Method, device and equipment of expression recognition

Abstract

The embodiment of the invention discloses a method, device and equipment of expression recognition. Human face expressions are expressed more accurately according to probability distribution of a plurality of expressions. The expression recognition method comprises the steps that human face area images are extracted from obtained images which is to be recognized and includes human faces; low levelfeatures and high level features are extracted from the human face area images through deep convolutional neural network models, wherein the low level features comprise outlines of facial lines of the human face area images and intensity of each pixel point gray level, the high level features comprise forms of facial features of the human face area images; and the extracted features are combinedthrough the deep convolutional neural network models, the combined features are judged through decision rules obtained from training learning in the deep convolutional neural network models to obtainprobability of M kinds of expression, and the M is a positive integer greater than 1.

Description

technical field [0001] The present invention relates to the technical field of machine learning, in particular to an expression recognition method, device and equipment. Background technique [0002] Expression recognition refers to recognizing the expression of the recognized object in the image from a given image, so as to determine the psychological emotion of the recognized object, generally refers to face recognition. At present, the expression recognition process usually includes the following two steps: [0003] (1) Extract the features of the recognized object from the image through the feature extraction algorithm. Wherein, the feature extraction algorithm may be, for example, a scale-invariant feature transform (Scale-invariant feature transform, SIFT) algorithm or a local binary pattern (Local Binary Pattern, LBP) algorithm. [0004] (2) Input the extracted features into the classifier, and classify the expressions through the classifier. Wherein, the classifie...

AI Technical Summary

Problems solved by technology

[0005] It can be seen that the recognition results obtained in the current expression recognition process are based on classification, that is to say, the final expression of the recognized object can only be happy or sad, etc. However, the expressions presented on the human face are essentially very rich. For example, the expression of surprise may be a combination of happy and surprised expressions, so simply classifying the expressions is not enough to accurately express the facial expressions

Images