MGFF-TDNN: A Multi-Granularity Feature Fusion TDNN Model with Depth-Wise Separable Module for Speaker Verification

TL;DR

This paper presents MGFF-TDNN, a novel speaker verification model that combines multi-granularity feature fusion with depth-wise separable convolutions to improve discriminative power and efficiency.

Contribution

The paper introduces MGFF-TDNN, integrating multi-granularity feature fusion with depth-wise separable modules for enhanced and efficient speaker verification.

Findings

Achieves superior performance on VoxCeleb dataset.

Maintains efficiency with fewer parameters and lower computational cost.

Effectively captures both global and fine-grained speaker features.

Abstract

In speaker verification, traditional models often emphasize modeling long-term contextual features to capture global speaker characteristics. However, this approach can neglect fine-grained voiceprint information, which contains highly discriminative features essential for robust speaker embeddings. This paper introduces a novel model architecture, termed MGFF-TDNN, based on multi-granularity feature fusion. The MGFF-TDNN leverages a two-dimensional depth-wise separable convolution module, enhanced with local feature modeling, as a front-end feature extractor to effectively capture time-frequency domain features. To achieve comprehensive multi-granularity feature fusion, we propose the M-TDNN structure, which integrates global contextual modeling with fine-grained feature extraction by combining time-delay neural networks and phoneme-level feature pooling. Experiments on the VoxCeleb dataset demonstrate that the MGFF-TDNN achieves outstanding performance in speaker verification while remaining efficient in terms of parameters and computational resources.