Intermediate Fine-Tuning Using Imperfect Synthetic Speech for Improving Electrolaryngeal Speech Recognition

TL;DR

This paper introduces an intermediate fine-tuning method using imperfect synthetic speech to bridge the domain gap in electrolaryngeal speech recognition, significantly improving recognition accuracy.

Contribution

It proposes a novel intermediate fine-tuning step with synthetic speech to enhance ASR performance for electrolaryngeal speakers, addressing domain shift issues.

Findings

Achieved 6.1% improvement over baseline without synthetic data

Intermediate fine-tuning helps learn high-level features rather than low-level details

Effective despite the imperfect nature of synthetic speech

Abstract

Research on automatic speech recognition (ASR) systems for electrolaryngeal speakers has been relatively unexplored due to small datasets. When training data is lacking in ASR, a large-scale pretraining and fine tuning framework is often sufficient to achieve high recognition rates; however, in electrolaryngeal speech, the domain shift between the pretraining and fine-tuning data is too large to overcome, limiting the maximum improvement of recognition rates. To resolve this, we propose an intermediate fine-tuning step that uses imperfect synthetic speech to close the domain shift gap between the pretraining and target data. Despite the imperfect synthetic data, we show the effectiveness of this on electrolaryngeal speech datasets, with improvements of 6.1% over the baseline that did not use imperfect synthetic speech. Results show how the intermediate fine-tuning stage focuses on learning the high-level inherent features of the imperfect synthetic data rather than the low-level features such as intelligibility.