معرفی
Mostafa Sadeghi is a Research Fellow in the PERCEPTION team at Inria Grenoble Rhône-Alpes, where he has been conducting research since August 2018. His work bridges machine learning and signal processing with a strong emphasis on developing advanced speech enhancement techniques.
Dr. Sadeghi's primary research interests include:
- Unsupervised probabilistic generative models, particularly variational autoencoders (VAEs)
- Audio-visual speech enhancement and separation systems
- Diffusion-based generative models for speech processing
- 3D face alignment and frontalization techniques for visual speech processing
- End-to-end multichannel speech processing architectures
Analysis of his recent publications reveals a clear trajectory toward increasingly sophisticated generative models for speech processing. His work has evolved from traditional VAE approaches to cutting-edge diffusion models, with growing emphasis on multimodal (audio-visual) integration. A significant portion of his research addresses real-world challenges in conversational speech enhancement, meeting transcription, and multi-speaker environments, as evidenced by his contributions to the CHiME challenges.
Dr. Sadeghi maintains an active research profile with consistent publication output in top-tier venues including IEEE ICASSP, INTERSPEECH, and IEEE Transactions on Signal Processing. His collaborative network spans multiple institutions, with frequent co-authorship with researchers from INRIA and other European research centers.
