CONSTRUCTION METHOD OF CLASSIFICATION MODEL FOR EMOTION RECOGNITION AND APPARATUS THEREOF

PURPOSE: A class model configuration method for emotion recognition and apparatus thereof are provided to classify a plurality of emotions from biometric data by using a binary system, thereby making the classification easier and improving classification performance. CONSTITUTION: A feature value extractor(110) extracts a multi-dimensional feature value for each emotion class from biometric data corresponding to emotion classes. A class grouper(120) groups the emotion classes into a first and second class groups. A Fisher space mapper(130) maps the first and second class groups to Fischer space. A binary classifier(140) repetitively cultivates the mapped first and second class groups through Adaboost algorithm, and classifies the groups using a binary system. A classification level verifier(150) individually verifies the classification level of the binary classification. [Reference numerals] (110) Feature value extractor; (120) Class grouper; (130) Space mapper of a fisher; (140) Binary classifier; (150) Classification level verifier; (160) Class group selecting unit; (170) Emotional classifier