What is the clinical evidence from this study?

Study design: Systematic Review. Population: Emotion Recognition (n=233). Intervention: Deep Learning Techniques.

November 18, 2025F1000Research1 citationsOpen Access

A Review of Deep Learning Techniques for EEG-Based Emotion Recognition: Models, Methods, and Datasets

PSP. SreehariManipal Academy of Higher Education URU. RaghavendraManipal Academy of Higher Education AGAnjan GudigarManipal Academy of Higher Education

Key Result

Deep learning techniques for EEG-based emotion recognition have advanced significantly, but significant heterogeneity in datasets, model architectures, and evaluation strategies currently prevents formal meta-analysis.

Study Design

Type

Systematic Review (n=233)

Structured PICO

Population

Studies with EEG signals for emotion recognition using deep learning architectures published from 2020 to 2025

Intervention

Deep learning architectures for EEG-based emotion recognition

Outcome

Overview of DL-based EEG emotion recognition approaches, models, methods, and datasets

This systematic review provides a roadmap for developing EEG-driven emotion recognition systems using deep learning, guiding researchers toward more reliable and scalable systems.

Limitations

Significant heterogeneity in datasets, model architectures, and evaluation strategies prevented meaningful pooling of results for a meta-analysis.
LSTM networks can be prone to overfitting when working with smaller datasets.
LSTMs might not fully capture spatial correlations between EEG channels and may be sensitive to irrelevant features and subject variability.

Abstract

Emotion Recognition (ER) with Electroencephalography (EEG) has become a major area of focus in affective computing due to its direct measurement of the activity of the brain. ER based on EEG has also advanced with the popularity of Deep Learning (DL) and its advancements related to classification accuracy and model efficiency. This systematic review is conducted following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines and aims to provide an overview of DL-based EEG emotion recognition approaches. A comprehensive literature search was conducted across five major databases covering the publications from 2020 to 2025. The studies with EEG signals for ER using DL architectures were included in the present review. Finally, a total of 233 articles were considered after eligibility screening. To enhance the diversity of investigation, we assessed the public datasets utilized for ER based on EEG in terms of their stimulation procedures and emotional representation. Further, the provided analysis attempts to direct future research toward EEG-based emotion identification systems that are more interpretable, generalizable, and data-efficient. This systematic review aims to provide a roadmap for developing EEG-driven ER, guiding researchers toward more reliable, scalable, and practically useful systems.

问 AI

Bookmark

View Full Paper