Effect-size-based electrode and feature selection for emotion recognition from EEG

Abstract

Emotion recognition from EEG signals allows the direct assessment of the "inner" state of the user which is considered an important factor in Human-Machine-Interaction. Given the vast amount of possible features from scalp recordings and the high variance between subjects, a major challenge is to select electrodes and features that separate classes well. In most cases, this decision is made based on neuroscientific knowledge. We propose a statistically-motivated electrode/feature selection procedure, based on Cohen's effect size f 2 . We compare inter-and intra-individual selection on a self-recorded database. Classification is evaluated using quadratic discriminant analysis (QDA). We found both feature selection versions based on f 2 yield comparable results. While highest accuracies up to 57,5% (5 classes) are reached by applying intra-individual selection, inter-individual analysis successfully finds features that perform with lower variance in recognition rates across subjects than combinations of electrodes/features suggested in literature.

EEG-Based BCI Emotion Recognition: A Survey

“…The following are some examples of algorithms for feature selection: Effect-size (ES)-based feature selection is a filter method. ES-based univariate: Cohen’s is an appropriate effect size for comparisons between two means [ 100 ]. So, if two groups’ means do not differ by 0.2 standard deviations or more, the difference is trivial, even if it is statistically significant.…”

Electroencephalogram-based emotion assessment system using ontology and data mining techniques

“…It can work with noisy data and missing data in dataset. C4.5 is one of the preeminent inductive inference algorithms and has been successfully applied to affective computing tasks [4,22,56]. In our research, the multiple EEG features are classified into low/high arousal (or low/high valence) by performing a gender-specific classification task as follows:…”

“…Linear methods, which include time-domain and frequency-domain analyses, attempt to model a time series of EEG signals with a specific mathematical expression, and study EEG in several classic non-overlapping frequency bands: theta wave (3-7 Hz), alpha wave (8-13 Hz) and beta wave (14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29). Recently, various non-linear approaches have been introduced.…”

“…In the emotional recall context, Smith et al [21] identifies an augmentation of activity in the connections between the hippocampus and the amygdala during the recollection of negative events compared to neutral events. In [22], statistical t-test and a univariate feature selection method using Cohen's effect size f 2 from analysis of variance were implemented for electrode and feature selection. Electrodes and features found by these approaches resulted in a small variance in classification accuracies across subjects.…”

An Enhanced Feature Extraction Method and Classification Method of EEG Signals using Artificial Intelligence

“…The fundamental downside of these techniques is that they don't function admirably for nonlinear order issues. As of late backpropagation neural systems (BPNNs) and Support vector machines (SVMs) have been indicated to enhance the classification accuracy and linear methods, for example, Linear discriminant analysis [11,12].…”