This dataset contains electroencephalogram (EEG), galvanic skin response (GSR), and electrocardiogram (ECG) recordings from 16 healthy participants during an affective music brain-computer interface training study. Participants listened to 40-second music clips designed to induce specific emotional states across three sessions, with subjective valence and arousal ratings collected. The dataset supports research on music-induced emotion recognition and brain-computer interface development for affective state modulation. This is the training dataset; companion datasets for system calibration and online real-time control are also available.