Micro-expression recognition method and system based on kernelized bigroup sparse learning

Abstract

The invention discloses a micro-expression recognition method and system based on nuclearized bigroup sparse learning. The method mainly includes: (1) extracting two sets of different types of feature vectors from micro-expression data set samples respectively, and constructing corresponding features (2) Assign a weight to each eigenvector to construct a kernelized bigroup sparse learning model for learning the weight of each eigenvector; (3) Solve the kernelized bigroup sparse learning model to obtain each eigenvector (4) For the input test video, extract two sets of different types of features, and stitch together the feature vectors with weights higher than the threshold as the micro-expression feature vector; (5) Use the classifier to analyze the micro-expression feature vector Classification to get the micro-expression category. The invention uses the training samples in the micro-expression data set to learn the weight of each feature vector, and automatically selects the optimal feature vector from two sets of different features for micro-expression recognition, which can effectively improve the recognition accuracy.

Description

technical field [0001] The invention relates to the technical fields of emotion recognition and artificial intelligence, and in particular, to a micro-expression recognition method and system based on nuclearized bigroup sparse learning. Background technique [0002] Facial expression recognition technology is the basis for computers to understand human emotions, and it is also an effective way to explore intelligent human-computer interaction. The robot's exploration of the user's real emotions using interactive information is conducive to improving the harmony and friendliness of human-computer interaction. However, due to the influence of their own defense mechanisms, human beings often try to suppress or hide their true emotional states, that is, in life, in addition to ordinary facial expressions (that is, macro expressions), there is another kind of difficult to detect. Expression - micro-expression. Compared with macro-expressions, the most notable features of micro...

AI Technical Summary

Problems solved by technology

[0006] Purpose of the invention: Aiming at the problems of difficult feature selection and low recognition accuracy in existing micro-expression recognition methods, the purpose of the present invention is to provide a micro-expression recognition method and system based on kernelized double-group sparse learning, by constructing a kernel The dual-group sparse learning model uses the training samples in the micro-expression data set to learn the importance weight of each feature vector in two sets of different features, and automatically selects the sparse optimal feature vector from the two sets of different features for micro-expression recognition to improve the accuracy of micro-expression recognition

Images