Scholarlink

The broad adoption and proliferation of the internet has led to issues such as Internet Addiction (IA). IA has become one of the main problems influencing the mental state and social well-being of an individual. The traditional way of diagnostics, which is based on standard questionnaires, has little use in many cases and sometimes even demonstrates inexpressive objectiveness and reliability. This paper aims at identifying IA by studying Electroencephalography (EEG) signal patterns and adopting Convolutional Neural Networks (CNN) model for effective interpretation. Moreover, to ensure the privacy of the data and for real-time diagnosis we extend the CNN model using Federated Learning (FL) and deploy the model on constrained edge devices. Findings show encouraging accuracy in IA diagnosis by a CNN model surpassing common classifiers. This is further strengthened by Raspberry Pi 4 deployment on edge devices which makes it possible for real-time diagnosis as well. Relying on edge computing reduces the overload in the central server which improves the response time and computation burden. The findings of this research will adjoin the groundwork for scalable and privacy-preserving diagnostic methods for IA, hence providing the basis for personalized interventions and prevention strategies in mental healthcare.

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

IAD-Fed: Internet Addiction Diagnosis from EEG Signals Using Federated Learning in Edge Environments

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • G. Anantha Krishnan,
  • John Paul Martin,
  • Christina Terese Joseph,
  • Tom Kurian

摘要

The broad adoption and proliferation of the internet has led to issues such as Internet Addiction (IA). IA has become one of the main problems influencing the mental state and social well-being of an individual. The traditional way of diagnostics, which is based on standard questionnaires, has little use in many cases and sometimes even demonstrates inexpressive objectiveness and reliability. This paper aims at identifying IA by studying Electroencephalography (EEG) signal patterns and adopting Convolutional Neural Networks (CNN) model for effective interpretation. Moreover, to ensure the privacy of the data and for real-time diagnosis we extend the CNN model using Federated Learning (FL) and deploy the model on constrained edge devices. Findings show encouraging accuracy in IA diagnosis by a CNN model surpassing common classifiers. This is further strengthened by Raspberry Pi 4 deployment on edge devices which makes it possible for real-time diagnosis as well. Relying on edge computing reduces the overload in the central server which improves the response time and computation burden. The findings of this research will adjoin the groundwork for scalable and privacy-preserving diagnostic methods for IA, hence providing the basis for personalized interventions and prevention strategies in mental healthcare.