Scholarlink

<p>A huge growth in global competition and relentless lifestyle leads to mental stress that affects the physical and mental health of people. Stress makes the brain more sensitive and may lead to a variety of illnesses affecting a person's physical, mental, and behavioural health. Since electroencephalogram (EEG) signals are the most straightforward, portable, and economical way to capture brain signals, they are widely employed for mental stress detection and emotion identification. Over the last ten years, a several of deep learning (DL) based stress detection methods have been used. However, inadequate intra-class disparity, insufficient distinguishing traits, poor temporal representation, and inappropriate channel selection restrict the systems' efficacy. In order to improve temporal representation, feature distinctiveness, and intra-inter class disparity, this research describes the use of deep convolutional neural networks (DCNNs) with 2-D wavelet packet transform (WPT) for mental stress detection. Additionally, it employs enhanced Spider Monkey Optimization (ISMO) with two competitive learning strategies—replacement of weak member (RWM) and spiralizer elite learning (SEL)—to improve the diversity of SMO solutions, algorithmic convergence, and the algorithm's search space by striking a balance between exploration and exploitation. The ISMO-WPT-DCNN outperforms the prior state of the arts with an accuracy of 97.54%, recall of 0.99, precision of 0.95, and F1-score of 0.97 for stress detection on the public Database for Emotion Analysis using the Physiological&#xa0;Signals&#xa0;(DEAP).</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

Novel 2-D Wavelet-Based Spectral-Temporal Representation for Deep Learning Based Stress Detection Using EEG Signal

  • 期刊论文

  • 发表日期：

    2025年7月18日

查看全文
  • Dipali Dhake,
  • Yogesh Angal

摘要

A huge growth in global competition and relentless lifestyle leads to mental stress that affects the physical and mental health of people. Stress makes the brain more sensitive and may lead to a variety of illnesses affecting a person's physical, mental, and behavioural health. Since electroencephalogram (EEG) signals are the most straightforward, portable, and economical way to capture brain signals, they are widely employed for mental stress detection and emotion identification. Over the last ten years, a several of deep learning (DL) based stress detection methods have been used. However, inadequate intra-class disparity, insufficient distinguishing traits, poor temporal representation, and inappropriate channel selection restrict the systems' efficacy. In order to improve temporal representation, feature distinctiveness, and intra-inter class disparity, this research describes the use of deep convolutional neural networks (DCNNs) with 2-D wavelet packet transform (WPT) for mental stress detection. Additionally, it employs enhanced Spider Monkey Optimization (ISMO) with two competitive learning strategies—replacement of weak member (RWM) and spiralizer elite learning (SEL)—to improve the diversity of SMO solutions, algorithmic convergence, and the algorithm's search space by striking a balance between exploration and exploitation. The ISMO-WPT-DCNN outperforms the prior state of the arts with an accuracy of 97.54%, recall of 0.99, precision of 0.95, and F1-score of 0.97 for stress detection on the public Database for Emotion Analysis using the Physiological Signals (DEAP).