Model Compression for DNN-based Speaker Verification Using Weight Quantization

TL;DR

This paper demonstrates that weight quantization effectively compresses DNN-based speaker verification models like ECAPA-TDNN and ResNet, significantly reducing size with minimal performance loss, especially in ResNet due to its smooth weight distribution.

Contribution

The study applies weight quantization to compress popular SV models and analyzes their robustness and knowledge retention, highlighting ResNet's superior robustness due to its weight distribution.

Findings

Model size reduced multiple times without performance degradation

ResNet shows more robust compression than ECAPA-TDNN

Quantized models retain most speaker-relevant knowledge

Abstract

DNN-based speaker verification (SV) models demonstrate significant performance at relatively high computation costs. Model compression can be applied to reduce the model size for lower resource consumption. The present study exploits weight quantization to compress two widely-used SV models, namely ECAPA-TDNN and ResNet. Experimental results on VoxCeleb show that weight quantization is effective for compressing SV models. The model size can be reduced multiple times without noticeable degradation in performance. Compression of ResNet shows more robust results than ECAPA-TDNN with lower-bitwidth quantization. Analysis of the layer weights suggests that the smooth weight distribution of ResNet may be related to its better robustness. The generalization ability of the quantized model is validated via a language-mismatched SV task. Furthermore, analysis by information probing reveals that the quantized models can retain most of the speaker-relevant knowledge learned by the original models.