2020/10/14 by Simon Geirnaert, Geirnaert, Simon, Tom Francart +3 · 2 citations
Computer Science · Engineering · Neuroscience · #Blind Source Separation Techniques #CCD and CMOS Imaging Sensors #EEG and Brain-Computer Interfaces #FOS: Electrical engineering #Signal Processing (eess.SP) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2010.07171
openalex publication_date 2020/10/14 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Auditory attention decoding (AAD) algorithms decode the auditory attention\nfrom electroencephalography (EEG) signals that capture the listener's neural\nactivity. Such AAD methods are believed to be an important ingredient towards\nso-called neuro-steered assistive hearing devices. For example, traditional AAD\ndecoders allow detecting to which of multiple speakers a listener is attending\nto by reconstructing the amplitude envelope of the attended speech signal from\nthe EEG signals. Recently, an alternative paradigm to this stimulus\nreconstruction approach was proposed, in which the directional focus of\nauditory attention is determined instead, solely based on the EEG, using common\nspatial pattern filters (CSP). Here, we propose Riemannian geometry-based\nclassification (RGC) as an alternative for this CSP approach, in which the\ncovariance matrix of a new EEG segment is directly classified while taking its\nRiemannian structure into account. While the proposed RGC method performs\nsimilarly to the CSP method for short decision lengths (i.e., the amount of EEG\nsamples used to make a decision), we show that it significantly outperforms it\nfor longer decision window lengths.\n