MMSE-Based Missing-Feature Reconstruction With Temporal Modeling for Robust Speech Recognition | IEEE Journals & Magazine | IEEE Xplore