Artificial

Intelligence

Lab

Education

• 2017-2019: Master Data and Knowledge Engineering (DKE), Otto-von-Guericke University Magdeburg, Germany
• since 2020: Researcher at the AI-Lab of the Otto-von-Guericke University Magdeburg, Germany

Main Research Interests

• Perceptually-Informed Deep Learning: Developing loss functions and model objectives grounded in human perception to enhance image and speech quality, preserve emotional expression, and maintain prosodic fidelity.

• Inclusive Speech Anonymization: Designing privacy-preserving voice conversion systems that retain prosody, emotion, and clinical relevance across diverse speaker populations.

• Ethical and Accessible Speech Technology: Building interpretable and domain-adaptive models for real-world applications in digital health, HCI, and voice accessibility.

Funded Projects

• During Emonymous (2022–2023): Developed an emotion-preserving speech anonymization framework that retains expressive and prosodic characteristics during voice conversion.

• During Medinym (2023–2024): Extended anonymization techniques to under-resourced and pathological speech using prosody-aware and data-efficient models. Additionally, proposed a semi-supervised method for segmenting pathological markers such as dysfluencies.

• In Anti-Stotter (2024–2025): Currently developing a solution for independent speech therapy.

Reproducible Research

Selected Publications

• Anonymizing Elderly and Pathological Speech: Voice Conversion Using DDSP and Query-by-Example
  

  Suhita Ghosh, Melanie Jouaiti, Arnab Das, Yamini Sinha, Tim Polzehl, Ingo Siegert and Sebastian Stober
  In: Interspeech 2024 - Kos, Greece, September 1-5, 2024

• Emo-StarGAN: A Semi-Supervised Any-to-Many Non-Parallel Emotion-Preserving Voice Conversion
  Suhita Ghosh, Arnab Das, Yamini Sinha, Ingo Siegert, Tim Polzehl and Sebastian Stober
  In: Interspeech 2023 - Dublin, Ireland, August 20-24, 2023

• Improving Voice Quality in Speech Anonymization With Just Perception-Informed Losses
  Suhita Ghosh, Tim Thiele, Frederick Lorbeer, Frank Dreyer and Sebastian Stober
  In: NeurIPS Workshop 2024 (Audio Imagination: Workshop on AI-Driven Speech, Music, and Sound Generation) - Vancouver, Canada, December 10-15, 2024 

• Uncertainty-aware temporal self-learning (UATS) - semi-supervised learning for segmentation of prostate zones and beyond
  Anneke Meyer, Suhita Ghosh, Daniel Schindele, Martin Schostak, Sebastian Stober, Christian Hansen and Marko Rak
  In: Artificial intelligence in medicine: AIM - Amsterdam [u.a.]: Elsevier Science - AIM, Bd. 116, 2021