基于深度学习的多维度智能心理测评系统设计  

Design of multidimensional intelligent psychological evaluation system based on deep learning

作　　者：陈倩[1] 刘涛[1] CHEN Qian;LIU Tao(Shangluo University,Shangluo,Shaanxi 726000,China)

机构地区：[1]商洛学院,陕西商洛726000

出　　处：《自动化与仪器仪表》2023年第7期143-146,150,共5页Automation & Instrumentation

基　　金：商洛学院2023年学生工作研究课题《基于“四要素论”的高校中外合作办学学生思想政治教育模式探索》(XSGZ2306)。

摘　　要：针对当前心理测评方式单一、主观性强等问题,构建了多维度智能心理测评系统,并借助CNN创建人脸表情识别模型和文本情绪识别模型以完成特征提取和表情识别。相较于VGG-16、VGG-19、ResNet50、MobileNet四种CNN框架,MobileNetV2框架在jaffe、CK+两种数据集的准确率更高,提升准确率数值为1%左右。隐藏层节点数和Batch数量最佳值分别约为100个和25个。CNN模型在短词和长词文本分类下的准确率、F1值、查准率、召回率均在95%左右,该模型具有良好的区分能力和较高的预测精度,可将其应用于心理疾病人群的辅助诊断。Aiming at the problems of single psychological evaluation method and strong subjectivity,a multi-dimensional intelligent psychological evaluation system is studied and constructed,and facial expression recognition model and text emotion recognition model are created with CNN to complete feature extraction and expression recognition.Compared with the four CNN frameworks VGG-16,VGG-19,ResNet50,and MobileNet,MobileNetV2 has a higher accuracy rate in jaffe and CK+datasets,with an accuracy improvement of about 1%.The optimal number of hidden layer nodes and the number of Batch are about 100 and 25,respectively.The accuracy,F1 value,precision and recall rate of CNN model under short word and long word text classification are all about 95%.This model has good discrimination ability and high prediction accuracy,which can be applied to the auxiliary diagnosis of people with mental diseases.

关 键 词：心理测评 CNN 文本情绪 人脸表情 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]