For processing the recorded acceleration sensor data, we developed a framework which computes all necessary features such as the energy or entropy. The framework supports multi-threading and allows to specify several settings, e.g., concerning the windows. The framework is designed to operate as fast as possible and is also runnable on Android devices.
Java implementation of the online random forest classifier. Online machine learning enables to update an existing classification model continuously without keeping the processed data available or knowing all data a-priori. Supports threading and requires at least Java 1.7 (compatible with Android)