基于多模态数据挖掘与LSTM网络计算的心理压力分类模型

Psychological stress classification model based on multimodal data mining and LSTM network calculation

为提高心理健康评估的准确率,提出一种基于多通道采集和LSTM的心理压力检测方法。其中,首先使用红外脉搏传感器实现脉搏、皮肤电阻等信号的采集;利用采集到的数据分别构建1D-cnn和LSTM的心理分类检测模型;考虑到传统对上述的特征模型进行融合;实验中采用的训练集为PPG和EDA生理信号小波分解之后的小波系数。实验结果表明,相较于原始模型,提出的基于多通道采集和LSTM的心理压力检测方法的准确性明显更高,通过融合不同生理信号的方式改善了精度,从而体现出更大的应用潜力。

To improve the accuracy of mental health assessment,a psychological stress detection method based on multi-channel acquisition and LSTM is proposed.Furthermore,the infrared pulse sensor is used to collect the signals of pulse and skin resistance,and the collected data is used to construct the psychological classification detection model of 1D-cnn and LSTM.Considering the traditional feature models,the training set used in the experiment is the wavelet coefficient after the decomposition of PPG and EDA.The experimental results show that compared with the original model,the proposed psychological stress detection method based on multiple channel acquisition and LSTM is significantly higher,which improves the accuracy by incorporating different physiological signals,thus reflecting greater potential for application.