Performance of American Sign Language Templates Digital Image Recognition System Based on Support Vector Machine

• Main Authors: Lin, Deng-Jun, 林登鈞
• Other Authors: Huang, Tien-Yu
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/28361095873059996130

碩士 === 國立屏東教育大學 === 資訊科學系碩士班 === 102 === The purpose of this study is to design a gestures digital image recognition system based on Support Vector Machine, whose training and testing samples are American Sign Language Templates. This gesture Digital Image Recognition System is a kind of human computer interface which is used as a means of communication. The proposed system transfers RGB color space to YCbCr color space to get a digital image of the object in line with human skin color range, and uses morphology for hand objects to do noise filtering and crushing fill. Then the system removes excess arm features to reduce the amount of the excess information. Then the system selects the hand objects to do object center point angle adjustment, regularization, and normalizing hand object with our proposed algorithm in order to extract SVM feature vector. In order to extract the hand object feature vector, we use the feature vector as the basis for identification. We analyze 170 samples for American Sign Language. In this experimental test, we found the recognition accuracy can reach 98.18%. Experimental results reveal promising classification results for our proposed American Sign Language Templates Digital Image Recognition System based on Support Vector Machine.