An EEG-Guided Olfactory Interface: Prototype Design and Person-Specific Emotion-Decoding Validation

Just-in-time adaptive interventions require timely and low-burden state estimation, while olfaction offers a programmable output channel with limited attentional demand. We describe a prototype architecture that links electroencephalography (EEG)-based emotion estimation to a six-channel odorant device and evaluate only the EEG sensing and decoding module. Forty EEG sessions from 39 adults were recorded with a 14-channel Emotiv EPOC X headset (128 Hz) during six standardized emotion-induction conditions. No odor was administered. Band-power, frontal alpha asymmetry (FAA) and global field power (GFP) were analyzed with rank-based repeated-measures tests and explicit multiple-comparison correction. Emotion decoding used subject-aware cross-validation. Frontal beta power, the beta/alpha ratio and GFP differed across conditions after false-discovery-rate correction, although effect sizes were small (Kendall’s W = 0.089–0.155). On-line affective metrics showed larger effects (W = 0.130–0.365). Six-class accuracy was 45.1% ± 13.2% within participants (n = 29; chance 16.7%; p < 10−8) and 23.1% across participants after per-subject normalization (macro-F1 = 0.23; permutation p = 0.005). FAA did not differ. Consumer-headset EEG contained person-specific information about laboratory-induced states, but performance was not sufficient to establish a clinically usable regulator. The results validate neither a complete closed loop nor olfactory efficacy; end-to-end latency, artifact and temporal robustness, chemical characterization and controlled odor-regulation effects require prospective evaluation.