A Bayesian Optimized Biquad Filter-Burg Model for Real-Time EEG Phase Estimation
Abstract
Real-time phase estimation on stimulus-aligned transient EEG is critical for closed-loop neuromodulation systems targeting the theta-band, yet most real-time phase estimation methods have been developed and validated on resting-state, high-fidelity signals in the alpha band. Recent work has pursued deep learning architectures, while individualized classical bandpass filter tuning remains underexplored. This study characterizes the spectral and temporal properties of non-anticipative bandpass filtering and demonstrates the performance gains achieved through individualized tuning. We present a Bayesian-Optimized Biquad Filter-Burg (BOB) model evaluated on EEG recorded during peripheral vibrotactile stimulation of the left median nerve (theta, N=20 healthy subjects) and on resting-state recordings from the Max Planck Institute's LEMON dataset (alpha, N=20 healthy subjects). BOB was benchmarked against the Deep Dual-Branch Complex Neural Network (DBCNN), EEG Phase Prediction Network (EPN), Phastimate, and Phase Estimation with Autoregressive Padding (PEAP) under matched testing conditions: channel selection, individualized optimization, evaluation time points, and an offline reference phase obtained through zero-phase filtering and the Hilbert transform. Under these testing conditions, BOB achieved matched or better performance than these models at a fraction of the calibration time.
Related
- Embracing uncertainty in time: A Bayesian dynamic mixing model for isotopic diet estimation
- Estimation of Genomic Breeding Values for Economic Traits in Alxa-Type Cashmere Goats Using Bayesian Methods.
- A single day of natural daylight exposure and physical activity does not conclusively change circadian phase: A Bayesian crossover experiment
- Visualization of intra-thalamic nuclei with optimized white-matter-nulled MPRAGE at 7T
- Optimized Cardiorespiratory State at Tone Onset Enhances Sensory Responses but Not Associative Learning in Healthy Adults
BibTeX
@misc{martinezkay2026bayesian,
title = {{A Bayesian Optimized Biquad Filter-Burg Model for Real-Time EEG Phase Estimation}},
author = {Ivan Martinez-Kay and Tatiana Kardashina and ShiNung Ching and Eric C. Leuthardt and Benjamin T. Acland},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.05374},
eprint = {2611.05374},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.05374},
}
RIS
TY - UNPB
TI - A Bayesian Optimized Biquad Filter-Burg Model for Real-Time EEG Phase Estimation
AU - Ivan Martinez-Kay
AU - Tatiana Kardashina
AU - ShiNung Ching
AU - Eric C. Leuthardt
AU - Benjamin T. Acland
PY - 2026
DA - 2026/11/03
PB - arXiv
DO - 10.48550/arxiv.2611.05374
UR - https://arxiv.org/abs/2611.05374
AB - Real-time phase estimation on stimulus-aligned transient EEG is critical for closed-loop neuromodulation systems targeting the theta-band, yet most real-time phase estimation methods have been developed and validated on resting-state, high-fidelity signals in the alpha band. Recent work has pursued deep learning architectures, while individualized classical bandpass filter tuning remains underexplored. This study characterizes the spectral and temporal properties of non-anticipative bandpass filtering and demonstrates the performance gains achieved through individualized tuning. We present a Bayesian-Optimized Biquad Filter-Burg (BOB) model evaluated on EEG recorded during peripheral vibrotactile stimulation of the left median nerve (theta, N=20 healthy subjects) and on resting-state recordings from the Max Planck Institute's LEMON dataset (alpha, N=20 healthy subjects). BOB was benchmarked against the Deep Dual-Branch Complex Neural Network (DBCNN), EEG Phase Prediction Network (EPN), Phastimate, and Phase Estimation with Autoregressive Padding (PEAP) under matched testing conditions: channel selection, individualized optimization, evaluation time points, and an offline reference phase obtained through zero-phase filtering and the Hilbert transform. Under these testing conditions, BOB achieved matched or better performance than these models at a fraction of the calibration time.
ER -