基于多模态融合的心理测试仪自动评估系统检测研究

Study on the automatic evaluation system of psychological tester based on multimodal fusion

针对传统多模态融合模型对心理测评结果准确度不高的问题,提出一种基于注意力机制的心理测试系统。其中,针对采集到的多模态特征信息,包括面部和步态、语音信息等,通过注意力机制对特征权重进行分配,从而实现特征的融合,然后基于以上特征进行SVM识别,并比较融合前后的准确率。结果表明,通过注意力机制进行多模态特征融合后的模型,整体心理测评准确率达82.36%,提升6.45%,测评准确度较高;且针对测评难点人际关系敏感度指标,多模态特征融合后其测评准确率从66.19%提高至78.38%。由此得出,注意力机制的引入可弥补传统多模态融合测评的不足之处,提升改进的融合模型系统的心理测评准确性。

The attention mechanism for the problem of the traditional multimodal fusion model is introduced, and the multimodal fusion improvement model is proposed based on the attention mechanism.This paper focuses on testing the improved model and compares the experimental data with the original evaluation model.The experimental results show that the improved fusion model can increase the accuracy of psychological assessment by 6.45% over the traditional model, the accuracy of 82.36% and the accuracy of the improved model, from 66.19% to 78.38%.The experiment fully proves that the introduction of attention mechanism can make up for the shortcomings of the traditional multimodal fusion evaluation system, and significantly improve the psychological evaluation accuracy of the improved fusion model system.