[Multi-source adversarial adaptation with calibration for electroencephalogram-based classification of meditation and resting states].
Sheng wu yi xue gong cheng xue za zhi = Journal of biomedical engineering = Shengwu yixue gongchengxue zazhi August 25, 2025 Mingyu Gou, Haolong Yin, Tianzhen Chen et al.
A new machine-learning model, the model-calibrated multi-source adversarial adaptation network (CMAAN), improves the accuracy of monitoring meditation states from electroencephalogram (EEG) signals. The model trains multiple domain-adversarial neural networks between different individuals and then calibrates them with a small amount of labeled data from the target person. Tested on EEG data from 18 people undergoing methamphetamine rehabilitation, the model achieved 73.09% classification accuracy. The approach addresses the challenge of large individual differences in EEG signals, and the learned model also reveals key EEG frequency bands and brain regions involved in meditation.