Auxiliary Loss of Transformer with Residual Connection for End-to-End Speaker Diarization

TL;DR

This paper introduces RX-EEND, an improved transformer-based speaker diarization model that employs auxiliary losses and residual connections to enhance learning in lower encoder layers, significantly reducing diarization error rates.

Contribution

The paper proposes a novel residual auxiliary learning architecture for transformers in end-to-end speaker diarization, improving lower layer training and overall performance.

Findings

50.3% relative DER reduction on simulated data

21.0% relative DER reduction on CALLHOME dataset

Residual connections further improve DER by 8.1% on CALLHOME

Abstract

End-to-end neural diarization (EEND) with self-attention directly predicts speaker labels from inputs and enables the handling of overlapped speech. Although the EEND outperforms clustering-based speaker diarization (SD), it cannot be further improved by simply increasing the number of encoder blocks because the last encoder block is dominantly supervised compared with lower blocks. This paper proposes a new residual auxiliary EEND (RX-EEND) learning architecture for transformers to enforce the lower encoder blocks to learn more accurately. The auxiliary loss is applied to the output of each encoder block, including the last encoder block. The effect of auxiliary loss on the learning of the encoder blocks can be further increased by adding a residual connection between the encoder blocks of the EEND. Performance evaluation and ablation study reveal that the auxiliary loss in the proposed RX-EEND provides relative reductions in the diarization error rate (DER) by 50.3% and 21.0% on the simulated and CALLHOME (CH) datasets, respectively, compared with self-attentive EEND (SA-EEND). Furthermore, the residual connection used in RX-EEND further relatively reduces the DER by 8.1% for CH dataset.