Dynamic Kernels and Channel Attention for Low Resource Speaker Verification

TL;DR

This paper introduces an attention-based dynamic kernel approach in convolutional neural networks to enhance speaker verification performance with limited data, achieving significant improvements over existing models.

Contribution

It proposes a novel dynamic kernel method with channel attention and feature aggregation, improving model resolution and representation capacity efficiently for low-resource speaker verification.

Findings

Achieved 1.62% EER and 0.18 miniDCF on VoxCeleb1

17% relative improvement over ECAPA-TDNN

Effective self-adaptation of model parameters to input features

Abstract

State-of-the-art speaker verification frameworks have typically focused on developing models with increasingly deeper (more layers) and wider (number of channels) models to improve their verification performance. Instead, this paper proposes an approach to increase the model resolution capability using attention-based dynamic kernels in a convolutional neural network to adapt the model parameters to be feature-conditioned. The attention weights on the kernels are further distilled by channel attention and multi-layer feature aggregation to learn global features from speech. This approach provides an efficient solution to improving representation capacity with lower data resources. This is due to the self-adaptation to inputs of the structures of the model parameters. The proposed dynamic convolutional model achieved 1.62\% EER and 0.18 miniDCF on the VoxCeleb1 test set and has a 17\% relative improvement compared to the ECAPA-TDNN using the same training resources.