CUHK-EE Systems for the vTAD Challenge at NCMMSC 2025

TL;DR

This paper introduces advanced voice timbre attribute detection systems using WavLM-Large embeddings and Diff-Net variants, achieving high accuracy and EER in the NCMMSC 2025 challenge, with insights into factors affecting performance.

Contribution

The paper presents novel timbre detection systems leveraging WavLM-Large and Diff-Net architectures, demonstrating improved generalization and performance in speaker attribute tasks.

Findings

WavLM-Large+FFN system achieves 77.96% accuracy and 21.79% EER.

WavLM-Large+SE-ResFFN excels in 'Seen' setting with 94.42% accuracy.

Model complexity influences generalization and robustness in timbre detection.

Abstract

This paper presents the Voice Timbre Attribute Detection (vTAD) systems developed by the Digital Signal Processing & Speech Technology Laboratory (DSP&STL) of the Department of Electronic Engineering (EE) at The Chinese University of Hong Kong (CUHK) for the 20th National Conference on Human-Computer Speech Communication (NCMMSC 2025) vTAD Challenge. The proposed systems leverage WavLM-Large embeddings with attentive statistical pooling (ASTP) to extract robust speaker representations, followed by two variants of Diff-Net, i.e., Feed-Forward Neural Network (FFN) and Squeeze-and-Excitation-enhanced Residual FFN (SE-ResFFN), to compare timbre attribute intensities between utterance pairs. Experimental results demonstrate that the WavLM-Large+FFN system generalises better to unseen speakers, achieving 77.96% accuracy and 21.79% equal error rate (EER), while the WavLM-Large+SE-ResFFN model excels in the 'Seen' setting with 94.42% accuracy and 5.49% EER. These findings highlight a trade-off between model complexity and generalisation, and underscore the importance of architectural choices in fine-grained speaker modelling. Our analysis also reveals the impact of speaker identity, annotation subjectivity, and data imbalance on system performance, pointing to future directions for improving robustness and fairness in timbre attribute detection.