Emotion recognition from EEG signals is a major field of research in cognitive computing. The major challenges involved in the task are extracting meaningful features from the signals and building an accurate model. Artificial intelligence (AI) algorithms are used by specialists in particular domains to generate decision-support systems. In recent times, researchers studying brain–computer interfaces (BCIs) have employed deep learning (DL) algorithms to interpret and analyze neural data. It has additionally revealed a possible problem with low transparency because of the complex structure of the algorithms. To improve the interpretability of AI algorithms and their decision-making procedures, explainable artificial intelligence (XAI) presents itself as a viable solution. This study emphasizes the application of deep learning models LSTM (Long Short-Term Memory) and RNN (Recurrent Neural Network) for prediction, while explaining the model prediction outcome and analyzing the feature importance for each feature through different XAI methods. Specifically, the LIME (Local Interpretable Model Agnostic Explanation), SHAP (Shapley Additive exPlanations), ELI5 (Explain Like I’m 5), and Partial Dependence Plot (PDP) techniques are utilized to describe the fundamental features that contribute to the classification results. The efficacy of the aforementioned XAI methods is validated on the publicly available EEG Brainwave Dataset: Feeling Emotions. The overall objective of this research is to strengthen trust and interpret the challenging neural network models in the real-world scenarios by bridging the gap between human interpretability and different ML and DL algorithms.
Explainable AI Methods for Interpreting Emotions in Brain–Computer Interface EEG Data
Discovering the Frontiers of Human-Robot Interaction ; Chapter : 18 ; 419-436
2024-07-24
18 pages
Article/Chapter (Book)
Electronic Resource
English
Brain Activity and Perseveration of Emotions
British Library Conference Proceedings | 1995
|Brain computer interface systems for neurorobotics: methods and applications : editorial
BASE | 2017
|Brain computer interface systems for neurorobotics: methods and applications : editorial
BASE | 2017
|