Date of Publication :17th January 2024

Abstract: This research paper proposes a novel gated recurrent unit (GRU) neural network for speech emotion recognition (SER) in children, using the FAU-Aibo dataset of children’s interactions with the Aibo robot. The GRU model shows promise in accurately predicting negative emotions in children's speech. The paper compares the GRU model with other deep learning and machine learning models, such as LSTM, SVM, and boosted trees, and shows that the GRU model achieves better accuracy, speed, and computational footprint. The paper also discusses the challenges and implications of using natural and spontaneous speech data for emotion recognition and how the GRU model can help detect negative emotions in children as a potential step toward child abuse detection.

Reference :