Improving Short Utterance Anti-Spoofing with AASIST2

TL;DR

This paper enhances speech anti-spoofing systems to better detect short utterances by modifying network architecture and training strategies, leading to improved performance across various durations.

Contribution

The paper introduces AASIST2 with Res2Net blocks and applies DCS and ALMFT strategies, significantly improving short utterance anti-spoofing performance.

Findings

AASIST2 outperforms previous models on short utterance detection.

Modified Res2Net blocks extract multi-scale features effectively.

Adaptive training strategies enhance robustness across datasets.

Abstract

The wav2vec 2.0 and integrated spectro-temporal graph attention network (AASIST) based countermeasure achieves great performance in speech anti-spoofing. However, current spoof speech detection systems have fixed training and evaluation durations, while the performance degrades significantly during short utterance evaluation. To solve this problem, AASIST can be improved to AASIST2 by modifying the residual blocks to Res2Net blocks. The modified Res2Net blocks can extract multi-scale features and improve the detection performance for speech of different durations, thus improving the short utterance evaluation performance. On the other hand, adaptive large margin fine-tuning (ALMFT) has achieved performance improvement in short utterance speaker verification. Therefore, we apply Dynamic Chunk Size (DCS) and ALMFT training strategies in speech anti-spoofing to further improve the performance of short utterance evaluation. Experiments demonstrate that the proposed AASIST2 improves the performance of short utterance evaluation while maintaining the performance of regular evaluation on different datasets.