Muskaan Singh

Multimodal fusion in ORBIT achieves superior zero-shot performance in Alzheimer's detection across languages, revealing the power of language-invariant representations.

COBALT outperforms existing methods by integrating cough audio features with advanced spectral analysis, achieving unprecedented accuracy in tuberculosis screening.

Existing audio deepfake detectors falter in healthcare contexts, but PHOENIX-Mamba achieves over 97% accuracy in identifying codec-fakes across various clinical conditions.

Forget relying on labeled speech data: this work shows you can train multilingual speech emotion recognition systems by transferring knowledge from *non-verbal* vocalizations.