This master thesis presents usage of facial expression methods for the purpose of discovering food taste reaction. There are currently a few research articles where using existing software that detects six basic emotions and neutral expression, correlation between intensity of detected emotions and food taste reaction was investigated. In this theses we presented this problem as binary, so that a video shows either pleasant or unpleasant food taste reaction. We also trained our method on this data, while the existing research used pretrained data.
We separately tested response of reactions on food taste and drink taste. Because we could not obtain any database of food tasting, we prepared the database ourselves. We first tested some facial expression recognition methods on standard databases, that contian several different emotions. The most successful method was then used on our database, where it performed with good results. In the last chapter of thesis, we also presented and tested methods for automatic detection of face occlusion, which is the results of food intake. While performing test on our database we only used the part of video after the end of face occlusion.
|
