Target Speech Extraction with Pre-Trained Self-Supervised Learning Models

article in a collection out of WoS and Scopus

English

Pre-trained self-supervised learning (SSL) models have achieved re- markable success in various speech tasks. However, their potential in target speech extraction (TSE) has not been fully exploited. TSE aims to extract the speech of a target speaker in a mixture guided by enrollment utterances. We exploit pre-trained SSL models for two purposes within a TSE framework, i.e., to process the input mixture and to derive speaker embeddings from the enrollment. In this paper, we focus on how to effectively use SSL models for TSE. We first in- troduce a novel TSE downstream task following the SUPERB princi- ples. This simple experiment shows the potential of SSL models for TSE, but extraction performance remains far behind the state-of-the- art. We then extend a powerful TSE architecture by incorporating two SSL-based modules: an Adaptive Input Enhancer (AIE) and a speaker encoder. Specifically, the proposed AIE utilizes intermedi- ate representations from the CNN encoder by adjusting the time res- olution of CNN encoder and transformer blocks through progressive upsampling, capturing both fine-grained and hierarchical features. Our method outperforms current TSE systems achieving a SI-SDR improvement of 14.0 dB on LibriMix. Moreover, we can further improve performance by 0.7 dB by fine-tuning the whole model in- cluding the SSL model parameters.

Target speech extraction, pre-trained models, self-supervised learning, feature aggregation

PENG, J.; DELCROIX, M.; OCHIAI, T.; PLCHOT, O.; ARAKI, S.; ČERNOCKÝ, J.

14. 4. 2024

IEEE Signal Processing Society

Seoul

979-8-3503-4485-1

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

10421

10425

5

https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10448315

peng_icassp2024_Target_Speech_Extraction_with_Pre-Trained.pdf