Björn W. Schuller

Imperceptible fingerprints prove to be far more reliable for speech deepfake attribution, even as perceptible ones fluctuate with emotional content and model changes.

PriMD revolutionizes multimodal emotion recognition by enabling robust performance even when critical data modalities are missing.

Naive application of transformer-based AI-text detectors can be brittle under distribution shift, but attention-based fusion of readability and vocabulary features can significantly improve robustness.

Forget coarse emotion categories: AffectSpeech unlocks richer, more nuanced speech emotion understanding and generation via fine-grained textual descriptions.

Control the emotional tone of generated speech without any training by directly manipulating specific neurons within large audio-language models.