Robust Speech Recognition Using Generative Adversarial Networks
This addresses the problem of speech recognition in noisy environments for users of AI systems, offering a data-driven approach that is incremental over existing methods.
The paper tackles robust speech recognition by using a generative adversarial network (GAN) framework to improve invariance in encoders, mapping noisy audio to clean audio embeddings, and shows improvements in simulated far-field speech recognition for sequence-to-sequence models without specialized front-ends.
This paper describes a general, scalable, end-to-end framework that uses the generative adversarial network (GAN) objective to enable robust speech recognition. Encoders trained with the proposed approach enjoy improved invariance by learning to map noisy audio to the same embedding space as that of clean audio. Unlike previous methods, the new framework does not rely on domain expertise or simplifying assumptions as are often needed in signal processing, and directly encourages robustness in a data-driven way. We show the new approach improves simulated far-field speech recognition of vanilla sequence-to-sequence models without specialized front-ends or preprocessing.