基于改进卷积神经网络的心理状态预警技术  

Mental state pre⁃warning technology based on improved convolutional neural network

作　　者：王克[1] WANG Ke(Xi’an Aeronautical Polytechnic Institute,Xi’an 710089,China)

机构地区：[1]西安航空职业技术学院,陕西西安710089

出　　处：《电子设计工程》2024年第10期49-53,共5页Electronic Design Engineering

基　　金：陕西省教育工委项目(2022XKT54)。

摘　　要：针对传统问卷法难以真实反映被调查者心理状态的问题,基于光流法和卷积神经网络提出了一种微表情判断方法,并将其作为心理状态预警技术的核心模块。对于数据集中人脸数据离散的问题,该方法采用人眼权重法对图像进行预处理,且通过金字塔光流算法提取预处理图像序列的光流特征,再利用三维卷积神经网络对该特征加以训练。与传统算法相比,所提方法在减少模型训练参数与运算时间的同时还具有更优的学习能力。实验测试结果表明,该算法在CASME数据集上的微表情识别准确率为89.2%,F1值为0.6751,均优于其他对比方法。由此证明,该算法可实现对人脸微表情的准确识别,进而为学生心理状态预警提供客观的数据支撑。Aiming at the defect that the traditional questionnaire method is difficult to truly reflect the psychological state of the respondents,a micro expression judgment technology based on optical flow method and convolution neural network algorithm is proposed,and takes it as the core module of the psychological state early warning algorithm.For the problem of discrete face data in the dataset,the human eye weight algorithm is used to preprocess the image.The optical flow characteristics of the preprocessed image sequence are extracted by the pyramid optical flow algorithm,and then the optical flow characteristics are trained by the three⁃dimensional convolution neural network.Compared with traditional methods,the proposed expression judgment technology has better learning ability while reducing model training parameters and operation time.The experimental test results show that the accuracy of this method in the recognition of microexpressions in CASME data set is 89.2%,and the F1 value is 0.6751,which is higher than other comparison methods,indicating that it can accurately recognize facial microexpressions and provide objective data support for students’psychological state early warning.

关 键 词：金字塔光流法 三维卷积神经网络 微表情识别 人脸识别 心理预警 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]