arXiv subjects

Bingwen Feng

Publications and source records attributed to Bingwen Feng.

3 recordsLinked to original sources

arXiv2026

WST-Graph: Topology-Preserving Wavelet Scattering Front-End for Speech Deepfake Detection

The acoustic front-end determines which forensic cues a speech deepfake detector can exploit. The wavelet scattering transform (WST) provides stable multiscale coefficients with explicit coordinates, yet direct flattening obscures the parent relation between paths. We introduce WST-Graph, reconstructing these paths as a sparse modulation-carrier grid for an AASIST graph backend. Modulation-level normalization and length-aware adaptive local attention pooling produce fixed relative-time representations while retaining the acoustic axes before learned adaptation. This yields a waveform-to-graph interface with a fixed, parameter-free WST. Our configurations remain competitive with AASIST while using approximately 60% fewer trainable parameters and show clear gains on selected out-of-domain benchmarks. These results underscore the value of preserving parent-child relations within the carrier-modulation topology when constructing a compact, physically grounded interface for graph-based speech deepfake detection. Code will be released at https://github.com/saki-ciallo/wst-graph.

eess.AS↗

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↗

arXiv2021

Lattice-Based Minimum-Distortion Data Hiding

Lattices have been conceived as a powerful tool for data hiding. While conventional studies and applications focus on achieving the optimal robustness versus distortion tradeoff, in some applications such as data hiding in medical/physiological signals, the primary concern is to achieve a minimum amount of distortion to the cover signal. In this paper, we revisit the celebrated quantization index modulation (QIM) scheme and propose a minimum-distortion version of it, referred to as MD-QIM. The crux of MD-QIM is to move the data point to only the boundary of the Voronoi region of the lattice point indexed by a message, which suffices for subsequent correct decoding. At any fixed code rate, the scheme achieves the minimum amount of distortion by sacrificing the robustness to the additive white Gaussian noise (AWGN) attacks. Simulation results confirm that our scheme significantly outperforms QIM in terms of mean square error (MSE), peak signal to noise ratio (PSNR) and percentage residual difference (PRD).

cs.IT↗