Description: Human speech as a means of communication between humans and machines is gaining significance. Nevertheless, automatic speech recognition in particular still does not show adequate performance in many cases. In order to explore routes to novel automatic speech recognition systems, we conduct fundamental research focusing on iterative decoding approaches. Beyond classical single-channel automatic speech recognition, we especially investigate multi-channel and multimodal - such as audio-visual - systems. The aim is to exploit as many available sources of information as possible, in order to obtain an improved recognition result.