基于面部动态数据捕获的心理自动测评系统设计

Design of psychological automatic evaluation system based on facial dynamic data capture

针对当代人工智能对面部动态细微变化捕获不足和测评置信度不佳的问题,提出了基于可控金字塔特征提取和时空双流网络模型的自动化心理测评系统。首先,构建心理自动测评系统整体框架;然后通过可控金字塔模型对采集到的面部动态数据进行特征提取,利用时空双流网络模型对进行情绪识别;最后对上述方法进行验证。实验评结果表明,可控金字塔特征提取网络数学模型可以在图像局部改变时保持捕捉数据完整性;时空双流网络模型能够提升系统测评的置信度,优化后的时空双流网络将测评置信度提升到0.9以上。最后通过对系统性能和功能的测评,验证本系统可成功反馈出用户的心理状态。

In view of the problems of insufficient capture of facial dynamics, a network model is proposed based on controllable pyramid feature extraction and spatial-temporal dual-stream network model.First the overall framewort of auto matic psychological evaluation system is constructed, then the spatial-temporal dual-stream network model is used to transform the gray-scale image and RGB image for emotion recognition, and finally the systematic evaluation experiment is conducted.The experimental evaluation results show that the mathematical model of controllable pyramid feature extraction network can maintain the capture data integrity when the image changes locally;the space-time dual-stream network model can improve the confidence of the system evaluation, and the optimized space-time dual-stream network is above 0.9.Finally, through the evaluation of the system performance and function, the verification of the system can successfully feedback the user’s psychological state.