Facial Expression Recognition Method Based on Video Sequence

Abstract

The present invention is based on the human facial expression recognition method of video sequence, relates to the method for recognizing figure, is a kind of human facial expression recognition method that utilizes HCBP-TOP algorithm to extract the dynamic spatiotemporal texture feature of human facial expression sequence, and the steps are: human face Expression sequence preprocessing; use the spatial pyramid segmentation method to process the facial expression sequence images in layers and blocks; use the HCBP‑TOP algorithm to extract the dynamic spatiotemporal texture features of the facial expression sequence images; use the SVM classifier for facial expression training and predict. The method of the invention overcomes the defect that the prior art does not consider the central pixel, ignores local detail information, has low efficiency and recognition accuracy of facial expression recognition, and does not have universal applicability.

Description

technical field [0001] The technical solution of the present invention relates to a method for recognizing graphics, in particular to a method for recognizing human facial expressions based on video sequences. Background technique [0002] Expression is the most effective way in human emotional communication. In recent years, facial expression recognition systems have important applications in fields involving visual systems and pattern recognition, such as psychological research, video conferencing, emotional computing, and intelligent human-computer interaction. and the medical industry. With the comprehensive improvement of human-computer interaction technology, research on how to make computer systems more capable of perceiving human expressions is the focus of artificial intelligence at present, and the research and development of facial expression recognition systems is of great significance. [0003] Early facial expression recognition methods focused on studying fac...

AI Technical Summary

Problems solved by technology

However, the Gabor feature has the disadvantages of high computational complexity, high dimensionality, and susceptibility to light interference; Wang Yubo from Tsinghua University extracted the Haar-like features of face images in 2003, and then used the algorithm based on continuous Adaboost to classify facial expressions

Haar-like geometric features have some advantages of intuition, low dimensionality and strong description ability, but this method is sensitive to edge features and line features, and the feature extraction accuracy is not high. In addition, when the background environment of the image or video is complex, Adaboost classification device will produce a high false recognition

Liao from the University of North Carolina used the dominant local binary model (Dominant LBP, DLBP) and the Gabor method to extract features in 2009, and selected the main features of the LBP algorithm to make the operation faster, and combined the DLBP and Gabor methods Texture classification after feature extraction has achieved good results, but this method has two shortcomings: on the one hand, LBP does not take into account the influence of central pixels in the expression of image texture features; on the other hand, this method does not fully consider Due to the effect of time domain information, part of the information is lost, resulting in unsatisfactory recognition rate

LBP is widely used in the field of expression recognition due to its advantages of gray invariance and rotation invariance, but its disadvantage is that it is difficult to obtain a large spatial support area, and it is not robust to changes in illumination direction and viewing angle. , the performance is not satisfactory in terms of texture classification

Images