Bayesian Odds-Ratio Filters: A Template-Based Method for Online Detection of P300 Evoked Responses

TL;DR

This paper introduces a Bayesian odds-ratio filter for online detection of P300 responses in EEG signals, demonstrating improved accuracy and speed over traditional correlation methods in BCI/BMI applications.

Contribution

It presents a novel Bayesian odds-ratio based method for signal detection, directly addressing model selection rather than correlation, with demonstrated improvements in P300 detection.

Findings

Significant improvement in ROC curves compared to correlation methods

Enhanced accuracy and speed in detecting evoked brain responses

Applicable to general template signal detection in real-time systems

Abstract

Template-based signal detection most often relies on computing a correlation, or a dot product, between an incoming data stream and a signal template. While such a correlation results in an ongoing estimate of the magnitude of the signal in the data stream, it does not directly indicate the presence or absence of a signal. Instead, the problem of signal detection is one of model-selection. Here we explore the use of the Bayesian odds-ratio (OR), which is the ratio of posterior probabilities of a signal-plus-noise model over a noise-only model. We demonstrate this method by applying it to simulated electroencephalographic (EEG) signals based on the P300 response, which is widely used in both Brain Computer Interface (BCI) and Brain Machine Interface (BMI) systems. The efficacy of this algorithm is demonstrated by comparing the receiver operating characteristic (ROC) curves of the OR-based (logOR) filter to the usual correlation method where we find a significant improvement in P300 detection. The logOR filter promises to improve the accuracy and speed of the detection of evoked brain responses in BCI/BMI applications as well the detection of template signals in general.