This paper proposes a speech enhancement method that has no restrictions with regard to sound-field conditions or signal characteristics. It is specially designed to model the Cocktail-Party-Effect, that is, to suppress interfering speech signals, which is the most challenging problem concerning speech enhancement. This important aim is achieved by involving the knowledge of binaural signal processing of the human auditory system. Combining tools that simulate binaural signal processing with a powerful, adapted noise cancelling algorithm leads to a new concept for a unique system. The individual steps of the processing scheme are described and promising, preliminary results are presented.