A Deep Learning Model for Personalized Human Activity Recognition

Human Activity Recognition (HAR) is a time series classification task that involves predicting the movement or action of a person based on sensor data. In the past the problem has been tackled by hand crafting features, which is time consuming and doesn’t generalize well. In this thesis we firstly analyze a Deep Learning model created for Time Series data that has set the state of the art in HAR. We then propose a customized version of the framework that is capable of adapting to a user.