Multi-Stream ASR Trained With Heterogeneous Reverberant Environments

doi 10.1109/icassp.2001.940815