Temporal-Channel Modeling in Multi-head Self-Attention for Synthetic Speech Detection

TL;DR

This paper introduces a Temporal-Channel Modeling (TCM) module to enhance multi-head self-attention in Transformer-based synthetic speech detectors, significantly improving detection accuracy with minimal additional parameters.

Contribution

The paper proposes a novel TCM module that captures temporal-channel dependencies in MHSA, leading to improved synthetic speech detection performance.

Findings

TCM module outperforms state-of-the-art by 9.25% in EER on ASVspoof 2021.

Only 0.03M additional parameters are needed for the TCM module.

Utilizing both temporal and channel information yields the best detection results.

Abstract

Recent synthetic speech detectors leveraging the Transformer model have superior performance compared to the convolutional neural network counterparts. This improvement could be due to the powerful modeling ability of the multi-head self-attention (MHSA) in the Transformer model, which learns the temporal relationship of each input token. However, artifacts of synthetic speech can be located in specific regions of both frequency channels and temporal segments, while MHSA neglects this temporal-channel dependency of the input sequence. In this work, we proposed a Temporal-Channel Modeling (TCM) module to enhance MHSA's capability for capturing temporal-channel dependencies. Experimental results on the ASVspoof 2021 show that with only 0.03M additional parameters, the TCM module can outperform the state-of-the-art system by 9.25% in EER. Further ablation study reveals that utilizing both temporal and channel information yields the most improvement for detecting synthetic speech.