Most noise removal techniques target common audio interference. However, this project requires a different approach: exploring how to eliminate a rare and distinctive noise to challenge conventional methods.

The goal is to explore machine learning methods, such as AutoEncoders or other suitable architectures, to develop models capable of isolating and eliminating this rare noise. You will design and optimize these models for use on low-end devices, similar to mobile phones, where computational resources are limited.

This project combines the precision of audio signal processing with the creativity and scalability of machine learning, pushing the boundaries of common understanding.
 

Duration: flexible, to be agreed (typically 3-4 months), starting time is flexible

Location: Antwerp (Belgium) 

Student enrolled in Ph.D. Computer Science/Engineering in Machine Learning, Signal Processing or related Proficiency in Python and/or Rust and familiarity with ML libraries Language skills: English Ability to analyze noise patterns and develop innovative strategies to address them. Willing to learn new tools that typically live outside the scope of ML tasks. A strong publication record is also a big plus.

Your tasks and hat you’ll gain

You will design and optimize these AutoEncoder models for use on low-end devices, similar to mobile phones. Experience solving specialized challenges in audio processing using cutting-edge ML techniques. Hands-on expertise in building and optimizing ML models for constrained environments, like mobile devices. You evaluate and benchmark techniques in the context of specific requirements.