AI-based Bayesian inference scheme to recognize electroencephalogram signals for smart healthcare

摘要

Canonical Correlation Analysis (CCA) is a popular way to analyze the underlying frequency components of an electroencephalogram (EEG) signal that contains Steady-State Visual Evoked Potentials (SSVEP). But solely itself may not be significant to detect the SSVEP frequency correctly. To improve its accuracy, several methods for processing the signal to optimize reference signals have been introduced. On the other hand, this paper is about "post-processing", in another word, improving accuracy after performing CCA. This paper proposes a method to improve the accuracy of CCA recognition by breaking the EEG signal into several folds, using the recognition result of each fold to update a probability distribution by Bayesian Inference and select the target with the highest probability. The experiment is conducted on a publicly available dataset and the proposed method shows significant improvement in the overall recognition rate. For better communication and status monitoring of the paralyzed patient, this work can be used to improve the current application of the non-invasive Brain-Computer Interface in smart healthcare.