Simon Geirnaert

Research

Signal Processing Algorithms for EEG-based Auditory Attention Decoding

People with hearing impairment have major difficulties understanding speech in noisy environments, which is why most hearing aids and cochlear implants are nowadays equipped with a speech enhancement algorithm to suppress background noise. However, in so-called ‘cocktail party’ scenarios with multiple speakers, such an algorithm must know which speaker the user aims to attend to and which speaker(s) should be treated as noise. The goal of this PhD project is to identify the attended speaker by simultaneously analyzing brain signals collected by electroencephalography (EEG) sensors and noisy speech mixtures collected by a microphone array. To this end, we will design novel multi-modal signal processing algorithms to extract the speech-related signal components that are coupled across both modalities (audio and EEG). We focus on an early integration of the audio and EEG signals in a coupled optimization problem to allow both modalities to assist and inform each other, leading to a more accurate and faster auditory attention detection. Furthermore, we aim for blind algorithms that require no or minimal prior training data, and which are able to adapt over time in a data-driven fashion.

Active projects

• Listen very carefully - Online tracking of auditory attention via electroencephalography (EEG) in individuals and groups of learners

Awards

• Best Paper Award WIC-IEEE Symposium 2019

Publications