Scholarlink

<p>Anxiety and depression disorders are increasingly common, necessitating methods for real-time assessment and early identification. This study investigates gait analysis as a potential indicator of mental health, using the Microsoft Kinect sensor to capture movement parameters. Fifty participants (26 males, 24 females) walked before the Kinect sensor, and their gait data were recorded. Before the walking task, participants completed the General Anxiety Disorder (GAD-7) and Beck Depression Inventory (BDI-II) scales to assess their anxiety and depression levels. Data augmentation used an additive Gaussian noise method to generate synthetic data and address dataset imbalances. Key kinematic features such as step length, step width, cadence, and eight additional gait parameters were extracted from the recorded data. Various machine learning techniques were applied to classify and estimate anxiety and depression levels, including Linear Discriminant Analysis (LDA), Naive Bayes, Multi-class Support Vector Machines (SVM) with a polynomial kernel, and a Deep Neural Network (DNN) for classification. Classification accuracy reached 61.67% for anxiety and 86.53% for depression. These results demonstrate the potential of gait analysis combined with machine learning in recognizing anxiety and depression, presenting a promising non-invasive method for mental health assessment.</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

Recognition of anxiety and depression using gait data recorded by the kinect sensor: a machine learning approach with data augmentation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月2日

查看全文
  • Milad Shoryabi,
  • Ahmad Hajipour,
  • Afshin Shoeibi,
  • Ali Foroutannia

摘要

Anxiety and depression disorders are increasingly common, necessitating methods for real-time assessment and early identification. This study investigates gait analysis as a potential indicator of mental health, using the Microsoft Kinect sensor to capture movement parameters. Fifty participants (26 males, 24 females) walked before the Kinect sensor, and their gait data were recorded. Before the walking task, participants completed the General Anxiety Disorder (GAD-7) and Beck Depression Inventory (BDI-II) scales to assess their anxiety and depression levels. Data augmentation used an additive Gaussian noise method to generate synthetic data and address dataset imbalances. Key kinematic features such as step length, step width, cadence, and eight additional gait parameters were extracted from the recorded data. Various machine learning techniques were applied to classify and estimate anxiety and depression levels, including Linear Discriminant Analysis (LDA), Naive Bayes, Multi-class Support Vector Machines (SVM) with a polynomial kernel, and a Deep Neural Network (DNN) for classification. Classification accuracy reached 61.67% for anxiety and 86.53% for depression. These results demonstrate the potential of gait analysis combined with machine learning in recognizing anxiety and depression, presenting a promising non-invasive method for mental health assessment.