نهمین کنگره نقشه برداری مغز

نمایش مقاله

Title:

Continuous Estimation of Meditation Level Using Single-channel Wearable EEG

Subject:

Brain–Computer Interfaces (BCIs)

Presenter Author:
Name:	Fatemeh Davarinia

Author:
Name 1:	Fatemeh Davarinia - Corresponding - Presenter
Affiliation 1:	Neurochallenge Company, Mashhad, Iran

Name 2:	Farzaneh Yousefipoor
Affiliation 2:	Neurochallenge Company, Mashhad, Iran

Name 3:
Affiliation 3:

Name 4:
Affiliation 4:

Name 5:
Affiliation 5:

Name 6:
Affiliation 6:

Name 7:
Affiliation 7:

Name 8:
Affiliation 8:

Name 9:
Affiliation 9:

Name 10:
Affiliation 10:

Abstract

Original Research

introduction :

Continuous estimation of meditation level is challenging due to the gradual, non-stationary, and temporally dynamic nature of relaxation states. Although low-cost single-channel wearable EEG devices offer portability and ease of use, they are particularly sensitive to noise when dry electrodes are employed. In the MindWave Mobile 2 headset, signal acquisition is performed at the FP1 frontal location, which plays a key role in attention and emotional regulation but is inherently prone to high signal variability caused by eye movements, facial muscle activity, and frontal cortical dynamics. Overcoming this variability is essential for reliable real-time meditation monitoring in wearable applications. This study aims to develop and evaluate a lightweight and interpretable framework for continuous meditation level estimation using single-channel wearable EEG, with an emphasis on robustness to signal variability at FP1and suitability for real-time deployment.

Methods:

EEG signals were collected from 10 healthy participants during a structured cognitive–emotional protocol including baseline, relaxation interventions, and cognitively demanding tasks. Seven physiologically interpretable spectral features were extracted from the alpha, theta, and beta frequency bands, selected based on their established sensitivity to relaxation-related brain activity. To suppress transient fluctuations inherent to frontal EEG recordings, features were temporally smoothed using 5-second windows with a 4-second overlap. A time-delay neural network (TDNN) was employed to capture temporal dependencies, with architectures comprising 10, 15, and 20 neurons and delay lengths ranging from 1 to 7. Model performance was evaluated using leave-one-subject-out cross-validation to assess robustness to inter-subject variability and channel sensitivity.

Results:

The optimal TDNN configuration with 15 neurons achieved a mean correlation coefficient of 75.57% ± 8.69% between estimated and reference meditation scores, indicating reliable tracking of gradual relaxation dynamics despite pronounced variability in FP1 recordings. When meditation scores were discretized into five levels ranging from very low to very high, the proposed framework achieved an average area under the ROC curve (AUC) of 79.92% ± 3.49%. Considering the single-channel acquisition and the high sensitivity of FP1 signals recorded with a dry electrode, these results can be regarded as competitive. The findings demonstrate that temporal smoothing combined with TDNN-based modeling effectively attenuates transient EEG fluctuations, enabling stable and consistent online estimation.

Conclusion:

This study demonstrates the feasibility of continuous meditation monitoring using a low-cost, single-channel wearable EEG device under conditions of substantial signal variability. By integrating physiologically meaningful spectral features with temporal modeling, the proposed framework achieves stable real-time performance while maintaining simplicity and ease of deployment. Compared to multi-channel EEG systems, this approach prioritizes portability and computational efficiency, albeit with lower absolute accuracy. Limitations include reliance on a proprietary meditation reference index and constraints inherent to frontal single-channel acquisition, restricting applicability to non-clinical contexts such as mental wellness monitoring, neurofeedback, and educational applications.

Keywords:

Meditation Level Estimation; Wearable EEG; Time-Delay Neural Network; Continuous Cognitive Monitoring;