Scholarlink

<p>To achieve feature fusion of ballistocardiogram (BCG) signals for the purpose of predicting mental fatigue states, a prediction model based on convolutional neural network-long short-term memory (CNN-LSTM) architecture is built in this paper. The model combines the BCG signal Heart Rate Variability (HRV) parameter values with multiscale features to achieve accurate prediction of the BCG signal state. The results show that the fatigue classification proposed in this paper is highly consistent with the subjective evaluation, and all the evaluation indexes are better than the traditional methods. It achieved an accuracy of 95%, a precision of 96%, a linear regression of 0.9, and a G-mean of 0.93. In addition, it is able to automatically accelerate the learning of features without human intervention. There is no need to perform a tedious feature extraction process compared to traditional models. The accuracy and real-time performance of mental fatigue monitoring can be improved by using this model.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction of mental fatigue states based on the CNN-LSTM architecture for BCG signals

  • 期刊论文

  • 发表日期：

    2025年12月11日

查看全文
  • Liu Zhichao,
  • Liu Ziqi,
  • Li Xin

摘要

To achieve feature fusion of ballistocardiogram (BCG) signals for the purpose of predicting mental fatigue states, a prediction model based on convolutional neural network-long short-term memory (CNN-LSTM) architecture is built in this paper. The model combines the BCG signal Heart Rate Variability (HRV) parameter values with multiscale features to achieve accurate prediction of the BCG signal state. The results show that the fatigue classification proposed in this paper is highly consistent with the subjective evaluation, and all the evaluation indexes are better than the traditional methods. It achieved an accuracy of 95%, a precision of 96%, a linear regression of 0.9, and a G-mean of 0.93. In addition, it is able to automatically accelerate the learning of features without human intervention. There is no need to perform a tedious feature extraction process compared to traditional models. The accuracy and real-time performance of mental fatigue monitoring can be improved by using this model.