Deep Residual Networks with Auditory Inspired Features for Robust Speech Recognition