This paper presents a system for another input modality in a multimodal human-machine interaction scenario. In addition to other common input modalities, e.g. speech, we extract head gestures by image interpretation techniques based on machine learning algorithms to have a nonverbal and familiar way of interacting with the system. Our experimental evaluation proofs the capability of the presented approach to work in real-time and reliable.1 1 Motivation Multimodal communication ways are becoming more important for a robust and flexible human-machine interaction in everyday surroundings. Therefore, our objective is to introduce a communication channel providing a natural and intuitive way of simple nonverbal communication. It emulates a common way of showing agreement / disagreement via head gestures, like it is known from human-human dialogs. Due to its simplicity and omnipresence in every-day life, this contributes to making dialog systems more efficient. 2 Use-Cases As already mentio...