Abstract
Due to a large number of potential applications, a good deal of effort has been recently made toward creating machine learning models that can recognize evoked emotions from one's physiological recordings. In particular, researchers are investigating the use of EEG as a low-cost, non-invasive method. However, the poor homogeneity of the EEG activity across participants hinders the implementation of such a system by a time-consuming calibration stage. In this study, we introduce a new participant-based feature normalization method, named stratified normalization, for training deep neural networks in the task of cross-subject emotion classification from EEG signals. The new method is able to subtract inter-participant variability while maintaining the emotion information in the data. We carried out our analysis on the SEED dataset, which contains 62-channel EEG recordings collected from 15 participants watching film clips. Results demonstrate that networks trained with stratified normalization significantly outperformed standard training with batch normalization. In addition, the highest model performance was achieved when extracting EEG features with the multitaper method, reaching a classification accuracy of 91.6% for two emotion categories (positive and negative) and 79.6% for three (also neutral). This analysis provides us with great insight into the potential benefits that stratified normalization can have when developing any cross-subject model based on EEG.
Highlights
Emotion recognition has gained great attraction due to its large number of potential applications in fields such as human-computer interaction (Brave and Nass, 2009), interactive storytelling (Fels et al, 2011), and mood disorders (El Keshky, 2018)
In order to solve this problem with EEG, current methods rely on participant-dependent models tuned with tedious and time-consuming calibration sessions implemented before each experiment
This section first presents the results of the experiment, analyzes the between-participant variance and cross-subject emotion recognition in the layers of the neural networks, and compares the results of this work with state-ofart literature
Summary
Emotion recognition has gained great attraction due to its large number of potential applications in fields such as human-computer interaction (Brave and Nass, 2009), interactive storytelling (Fels et al, 2011), and mood disorders (El Keshky, 2018). The main bottleneck in the development of models trained with EEG signals is the poor homogeneity of between-sessions data and between-participants data, which, interestingly, is not Neural Networks With Stratified Normalization apparent in the literature in the context of emotion recognition from facial expressions or other physiological data (Cimtay and Ekmekcioglu, 2020). One example is the study by Li et al (2020), where researchers used an unsupervised deep generative model to capture the emotion-related information between participants. Another example is from Yin et al (2017), who present an EEG feature selection approach to determine a set of the most robust EEG indicators with stable geometrical distribution across a group of participants. The study by Song et al (2020) proposes a graph to model the multichannel EEG features and perform EEG emotion classification based on this model
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
