Real-time classification of attended source with EEG and EarEEG brain-computer interface
Title: Real-time classification of attended source with EEG and EarEEG brain-computer interface
SNIC Project: LiU-storage-2020-3
Project Type: LiU Storage
Principal Investigator: Emina Alickovic <emina.alickovic@liu.se>
Affiliation: Linköpings universitet
Duration: 2020-02-05 – 2021-03-01
Classification: 20205
Keywords:

Abstract

How do we select behaviorally relevant voice from a dense mixture of many incessant streams of real-world sounds ('cocktail party' sounds)? As of the date, the major contributions for resolving this question are rather limited. This project will open a new perspective on selective attention; as it we will suggest a technique grounded on the model-based learning (model fitting) to select the attended voice (AV) from multiple sound streams. This project is a collaboration between Linkoping University and Eriksholm (Independent Research center of the hearing aid manufacturer Oticon). The long-term goal of this project would be to evaluate this technique on hearing impaired persons, and to develop a prosthesis that amplifies only AV. Here we will focus on forward modelling from the attended voice to EEG, and in the later stage of the project, to wearable EEG (EarEEG). It is the intention of this project to design a model with high performance using advanced machine learning algorithms, what will require an extensive experimentation with data utilities ranging from data collection and storage (in huge volumes) to data analysis. It is computationally expensive to assess our model(s) on the ordinary laptop. To make the calculations for ONE sound stream of 120 minutes (40 trials, each 3 minutes long) and to process more than 1000 different combinations require  150 hours of computing on an ordinary laptop. And we need to process a greater number of different combinations of shorter sounds and the same procedure needs to be repeated for 30 (or more) subjects that would therefore need our computers to run for several months. We need a faster turn-around time and therefore we are seeking this supercomputer opportunity.