Scholarlink

<p>Mental stress is on the rise in all professions due to the fast pace of modern life and the transition in employment from physical to cognitive work. Mental stress is increasingly a major source of work-related illness. A sedentary lifestyle and work environment require individuals to work for extended periods, which can lead to stress. Working under mental stress for long periods increases the risk of life-threatening conditions such as cardiovascular disease and mental health issues. Consequently, early detection of mental stress is critical for effective diagnosis and intervention by healthcare professionals to categorize individuals into distinct mental states such as ‘Normal’, ‘Mildly Stressed’ and ‘Highly Stressed’. In this study, a hybrid deep learning model combining Convolution neural network (CNN) with Multilayer perceptron (MLP) is proposed to classify mental stress levels based on physiological signal data. The dataset comprises a diverse set of physiological biomarkers correlated with different mental states. The algorithm begins by segmenting preprocessed EEG data and extracting time–frequency features using Short-Time Fourier Transform (STFT). These spectrograms are then processed by CNN layers to extract spatial patterns, followed by MLP layers for mental stress state classification. Experimental results demonstrates that the hybrid CNN-MLP model achieves superior classification accuracy of 99.45% for training and 99.99% for validation outperforming other conventional models such as Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN) and Transformers. Additionally, the proposed hybrid model exhibits an average precision, recall, and F1-score of 99% for identifying'Highly Stressed' and'Mildly Stressed' states, indicating a highly promising approach for mental health practitioners in early stress detection and diagnosis.</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

Physiological signal-based mental stress detection using hybrid deep learning models

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月23日

查看全文
  • Nandini Modi,
  • Yogesh Kumar,
  • Kapil Mehta,
  • Neelam Chaplot

摘要

Mental stress is on the rise in all professions due to the fast pace of modern life and the transition in employment from physical to cognitive work. Mental stress is increasingly a major source of work-related illness. A sedentary lifestyle and work environment require individuals to work for extended periods, which can lead to stress. Working under mental stress for long periods increases the risk of life-threatening conditions such as cardiovascular disease and mental health issues. Consequently, early detection of mental stress is critical for effective diagnosis and intervention by healthcare professionals to categorize individuals into distinct mental states such as ‘Normal’, ‘Mildly Stressed’ and ‘Highly Stressed’. In this study, a hybrid deep learning model combining Convolution neural network (CNN) with Multilayer perceptron (MLP) is proposed to classify mental stress levels based on physiological signal data. The dataset comprises a diverse set of physiological biomarkers correlated with different mental states. The algorithm begins by segmenting preprocessed EEG data and extracting time–frequency features using Short-Time Fourier Transform (STFT). These spectrograms are then processed by CNN layers to extract spatial patterns, followed by MLP layers for mental stress state classification. Experimental results demonstrates that the hybrid CNN-MLP model achieves superior classification accuracy of 99.45% for training and 99.99% for validation outperforming other conventional models such as Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN) and Transformers. Additionally, the proposed hybrid model exhibits an average precision, recall, and F1-score of 99% for identifying'Highly Stressed' and'Mildly Stressed' states, indicating a highly promising approach for mental health practitioners in early stress detection and diagnosis.