Children's Speech Recognition through Discrete Token Enhancement

TL;DR

This paper explores the use of discrete speech tokens to improve children's speech recognition systems, addressing data scarcity and privacy issues while maintaining performance and reducing model complexity.

Contribution

It introduces a method for integrating discrete tokens into children's ASR systems and evaluates strategies for creating these labels and their generalization capabilities.

Findings

Discrete token ASR achieves nearly equivalent performance to traditional models.

Model complexity is reduced by approximately 83%.

Effective generalization to unseen domains and nativity datasets.

Abstract

Children's speech recognition is considered a low-resource task mainly due to the lack of publicly available data. There are several reasons for such data scarcity, including expensive data collection and annotation processes, and data privacy, among others. Transforming speech signals into discrete tokens that do not carry sensitive information but capture both linguistic and acoustic information could be a solution for privacy concerns. In this study, we investigate the integration of discrete speech tokens into children's speech recognition systems as input without significantly degrading the ASR performance. Additionally, we explored single-view and multi-view strategies for creating these discrete labels. Furthermore, we tested the models for generalization capabilities with unseen domain and nativity dataset. Results reveal that the discrete token ASR for children achieves nearly equivalent performance with an approximate 83% reduction in parameters.