Frequency and Multi-Scale Selective Kernel Attention for Speaker Verification

TL;DR

This paper introduces a selective kernel attention mechanism for speaker verification models, enabling adaptive kernel size selection in convolutional layers to improve performance across multiple evaluation protocols.

Contribution

It proposes a novel SKA mechanism that adaptively chooses kernel sizes based on frequency and channel information, enhancing existing architectures like ECAPA-TDNN and Res2net.

Findings

Consistent performance improvements across three evaluation protocols.

Effective integration of SKA with ECAPA-TDNN and Res2net backbones.

Enhanced model adaptability and accuracy in speaker verification.

Abstract

The majority of recent state-of-the-art speaker verification architectures adopt multi-scale processing and frequency-channel attention mechanisms. Convolutional layers of these models typically have a fixed kernel size, e.g., 3 or 5. In this study, we further contribute to this line of research utilising a selective kernel attention (SKA) mechanism. The SKA mechanism allows each convolutional layer to adaptively select the kernel size in a data-driven fashion. It is based on an attention mechanism which exploits both frequency and channel domain. We first apply existing SKA module to our baseline. Then we propose two SKA variants where the first variant is applied in front of the ECAPA-TDNN model and the other is combined with the Res2net backbone block. Through extensive experiments, we demonstrate that our two proposed SKA variants consistently improves the performance and are complementary when tested on three different evaluation protocols.