Riemannian Geometry-Preserving Variational Autoencoder for MI-BCI Data Augmentation
This work addresses data augmentation needs for MI-BCI applications, offering potential benefits for signal privacy and scalability, but it is incremental as it adapts existing VAE methods to preserve geometric properties.
The paper tackled the challenge of generating synthetic EEG covariance matrices for motor imagery brain-computer interfaces by proposing a Riemannian geometry-preserving variational autoencoder (RGP-VAE), which successfully produced valid and representative matrices while learning a subject-invariant latent space, with practical utility depending on the classifier used.
This paper addresses the challenge of generating synthetic electroencephalogram (EEG) covariance matrices for motor imagery brain-computer interface (MI-BCI) applications. Objective: We aim to develop a generative model capable of producing high-fidelity synthetic covariance matrices while preserving their symmetric positive-definite nature. Approach: We propose a Riemannian geometry-preserving variational autoencoder (RGP-VAE) integrating geometric mappings with a composite loss function combining Riemannian distance, tangent space reconstruction accuracy and generative diversity. Results: The model generates valid, representative EEG covariance matrices, while learning a subject-invariant latent space. Synthetic data proves practically useful for MI-BCI, with its impact depending on the paired classifier. Contribution: This work introduces and validates the RGP-VAE as a geometry-preserving generative model for EEG covariance matrices, highlighting its potential for signal privacy, scalability and data augmentation.