Emotion recognition method and device

Abstract

The present invention provides an emotion recognition method and device. The method comprises the steps of: collecting recognition object face images, then performing data processing, performing database storage, establishing an emotion distinguishing feature model, and obtaining a method of emotion output. The emotion recognition device comprises a recognition formed by a collection unit, a dataprocessing unit and a database unit. The emotion recognition method is high in differentiation degree, and the emotion feature distinguishing model is high in integrity and reliability, and an emotionresult obtained by the model is more accurate.

Description

technical field [0001] The invention relates to the field of image processing applications, in particular to an emotion recognition method and a device thereof. Background technique [0002] With the rapid development of artificial intelligence science, how to enable computers to recognize human expressions and obtain human emotional states has attracted more and more attention from computer science, psychology and other disciplines. [0003] At present, many emotional models have appeared in the field of emotional computing, but most of them are limited to the overall fusion training of images. The results of emotional models are not accurate enough to meet the judgment of emotional states under subtle expression states. [0004] The Chinese patent application number 201611014343.9 discloses a deep learning-based emotion recognition method and its device. The device includes a face image acquisition module, a deep learning-based expression recognition module, and an express...

AI Technical Summary

Problems solved by technology

By collecting facial images every time an employee checks in, the expression analysis algorithm of the technical deep learning algorithm is used to analyze the employee's emotion, and compare and analyze it with the historical emotion. When the emotion is abnormal, the device sends an alarm message to the relevant personnel; Obtain the frontal image of the employee's face and store the image in the image database according to the employee's ID and punch-in date. This method of fixed ID is inconvenient to use, and the image processing will adjust the face image to a size of 227*227 The RGB image of the image is classified through the trained VGGNet network. The image processing is relatively rough, and the micro-expressions cannot be accurately distinguished. The effect of the emotion classification is not ideal, and it is urgent to improve it.

Images