Exploring Effective Fusion Algorithms for Speech Based Self-Supervised Learning Models

TL;DR

This paper investigates various fusion algorithms to combine speech-based self-supervised learning models, specifically Hubert and Data2vec, to enhance performance on speaker identification and speech recognition tasks.

Contribution

It introduces and compares new fusion algorithms for combining SSL models, demonstrating their effectiveness in improving downstream task performance.

Findings

Fusion algorithms significantly boost model performance

Combining Hubert and Data2vec improves SID and ASR results

Proposed methods outperform individual SSL models

Abstract

Self-supervised learning (SSL) has achieved great success in various areas including speech processing. Recently, it is proven that speech based SSL models are able to extract superior universal representations on a range of downstream tasks compared to traditional hand-craft feature (e.g. FBank, MFCC) in the SUPERB benchmark. However, different types of SSL models might exhibit distinct strengths on different downstream tasks. In order to better utilize the potential power of SSL models, in this work, we explore the effective fusion on multiple SSL models. A series of model fusion algorithms are investigated and compared by combining two types of SSL models, Hubert and Data2vec, on two representative tasks from SUPERB benchmark, which are speaker identification (SID) and automatic speech recognition (ASR) tasks. The experimental results demonstrate that our proposed fusion algorithms can further boost the individual model significantly.