Detection of Motor Seizures and Falls in Mobile Application using Machine Learning Classifiers

Abstract

We have developed a healthcare mobile application, for human activity recognition, monitoring of well-being and detection of individuals going towards a health hazard based on the data collected from sensors embedded in mobile phones and wearables. The data from sensors are processed within the mobile application to detect and classify different Activities of Daily Living. The developed framework is used to collect data in an unconstraint environment from individuals suffering from neurological disorders. The data is further tested using signal processing and machine learning algorithms. Results of in-app processing and classification are stored in a dedicated mobile database for later reference and analysis. This paper shows that statistical and Machine Learning methods can also be used within a mobile application for classification of ADLs. MyNeuroHealth has been designed in accordance with the scale of the prevalence of neurological disorders among the general population of developing countries and has become more relevant in COVID-19 pandemic as it offers real-time nonintrusive monitoring. Results show that MyNeuroHealth can detect and classify Motor Seizures and falls with an accuracy of 99%. The app is also able to detect if a patient had stumbled or fallen due to any reason and notifies caregiver accordingly.