Realtime Activity Recognition Using LSTM and Smartwatch Sensor Data

Abstract

The aim of this work is to gain knowledge about finding, providing, and classifying interaction and health data during the course of disease of people suffering from dementia. In the following, we present a prototype that records interaction data of dementia patients using the smartwatch “Apple Watch Series 6” and that uses a recurrent neural network to provide information about the respective activity in real time. Based on three very similar activities, we systematically compare the prediction accuracy of different sensors in combination with a recurrent neural network.