This work proposes a closed-form ridge residualization of S3M features against external VoxLingua107 language embeddings, fitted using only healthy-control speech, which consistently improves cross-lingual PD-detection performance while correcting the high-specificity/low-sensitivity failure.
Abstract
Self-supervised speech models (S3Ms) provide powerful representations for Parkinson's disease (PD) detection, making cross-lingual transfer attractive for languages lacking labeled patient speech. However, these representations also encode language identity, which can confound this transfer: without target-language PD speech, classifiers may separate languages rather than pathology, yielding high specificity but low sensitivity on target patients. We propose \emph{language orthogonalization}, a closed-form ridge residualization of S3M features against external VoxLingua107 language embeddings, fitted using only healthy-control (HC) speech. By removing language-predictable components while retaining pathology-related variation, it produces a less language-dependent geometry in which HC representations concentrate while PD representations disperse. Across five S3M backbones, three speech tasks, and three target languages, our method consistently improves cross-lingual PD-detection performance while correcting the high-specificity/low-sensitivity failure.
A layer-wise analysis of nine SSL speech backbones using a low-capacity logistic regression probe reveals that the transferred discriminative signal lacks pathological specificity, highlighting critical limitations that must be addressed before speech-based pathology recognition models can be reliably deployed in clini...
Serli Kopar, Sam Gijsen, Abner Hernandez et al.· 0 citations
This work investigates the effectiveness of internal representations from encoder and decoder components of Speech LLMs for Parkinson's Disease detection across multilingual and cross-lingual settings and proposes a Squeeze-and-Excitation (SE)-based dynamic layer aggregation framework, which surpasses best-layer select...
Sarthak Giri, Z. Pang, Tatsuya Kawahara· arXiv.org· 0 citations
This work evaluates pre-trained embeddings from four state-of-the-art open-source SFMs across three datasets in zero-shot and k-shot cross-lingual settings for multi-class PD severity assessment and shows that pre-trained speech embeddings enable meaningful cross-lingual transfer, although performance is sensitive to d...
Simon Pals, Cristian Tejedor-García· arXiv.org· 0 citations
Audio deepfake detectors need to transfer to languages absent from training, as multilingual speech synthesis outpaces labeled anti-spoofing resources. While detectors increasingly rely on self-supervised speech models (S3Ms), these backbones encode language-dependent structure that confounds spoof cues. We address thi...
CL-DAF is presented, a Cross-Lingual Disease-Alignment Framework that identifies acoustic dimensions whose disease effects remain consistent across languages and provides a foundation for multilingual clinical speech models emphasizing pathology over language-dependent variation.
Roksana Khanom, Raghib Asfak Tasnim, Bodrun Nahar Bithi et al.· arXiv.org· 0 citations
It is shown that strengthening the model's ability to discriminate languages during pretraining reduces and, on some measures, closes this multilingual gap on continuous phonetic and higher-level linguistic measures, while preserving substantial cross-language sharing.
Maureen de Seyssel, Jie Chi, Zakaria Aldeneh· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.