Emotion Estimation Algorithm of Facial Subtle Expressions Based on Intelligence
The subtle change of local face texture is the key visual cue to judge emotion.However,it is difficult to capture the subtle change of expressions.Therefore,an emotion estimation algorithm of facial subtle expressions based on intelligent perception was proposed.At first,the interference on the face image was removed by filtering.Then,the face region in the continuous image was extracted by the difference method.Moreover,Fisher linear transformation was used to determine the best projection direction of the RGB vector and distinguish local feature areas of the face,and thus to extract subtle expression features.Meanwhile,the variational auto-coder was introduced,and then the attention model was embedded into the variational auto-coder network.According to the network attention,the important char-acteristics of different emotions were inferred.After that,the unlabeled features were input into the variational encoder for training.At the same time,the extracted features were used as a feature extractor to intelligently perceive different types of facial expressions.Finally,the emotion estimation of facial subtle expressions was achieved.Simulation results show that the proposed algorithm can accurately extract the features of facial subtle expressions.The emotional accura-cy estimated by subtle expression is high as well.In addition,the recall and precision can reach 97%.
NETL
NSTL
万方数据
维普
