Frequency domain binaural model as the front end of speech recognition system.