Detection of Facial Expressions Using Mathematical Morphology and Genetic Algorithms

• Main Authors: YEH, KUAN-CHIEH, 葉冠捷
• Other Authors: CHEN, JONG-CHEN
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/s965jb

碩士 === 國立雲林科技大學 === 資訊管理系 === 106 === Based on an evolutionary learning system, call MORPH which is used to recognize alphabets, our goal is to develop an innovative method of facial expression recognition. We wish to develop a filter that can distinguish facial features and expressions through this evolutionary learning system. Therefore, we have to do the preprocessing, find out the facial feature (eyes, eyebrows, mouths and noses) based on this and collocate with seven different types of expressions from Jaffe dataset to do the evolutionary learning system. The first step is “Expand”, generate the morphology sequences to increase population and diversity. The second step is “Compose”, combine the morphology sequences to generate different filters. The third step is “Select”. Calculate the score from the filters and select the outstanding population. The forth step is “Copy”, copy the outstanding population. The fifth step is “Mutation”, mutate the copied population to develop diverse population. Finally, when we distinguish different expression features through evolutionary learning system, the recognition rate could be 76 percent, Although the accuracy is not very high, if we improve this learning system, I believe it can improve the recognition rate and apply it to life.