DEED: A Multimodel Dataset for Dream Emotion Classification
Lei Zheng, Di Zhou, Meng Zhang et al.
A new open-source dataset, the Dream Emotion Evaluation Dataset (DEED), provides multimodal dream-related information including neural data collected from 38 participants over 82 nights using polysomnography. After each REM awakening, participants reported dream content and affective states. Using DEED, the first dream emotion classification algorithms were implemented, combining feature extraction methods (power spectral density, differential entropy, multi-frequency band common spatial pattern) with support vector machine and convolutional neural networks. The combination of multi-frequency band common spatial pattern and support vector machine achieved the highest accuracy of 83.6%. The dataset aims to accelerate research on decoding neural mechanisms of dream emotions.