基于人脸视频图像的心理状态评估研究  

Research on psychological state assessment based on video sequences

作　　者：韩昊欣 黄玮 毛吉勇 李汪洋 黄东 HAN Haoxin;HUANG Wei;MAO Jiyong;LI Wangyang;HUANG Dong(No.3 Cadet Regiment,School of Basic Medical Sciences,Air Force Medical University,Xi'an 710032,China;No.4 Cadet Regiment,School of Basic Medical Sciences,Air Force Medical University,Xi'an 710032,China;Department of Military Medical Information Technology,School of Military Biomedical Engineering,Air Force Medical University,Xi'an 710032,China;Shaanxi Provincial Key Laboratory of Bioelectromagnetic Detection and Intelligent Perception,Air Force Medical University,Xi'an 710032,China)

机构地区：[1]空军军医大学基础医学院学员三大队,陕西西安710032 [2]空军军医大学基础医学院学员四大队,陕西西安710032 [3]空军军医大学军事生物医学工程学系军事医学信息技术教研室,陕西西安710032 [4]空军军医大学生物电磁检测与智能感知陕西省重点实验室,陕西西安710032

出　　处：《空军军医大学学报》2024年第11期1222-1226,共5页Journal of Air Force Medical University

基　　金：陕西省自然科学基础研究计划项目(2023-JC-QN-0704)。

摘　　要：目的设计开发一种基于视频序列的心理状态评估研究的算法和系统,弥补传统评估过程的不足,为评估者提供客观公正的评估过程,保证评估方法不受时间地域的限制,最终提高评估过程效率。方法评估系统主要由人脸图像采集、心理状态评估、结果可视化分析三部分构成。人脸图像采集,负责抽取视频中的视频帧并利用已有的人脸分割算法将图像中的人脸区域进行分割;心理状态评估,由VGG负责提取面部特征,由长短期记忆网络学习视频帧之间时间关系,得到心理状态结果;结果可视化分析层,对最终结果进行时间回溯计算每帧图像对于最终结果的权重分配,并对权重结果进行可视化输出。结果以8∶2将数据库划分训练集和测试集,在测试集下具有82.7%的准确率,曲线下面积为0.89。结论在一定程度下可以实现个体心理状态评估,能够解决在目前社会条件下个体日益关注心理健康等问题,拓宽了现有心理状态评估途径,为所有需要心理评估人群提供便捷、客观、高效的新型评估方式。Objective To design and develop an algorithm and system for psychological state assessment research based on video sequences,so as to make up for the shortcomings of traditional evaluation process,provide an objective and fair evaluation process for evaluators,ensure that the evaluation method is not limited by time and region,and ultimately improve the efficiency of the evaluation process.Methods The evaluation system was mainly composed of the following three parts:facial image acquisition,psychological state assessment,and result visualization analysis.The facial image acquisition part was used to extract the video frames in the video and use the existing face segmentation algorithm to segment the face region in the image.In the psychological state assessment part,VGG was used to extract facial features,and the long short-term memory network was used to learn the temporal relationship between video frames to obtain psychological state results.The result visualization analysis part performed time backtracking of the final results in order to calculate the weight allocation of each frame image to the final results,and visualized the weight results.Results The database was divided into training set and test set at a ratio of 8∶2,with the accuracy of 82.7%and the area under the curve of 0.89.Conclusion To a certain extent,individual psychological state assessment can be realized by the proposed method,which can solve the problem of individuals'increasing concern about mental health under the current social conditions,broaden the existing psychological state assessment approach,and provide a convenient,objective and efficient new assessment method for people who need psychological assessment.

关 键 词：心理状态评估 长短期记忆网络 卷积神经网络 人脸检测 

分 类 号：R395[哲学宗教—心理学] TP391.41[医药卫生—医学心理学]