Shitong Xu

About Me

I am a third-year Ph.D. candidate in Computer Science at the University of Oxford, supervised by Prof. Niki Trigoni and Prof. Andrew Markham. I obtained my Master and Bachlor degree in Imperial College London, where I completed my Master thesis under the supervision of Prof. Ben Glocker.

Research Interest

During my PhD, my research focuses on developing deep learning models that integrate information across multiple modalities, including audio, text, and electroglottographic (EGG) signals, to understand real-world sound sources' spatial, temporal, and characteristic features. More broadly, my research interests include:

• Audio Signal Processing
• Multimodal Large Language Models
• Physics-Informed Machine Learning

In my previous projects, I explored encoding the target speaker's speech characteristics from noisy audio enrollment pairs, addressing the scarcity of clean enrollment audio in real-world target speaker extraction scenario. I also coauthored with labmates on spatial audio processing projects, including reverberation effect editing and sound source localization with distributed microphones. Building on these experience, I am now extending large audio-language models (LALMs) to perform complex audio processing tasks such as audio retrieval and spatial audio understanding.

I'm actively looking for internship and collaboration opportunities — feel free to reach out via email (shitong.xu [at] cs.ox.ac.uk).

Education

Internship