arXiv subjects

Kwok-Ho Ng

Publications and source records attributed to Kwok-Ho Ng.

3 recordsLinked to original sources

arXiv2026

REIMU: Efficient Heterogeneous Hierarchical Reasoning for SSL-Based Speech Deepfake Detection

The increasing realism of speech generated by text-to-speech and voice conversion systems poses growing challenges to media integrity and voice authentication. Self-supervised learning (SSL) has substantially advanced speech deepfake detection, where downstream backbones conventionally process SSL representations through a single forward pass. This work investigates the practical effectiveness of recurrent hierarchical reasoning for this task. We term this controlled study REIMU and systematically compare conventional single-pass backbones, weight-shared recurrence, homogeneous HRM, and heterogeneous HRM across four Base-scale SSL frontends. We further examine heterogeneous high- and low-level modules that combine self-attention with linear attention. Experiments on the ASVspoof 2019 and 2021 evaluation sets show that recurrence and hierarchical decomposition do not inherently improve detection, whereas heterogeneous operator assignment provides a more competitive configuration. Notably, the heterogeneous design remains competitive while using 10.8\% fewer downstream parameters than the matched baseline, demonstrating its potential for parameter-efficient speech deepfake detection.

eess.AS↗

arXiv2026

XLSR-MamBo: Scaling the Hybrid Mamba-Attention Backbone for Audio Deepfake Detection

Advanced speech synthesis technologies have enabled highly realistic speech generation, posing security risks that motivate research into audio deepfake detection (ADD). While state space models (SSMs) offer linear complexity, pure causal SSMs architectures often struggle with the content-based retrieval required to capture global frequency-domain artifacts. To address this, we explore the scaling properties of hybrid architectures by proposing XLSR-MamBo, a modular framework integrating an XLSR front-end with synergistic Mamba-Attention backbones. We systematically evaluate four topological designs using advanced SSM variants, Mamba, Mamba2, Hydra, and Gated DeltaNet. Experimental results demonstrate that the MamBo-3-Hydra-N3 configuration achieves competitive performance compared to other state-of-the-art systems on the ASVspoof 2021 LA, DF, and In-the-Wild benchmarks. This performance benefits from Hydra's native bidirectional modeling, which captures holistic temporal dependencies more efficiently than the heuristic dual-branch strategies employed in prior works. Furthermore, evaluations on the DFADD dataset demonstrate robust generalization to unseen diffusion- and flow-matching-based synthesis methods. Crucially, our analysis reveals that scaling backbone depth effectively mitigates the performance variance and instability observed in shallower models. These results demonstrate the hybrid framework's ability to capture artifacts in spoofed speech signals, providing an effective method for ADD.

eess.AS↗

arXiv2025

Hybrid Quantum-Classical Convolutional Neural Networks for Image Classification in Multiple Color Spaces

The growing complexity and scale of image processing tasks challenge classical convolutional neural networks (CNNs) with high computational costs. Hybrid quantum-classical convolutional neural networks (HQCNNs) show potential to improve performance by accelerating processing speed, enhancing classification accuracy, and reducing model parameters, though studies have primarily focused on the RGB color space. However, the effectiveness of HQCNNs in non-RGB color spaces, such as Lab, YCrCb, and HSV, remains largely unexplored. We propose an HQCNN to evaluate image classification across diverse color spaces. The HQCNN integrates parameterized quantum circuits (PQCs) with a classical CNN, leveraging quantum entanglement and trainable gates to enhance expressiveness across varied color representations. We assess performance on MNIST, CIFAR-10, EuroSAT, and SAT-4 datasets. Experimental results demonstrate that the HQCNN outperforms the classical CNN across all tested color spaces for the ten-class MNIST task, achieving a best accuracy of $94.3\%$ in Lab compared to $92.8\%$ in RGB for the CNN, with superior performance on other datasets in various color spaces. These findings highlight the potential of non-RGB color spaces and optimized PQC designs to improve classification performance. We provide new insights for advancing hybrid quantum-classical computer vision through optimized PQC architectures and diverse color space applications.

quant-ph↗