自閉症兒童對於社交行為與溝通,但也在感知方面有一些障礙。此篇的論文研究重點在於利用電腦模擬神經元的方式,重新展示出自閉症兒童的人臉辨識之感知障礙。本實驗一共有三個階段。第一階段,我們先利用了人臉辨識演算法,來擷取人臉影像。第二階段利用SURF 演算法來擷取特徵點。第三階段,我們利用所找出來的SURF 特徵點,當作我們類神經網路輸入層的信號。我們的類神經網路分別有兩種不同的感知場。一種是自閉症模型另一種為非自閉症模型。我們的兩種輸入信號分別是兩種人臉,讓類神經網路去做二元分類問題。透過效率指標的計算,我們可以看出自閉症模型以及非自閉症模型,在學習行為上有不同的結果。
Children with autism show a wide variety of impairments of social behavior and communication, but also in perception. Our focus is here on psycho-physical experiments that compare face recognition of both autistic and non-autistic children. In order to reproduce the results in our model we use our self-programmed face recognition system in 2 slightly different modes and are in this way capable to reproduce both impairments and special capabilities of autistic children in our modified system. We use a 3 stage system. In the first stage we use face detection algorithm to find the face image. In the second stage we use SURF to extract the feature information from the face image. In the third stage a neural network is used to do the classification task. We give an insight into our model and also would like to further discuss analogies between autism and current weaknesses of present day service robots.