A mobile application with machine learning techniques based on data from inertial sensors for the monitoring and identification of occurrences of Parkinson's tremor

Abstract

This article aims to show the results of the development of a mobile app that controls the monitoring and detection of occurrences of Parkinson's tremor. One of the specific objectives is the development of a hybrid application that uses machine learning techniques (Machine Learning) and data from inertial sensors present in smartphones. This app could enable the medical specialist to diagnose the degree of parkinsonian tremor (a product of Parkinson's disease), likewise the system allows keeping track of the results. On the one hand, the patient will have a report sent by email, it will also allow control of the movement freezing time, and an alarm system for the dosage of medications prescribed in the treatment. The implementation of the app is based on the Ionic framework and uses linear regression techniques for early identification from inertial sensor data. Among the results, we can observe the completion and viability in the development of the application, the initial considerations for its operation, taking into account variables, specifications, models, architecture and use cases that allow optimal operation and coupling with the other two analysis functionalities that are freezing (measuring the time of the person without movement) and alarm (medication dosage control).