Fuzzy ensemble-based federated learning for EEG-based emotion recognition in Internet of Medical Things

Abstract

Emotion recognition based on electroencephalography (EEG) is a crucial research area in the Internet of Medical Things (IoMT), with significant applications in engineering and entertainment. Addressing challenges such as efficient EEG feature extraction, accurate classification, and data privacy, this study introduces a novel fuzzy ensemble-based federated learning framework for EEG-based emotion recognition. We integrate three deep learning models, including a temporal convolutional network (TCN), long short-term memory (LSTM), and gated recurrent unit (GRU), and employ a Gompertz function-based fuzzy rank approach to combine their predictions. Additionally, we propose an asynchronous dropout algorithm within the federated learning framework to aggregate a global model, ensuring data privacy and mitigating gradient staleness. Our approach is validated using three public datasets, including GAMEEMO, SEED and DEAP, demonstrating superior performance in accuracy and F1 score compared to existing methods.