DNN-based Acoustic Modeling for Robust Automatic Speech Recognition