Scholarlink

<p>Stress is a growing concern in modern society, significantly impacting health and productivity. Traditional stress assessment methods, such as self-reports and biochemical analysis, are limited by subjectivity and impracticality for continuous monitoring. To address these challenges, we propose a multi modal deep learning framework for stress severity classification using both facial and physiological data. Our approach introduces EffiFusion-Net model which integrates Efficient Net based convolutional neural network (CNN) feature extraction with a multi scale feature fusion to analyze stress-related facial features. Physiological signals, including heart rate and blood pressure, undergo normalization and statistical feature extraction to enhance model accuracy. A weighted soft-voting ensemble classifier, combining multi-layer perceptron (MLP), k-nearest neighbors (KNN), and random forest (RF), ensures robust stress level classification into five categories: No Stress, Low, Moderate, High, and Very High. Experimental results demonstrate the superiority of our model over conventional machine learning techniques, achieving an accuracy of 97.56% with 80% training data. This research highlights the potential of deep learning-based multi modal stress assessment for real-time monitoring and early intervention, paving the way for improved mental health management.</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

Stress Level Classification Using Multimodal Deep Learning and Physiological Signal Analysis

  • 期刊论文

  • 发表日期：

    2025年5月26日

查看全文
  • Zankhana Bhatt,
  • Ashwin Dobariya

摘要

Stress is a growing concern in modern society, significantly impacting health and productivity. Traditional stress assessment methods, such as self-reports and biochemical analysis, are limited by subjectivity and impracticality for continuous monitoring. To address these challenges, we propose a multi modal deep learning framework for stress severity classification using both facial and physiological data. Our approach introduces EffiFusion-Net model which integrates Efficient Net based convolutional neural network (CNN) feature extraction with a multi scale feature fusion to analyze stress-related facial features. Physiological signals, including heart rate and blood pressure, undergo normalization and statistical feature extraction to enhance model accuracy. A weighted soft-voting ensemble classifier, combining multi-layer perceptron (MLP), k-nearest neighbors (KNN), and random forest (RF), ensures robust stress level classification into five categories: No Stress, Low, Moderate, High, and Very High. Experimental results demonstrate the superiority of our model over conventional machine learning techniques, achieving an accuracy of 97.56% with 80% training data. This research highlights the potential of deep learning-based multi modal stress assessment for real-time monitoring and early intervention, paving the way for improved mental health management.