A major challenge caused by the aging of the modern society is the home care of elderly people. Many elders prefer staying in their own home over moving into a nursing home for as long as possible. For this reason, the project DeinHaus 4.0 Lower Bavaria researches possible technical solutions which can help the elderly to live in their own homes for a longer period of time.The topic of falls is one main part of the research because the number of falls and thereby caused injuries increase with rising age. In this context, this work presents a prototype for an optical fall detection system based on pose estimation, running in real time on an embedded device connected to a USB-camera. In the process of development, different approaches including machine and deep learning, as well as a rule-based algorithm are applied to solve this task. All methods reach an accuracy higher than 94 % when used on publicly available data sets, but machine learning seems to have drawbacks when applied in the real-time application due to limitations in the available training data. The rule-based approach, on the other hand, seems to be more capable of generalizing different types of fall events.