利用臉部表情診斷學習困難度之研究

Abstract

人臉偵測與人臉辨識目前廣泛應用於門禁監視系統(surveillance)、智慧型電腦介面控制(HCI)、身分認證(BIA)、安全資訊存取(SIA)等相關應用上。然而人臉偵測又是人臉表情辨識與人臉追蹤前置處理之重要動作，因此探討利用臉部表情作為診斷學習困難度亦為一相當有趣的研究課題。 本文結合影像前處理、人臉偵測及人臉表情辨識技術，應用於診斷學習者的學習困難度。將WebCam所攝取的臉部影像，經過整合型濾波器抑制雜訊，再透過Sobel邊緣偵測取得影像邊緣特徵，接著利用邊緣強化加強影像輪廓;然後利用膚色偵測擷取眼睛和嘴巴的特徵，最後利用迴避度、專注度及愉悅度等向量進行困難度表情辨識，以自行拍攝40人，2040張之人臉資料庫的實驗結果發現學習困難度的辨識率可達92%。可見臉部表情對於診斷學習困難度已有可行性，將來可以應用在E-learning適性學習系統。

Facial detection and recognition have been utilized in areas such as surveillance system ,human controller interface(HCI), biometric identity authentification(BIA) and Security Information Access(SIA). However, Facial detection is an important preprocessing before facial expression recognition and facial tracing , Therefore ,it is an interesting research to design diagnosing the learning difficulty with facial expression recognition. In this study, we combined the image processing technical ,included image preprocessing ,Facial detection and facial expression recognition. This paper discusses three phase .In the first phase ,It captured the facial picture from a WebCam , Removing noise from the facial picture through an integrated filter while preserving and enhancing edges is one of the most fundamental operations of image processing. Then, the second phase used the color of skin method to extract the eyes and mouth characteristic, Finally, based on the facial emotion vector theories (withdrawal,careness,happiness) , we used it to diagnose the learning difficulty with facial expression recognition.In our experiment , we captured 2040 piece of facial images by 40 students utilizing WebCam for facial recognition database.The experiment result showed an correctness rate of 92 percent . The experiment proved that we can utilizing the proposed method for E-learning system in the further .