Spatial Attention for Far-field Speech Recognition with Deep Beamforming Neural Networks
This is an incremental improvement for far-field automatic speech recognition systems.
The paper tackled the problem of redundant information in multi-direction neural beamformers for far-field speech recognition by introducing a spatial attention subnet to weigh features from different directions, resulting in up to 9% relative word error rate improvement.
In this paper, we introduce spatial attention for refining the information in multi-direction neural beamformer for far-field automatic speech recognition. Previous approaches of neural beamformers with multiple look directions, such as the factored complex linear projection, have shown promising results. However, the features extracted by such methods contain redundant information, as only the direction of the target speech is relevant. We propose using a spatial attention subnet to weigh the features from different directions, so that the subsequent acoustic model could focus on the most relevant features for the speech recognition. Our experimental results show that spatial attention achieves up to 9% relative word error rate improvement over methods without the attention.