Adversarial Learning of Raw Speech Features for Domain Invariant Speech Recognition

TL;DR

This paper demonstrates that adversarial training with Domain Adversarial Neural Networks can learn raw speech features that are invariant to domain shifts like gender and accent, improving speech recognition robustness.

Contribution

It introduces the application of DANNs to raw speech for domain-invariant feature learning, advancing unsupervised domain adaptation in ASR.

Findings

DANNs effectively learn domain-invariant features from raw speech.

Adversarial training improves ASR performance across different domains.

Promising results on datasets with gender and accent variations.

Abstract

Recent advances in neural network based acoustic modelling have shown significant improvements in automatic speech recognition (ASR) performance. In order for acoustic models to be able to handle large acoustic variability, large amounts of labeled data is necessary, which are often expensive to obtain. This paper explores the application of adversarial training to learn features from raw speech that are invariant to acoustic variability. This acoustic variability is referred to as a domain shift in this paper. The experimental study presented in this paper leverages the architecture of Domain Adversarial Neural Networks (DANNs) [1] which uses data from two different domains. The DANN is a Y-shaped network that consists of a multi-layer CNN feature extractor module that is common to a label (senone) classifier and a so-called domain classifier. The utility of DANNs is evaluated on multiple datasets with domain shifts caused due to differences in gender and speaker accents. Promising empirical results indicate the strength of adversarial training for unsupervised domain adaptation in ASR, thereby emphasizing the ability of DANNs to learn domain invariant features from raw speech.