Detection of Generalized Tonic Clonic Seizures and Falls in Unconstraint Environment Using Smartphone Accelerometer
Open Access
- 4 March 2021
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 39432-39443
Abstract
The detection of Generalized Tonic Clonic Seizures (GTCS) and Falls is of utmost importance due to the increase in prevalence of epilepsy and Sudden Death in Epileptic Patients during CoVID-19 pandemic, and prevention of serious injuries in Fall risk groups such as elderly requiring continuous monitoring for disease management and assisted living etc. Monitoring of Activities of Daily Living (ADLs) can assist in the detection of symptoms and onset of neurological disorders such as Alzheimer’s, stroke, and epileptic seizures. With a host of embedded sensors, improved memory, enhanced processing capabilities and availability to masses, smartphones can be used for Human Activity Recognition (HAR) through continuous monitoring of ADLs. This paper presents a tri-axial accelerometer-based approach to detect and classify activities performed by individuals by applying machine learning algorithms including RF, J48, NB, LMT and SVM to movement data. Movement data is collected in real-time from the embedded accelerometer of a smartphone worn by individual on upper-left arm in unconstraint environment. It is pre-processed using time and frequency domain analysis and spatial domain features are computed. Supervised machine learning techniques are applied to classify ADLs into five classes based on the intensity of movements: Stationary, Light Ambulatory, Intense Ambulatory, GTCS and Falls. We also used training data from MyNeuroHealth dataset collected from 23 individuals including epilepsy patients. Based on gathered results, Random Forest outperforms other classifiers with classification accuracy of 99.6% for stationary, 81.5% for light ambulatory, 99.8% for intense ambulatory and GTCS, and 97.2% for Falls corresponding to training data of 14000 samples. To date, activity classification in our system has been implemented on cloud instead of mobile phone application as subjects are using smartphones with dissimilar software and hardware specifications for assisted living applications.Keywords
Funding Information
- Higher Education Commission Pakistan through the National Research Program for Universities (8355/NRPU/R&D/HEC/2017)
This publication has 36 references indexed in Scilit:
- Identification of abnormal movements with 3D accelerometer sensors for seizure recognitionJournal of Applied Logic, 2017
- Analysis of Optimal Sensor Positions for Activity Classification and Application on a Different Data Collection ScenarioSensors, 2017
- Performance Analysis of Smartphone-Sensor Behavior for Human Activity RecognitionIEEE Access, 2017
- Novel Hierarchical Fall Detection Algorithm Using a Multiphase Fall ModelSensors, 2017
- Human activity recognition with smartphone sensors using deep learning neural networksExpert Systems with Applications, 2016
- Who to target in sudden unexpected death in epilepsy prevention and how? Risk factors, biomarkers, and intervention study designsEpilepsia, 2016
- Seizure detection, seizure prediction, and closed-loop warning systems in epilepsyEpilepsy & Behavior, 2014
- Detecting Falls with Wearable Sensors Using Machine Learning TechniquesSensors, 2014
- Accelerometer-based on-body sensor localization for health and medical monitoring applicationsPervasive and Mobile Computing, 2011
- Personalization and user verification in wearable systems using biometric walking patternsPersonal and Ubiquitous Computing, 2011