Unsupervised Neural Mask Estimator For Generalized Eigen-Value Beamforming Based Asr

The state-of-art methods for acoustic beamforming in multi-channel ASR is based on a neural mask estimator that attempts to learn the prediction of speech and noise using a paired corpus of clean and noisy recordings (teacher model). In this paper, we attempt to move away from the requirements of having a supervised clean recordings. The models based on signal enhancement and conventional beamforming methods serves as the required mask estimate. In this way, the model training can also be carried out on real recordings of noisy speech rather than simulated ones alone done in a teacher model. Several experiments performed on noisy and reverberant environments in the CHiME-3 corpus as well as the REVERB challenge corpus highlight the effectiveness of the proposed approach. The ASR results for the proposed approach provide performances that are significantly better than a teacher model trained on an out-of-domain dataset and on par with the oracle mask estimators in the in-domain dataset.