Unsupervised Domain Adaptation by Adversarial Learning for Robust Speech Recognition

Abstract

In this paper, we investigate the use of adversarial learning for unsupervised adaptation to unseen recording conditions, more specifically, single microphone far-field speech. We adapt neural networks based acoustic models trained with close-talk clean speech to the new recording conditions using untranscribed adaptation data. Our experimental results on Italian SPEECON data set show that our proposed method achieves 19.8% relative word error rate (WER) reduction compared to the unadapted models. Furthermore, this adaptation method is beneficial even when performed on data from another language (i.e. French) giving 12.6% relative WER reduction.

arXiv comment5 pages, 2 figures, the 13th ITG conference on Speech Communication

References in corpus (7)

• Adam: A Method for Stochastic Optimization
• Achieving Human Parity in Conversational Speech Recognition
• Speaker-Invariant Training via Adversarial Learning
• Unsupervised Adaptation with Domain Separation Networks for Robust Speech Recognition
• Invariant Representations for Noisy Speech Recognition
• Domain Adversarial Training for Accented Speech Recognition
• Distilling Knowledge Using Parallel Data for Far-field Speech Recognition

Cited by in corpus (3)

• Adaptation Algorithms for Neural Network-Based Speech Recognition: An Overview
• Deep Representation Learning in Speech Processing: Challenges, Recent Advances, and Future Trends
• Unsupervised Speech Domain Adaptation Based on Disentangled Representation Learning for Robust Speech Recognition