，一些社會化的因素亦會影響辨識結果，而人類可以從動態光點中辨識身分特徵的能力，應可以在人員的安全管制應用上有所幫助。Purpose: Examining the source of gender identification in the point light display based on the age and developmental principles. Methods: Fifty adult participants observed still pictures and dynamic videos of point light display showing human walking. The walkers in the videos consisted of 3 age groups: toddler, child, and adult. Both frontal and sagittal views of the walk were shown, each with 36 clips. Results: 1. Dynamic videos had significantly higher scores than the still pictures in identifying the walkers' gender. 2. Age and viewing direction both had significant effect on gender identification: the identification score was highest for the adult followed by the child, and then followed by the toddler; The identification score was also higher for the frontal view than the sagittal view. 3. The correlation between the identification score and consistency measure were significant in both frontal and sagittal view. Although the result of center of moment (Cm) measure did not agree with the prediction, there were other variables such as the stride length and pace in the adult group, the distance between the elbows of all age walkers, and the upper body forward angle showed some relation to the gender identification. In addition, wider distance between the knees, larger amplitude of arm swing and the forward position of the hip tended to lead to the identification as boy. Conclusion: The evidence that participants can identify gender from different viewing directions through a point light display supports the idea of the invariant property of information. The clue to gender identification via a point light display seem to have multiple sources which not only include the body segment, walking stride and pace, but also some social and cultural influence of the society. The ability of identifying different organism's characteristics from point-light display could be applied to entrance security.