Face Expression is the visible feature mask to concede person's behaviour, intension or reaction. The paper has dealt out a qualified data mining model for expression recognition called MPMFFT (Multiple Pattern Multiple Feature based Feature Transformation) integrated model. This proposed model transformed the facial information in 22 aggregative features. These features include textural, geometrical, mathematical and structural characteristics. The characteristics are captured in the entire facial region and expression sensitive selected facial segment. To acquire the features, three local region extraction models are used on both broad and responsive facial areas. This complete featured composition formed a set of 132 features. This transformed data form is treated under the danger theory. At First, DCA (Dendric Cell algorithm) is ready to generate feature patterns to identify safe and danger qualified features. After this, each cell group is processed under DBT (Dempster Belief Theory) for integrated feature weight assignment based on mapping to the expression class. This weighted and cell formed data set is finally processed under probabilistic implementation using Bayesian nets. The experiment is applied on JAFFE, CMU and CK datasets. The experimental observations are compiled for individual expressions against decision tree, SVM and KNN methods. The comparative results signify that the model improved the accuracy of each of distinctive facial expression.
Juneja, K. (2017). MPMFFT based DCA-DBT integrated probabilistic model for face expression classification. Journal of King Saud University - Computer and Information Sciences. King Saud bin Abdulaziz University. https://doi.org/10.1016/j.jksuci.2017.10.006