Incremental Non-Gaussian Analysis on Multivariate EEG Signal Data

Kam Swee Ng,Hyung-Jeong Yang,Sooyoung Kim,Sun-Hee Kim
DOI: https://doi.org/10.1587/TRANSINF.E95.D.3010
2012-12-01
Abstract:In this paper, we propose a novel incremental method for discovering latent variables from multivariate data with high efficiency. It integrates non-Gaussianity and an adaptive incremental model in an unsupervised way to extract informative features. Our proposed method discovers a small number of compact features from a very large number of features and can still achieve good predictive performance in EEG signals. The promising EEG signal classification results from our experiments prove that this approach can successfully extract important features. Our proposed method also has low memory requirements and computational costs. key words: EEG, independent component analysis, principle component analysis, incremental model
What problem does this paper attempt to address?