Key result
Contactless 77-GHz radar detects four emotional states with ~77% accuracy vs. ECG.
Why the study?
Non-contact emotion recognition using frequency-modulated continuous-wave radar is limited in office environments because background clutter, body motion, and respiratory harmonics obscure weak cardiac components, making reliable heartbeat extraction difficult.
Does a 77-GHz FMCW radar system with multi-feature fusion accurately extract heart rate variability and recognize emotional states compared to synchronized ECG measurements?
Does a 77-GHz FMCW radar system with multi-feature fusion accurately extract heart rate variability and recognize emotional states compared to synchronized ECG measurements?
Answers are grounded in this paper and cite their sources.
Multidimensional physiological features including heart rate variability derived from 77-GHz FMCW radar can accurately extract cardiac parameters and support unobtrusive four-class emotion recognition in office environments.
Non-contact emotion recognition has potential applications in smart offices, human–computer interaction, and mental health monitoring. Frequency-modulated continuous-wave (FMCW) radar provides a suitable sensing modality because it captures physiological motion without cameras or wearable sensors. Reliable heartbeat extraction remains difficult in office environments. Background clutter, body motion, and respiratory harmonics can obscure weak cardiac components. This paper presents a non-contact emotion recognition system based on a 77-GHz FMCW radar. The system uses a cascaded signal processing framework to suppress static clutter, localize the human target, and extract physiological features from radar echoes. A robust circle-fitting method based on the Tukey M-estimator is used to calibrate the I/Q trajectory and reduce motion-induced phase distortion. Noise-assisted multivariate empirical mode decomposition (NA-MEMD) is combined with frequency slice wavelet transform (FSWT) to separate respiratory and cardiac components. A correlation-based selective harmonic cancellation strategy is introduced to suppress respiratory harmonics in the heartbeat band while preserving cardiac waveform details. Heart rate variability (HRV), respiratory rhythm, the spectral movement-stability index (SMI), and the cardiopulmonary phase locking value (PLV) are fused for four-class emotion recognition using a random forest classifier. Compared with synchronized ECG measurements, the proposed method achieves mean extraction errors of 4.96 ms for SDNN and 14.46 ms for RMSSD. The average recognition accuracy for calm, happy, anxious, and sad states reaches 76.8%. These results show that multidimensional physiological features derived from FMCW radar can support unobtrusive emotion recognition in office environments.
No takes yet. Share an insight, caveat, or question.
Xu et al. (2026) studied Emotion states (calm, happy, anxious, and sad) in office environments. 77-GHz FMCW radar non-contact emotion recognition system with multi-feature fusion vs. Synchronized ECG measurements was evaluated on Four-class emotion recognition accuracy (calm, happy, anxious, and sad states). The 77-GHz FMCW radar system achieved 76.8% average accuracy for four-class emotion recognition and mean errors of 4.96 ms for SDNN and 14.46 ms for RMSSD vs ECG.
Synapse has enriched 10 related papers. Consider them for comparative context: