We studied whether deep learning methods from machine learning are suited to automatically derive features from raw accelerometer data that assess fall risk.