Challenge for Affective Brain-Computer Interfaces: Non-stationary Spatio-spectral EEG Oscillations of Emotional Responses

Abstract

Electroencephalogram (EEG)-based affective brain-computer interfaces (aBCIs) have been attracting ever-growing interest and research resources. Whereas most previous neuroscience studies have focused on single-day/-session recording and sensor-level analysis, less effort has been invested in assessing the fundamental nature of non-stationary EEG oscillations underlying emotional responses across days and individuals. This work thus aimed to use a data-driven blind source separation method, i.e., independent component analysis (ICA), to derive emotion-relevant spatio-spectral EEG source oscillations and assess the extent of non-stationarity. To this end, this work conducted an 8-day music-listening experiment (i.e., roughly interspaced over 2 months) and recorded whole-scalp 30-ch EEG data from 10 subjects. Given the large size of the data (i.e., from 80 sessions), results indicated that EEG non-stationarity was clearly revealed in the numbers and locations of brain sources of interest as well as their spectral modulation to the emotional responses. Less than half of subjects (two to four) showed the same relatively day-stationary (source reproducibility >6 days) spatio-spectral tendency towards one of the binary valence and arousal states. This work substantially advances the previous work by exploiting intra- and inter-individual EEG variability in an ecological multiday scenario. Such EEG non-stationarity may inevitably present a great challenge for the development of an accurate, robust, and generalized emotion-classification model.

Recurrent Neural Network-based Acute Concussion Classifier using Raw Resting State EEG Data

“…EEG signals are well known to be strongly non-stationary on timescales greater than ∼0.25 s, exhibiting significant variations, and even shifts, in the statistical properties of the signal over time [61,62]. This behaviour poses a significant challenge for extracting stable features from the signal as well as for designing reliable brain-computer interface systems within the context of real environments [85][86][87][88]. In the present study, we leverage this behaviour to generate multiple, effectively independent, realizations of the raw90-rsEEG datasets from each of the participants.…”

“…The problem is further compounded by the fact that EEG time series is strongly non-stationary and simple procedures, like filtering, do not completely remove the non-stationarity of features. This tendency is the source of the difficulties faced by researchers pursuing reliable brain-computer interfaces [86,88]. In light of this, and in the interest of ensuring that the data processing pipeline is as simple as possible, we felt it best to use the observed (raw) EEG signals and rely on a deep network to learn the signals.…”

Objective Assessment of Impulse Control Disorder in Patients with Parkinson’s Disease Using a Low-Cost LEGO-Like EEG Headset

“…That is, each recruited subject participated in the Go/NoGo protocol with and without DA treatment only once. However, intra-individual differences in task-related EEG activities may present ecologically on a daily basis [44][45][46][47]. Several behavioral and psychological states such as attention, stress, anxiety and/or sleep quality may contribute to the above EEG non-stationarity.…”

EEGs Vary Less Between Lab and Home Locations Than They Do Between People

“…Finally, Shen and Lin ( 2019 ) studied both inter and intra-subject variation in EEG during emotional responses. They found substantial inter- and intra-subject variation, not unlike what we show here.…”