Machine learning for audio with applications for health and wellbeing

The studentship’s central theme is Machine Learning for Audio. New AI for sound technology has major potential applications in security, health and wellbeing, environmental sensing, urban living, and the creative sector. Realising the potential of computational analysis of sounds presents particular challenges for machine learning technologies. For example, current research use cases are often unrealistic; modern AI methods, such as deep learning, can produce promising results, but are still poorly understood, and current datasets may have unreliable or missing labels. There is therefore great potential for advancing AI through research contributions in this area.

Solutions you will develop may involve, but not be limited to the following research themes:

• Bayesian Deep Learning:
  • Monte-Carlo dropout inference
  • HMC (Hamiltonian Monte Carlo)
  • Hybrid approaches
• Big Data analysis
• Sound Event Detection
• Signal Processing

Skills you will develop, which are sought after in further research in academia and industry, include:

• Time-series analysis
• Critical literature analysis and implementation
• Advanced data visualisation and presentation
• Reproducible research (RR)
• Open source contributions via GitHub
• Python programming:
  • Scientific computing in Scipy/Scikit-learn
  • Deep Learning with Keras/Tensorflow 2.0, and/or PyTorch
  • CUDA configuration and high-performance computing
  • (Optional) web communication with Apache/Django 
• Database languages:
  • MySQL/PostgreSQL
  • MongoDB
• Unix programming

Example up-and-running projects you may join to rapidly build up your research expertise and portfolio (including publications) include: 

• As part of CVSSP’s AI4Sound: 
  • The HumBug project with the University of Oxford, Imperial College London, and entomologists worldwide: acoustic mosquito monitoring for malaria vector mapping and intervention 
  • In the Surrey Clinical Research Building: analysing the role of audio and multimodal data on the effect of sleep; using audio to categorise sleep quality, and thus determine biomarkers for disease onset prediction 

Ongoing collaboration with UKHSA and The Alan Turing Institute: machine learning for COVID interventions, including a feasibility assessment of COVID-19 detection from audio, and extensions to other respiratory conditions

All applicants should have (or expect to obtain) a first-class degree in a numerate discipline (mathematics, science or engineering) or MSc with Distinction (or 70% average) and a strong interest in pursuing research in AI. A keen interest in signal processing, sound, or general research in audio is highly advantageous. Additional experience which is relevant to the area of research is also advantageous.

English language requirements

IELTS minimum 6.5 overall with 6.0 in Writing, or equivalent.

Centre for Vision, Speech and Signal Processing

Surrey Institute for People-Centred AI