A robust and fast seizure detector for IoT edge

Md Abu Sayeed, Saraju P. Mohanty, Elias Kougianos, Venkata Prasanath Yanambaka, Hitten Zaveri

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

20 Scopus citations

Abstract

Epilepsy is a neurological disorder which has negative impact on human life quality. Epilepsy affects almost 1% of the world population necessitating a unified system for fast seizure detection as well as remote health monitoring to enhance the daily lives of the epilepsy patients. We envision a smart seizure detection framework in the edge of the Internet of Things (IoT) which is capable of detecting seizures as well as monitoring the patient's healthcare activity remotely. Detection of seizure is performed using the discrete wavelet transform, statistical feature extraction, and a naive Bayes (NB) classifier. The proposed system was implemented and validated using Simulink, ThingSpeak, and off-the-shelf microcontrollers. Experimental results show that the proposed system reduces latency by 44% compared to a cloud-IoT based system and reports a classification accuracy of 98.65%.

Original languageEnglish
Title of host publicationProceedings - 2018 IEEE 4th International Symposium on Smart Electronic Systems, iSES 2018
Pages156-160
Number of pages5
ISBN (Electronic)9781538691724
DOIs
StatePublished - 2 Jul 2018
Event4th IEEE International Symposium on Smart Electronic Systems, iSES 2018 - Hyderabad, India
Duration: 17 Dec 201819 Dec 2018

Publication series

NameProceedings - 2018 IEEE 4th International Symposium on Smart Electronic Systems, iSES 2018

Conference

Conference4th IEEE International Symposium on Smart Electronic Systems, iSES 2018
Country/TerritoryIndia
CityHyderabad
Period17/12/1819/12/18

Keywords

  • Electroencephalogram (EEG)
  • Epilepsy
  • Feature Extraction
  • IoT
  • Naive Bayes Classifier
  • Seizure Detection

Fingerprint

Dive into the research topics of 'A robust and fast seizure detector for IoT edge'. Together they form a unique fingerprint.

Cite this