Blind Source Separation Techniques

Spatio-dynamical decomposition: Joint spatial and dynamical criteria for FMRI Blind source separation.

This paper introduces spatio-dynamical decomposition for fMRI BSS, showing it better recovers sources using both spatial and temporal criteria than methods that use only one.

PhaseICA: Resolving Individual-Specific Macroscale Brain Propagations

PhaseICA extracts reoccurring spatiotemporal brain patterns by incorporating phase information, improving the analysis of nonstationary brain activity.

A dynamic generative model can extract interpretable oscillatory components from multichannel neurophysiological recordings

This paper introduces a dynamic generative model to extract interpretable oscillatory components from multichannel neurophysiological recordings, improving component selection and interpretation.

Local Subspace Pruning (LSP) for Multichannel Data Denoising

This paper introduces a denoising algorithm for multichannel data that identifies and removes noise by analyzing local subspaces, improving upon linear methods that fail when sources exceed channels.

Narrowband multivariate source separation for semi-blind discovery of experiment contrasts

This paper introduces csGED, a method for separating neural signals in MEG data by combining frequency and condition-specific analysis, showing improved performance over existing methods.

Impact of Model Order Choice on the Results of Parallel Independent Component Analysis

This study evaluates how model order selection affects parallel independent component analysis (pICA) results in schizophrenia research, finding consistent sensitivity to disorder effects and partial capture of spatial patterns by the MDL algorithm.

Learning Complex Representations from Spatial Phase Statistics of Natural Scenes

This paper introduces a generative model for natural scenes that uses complex domain signals and non-uniform phase priors, showing improved performance in blind source separation and revealing phase-sensitive complex cells in the visual cortex.

Multi-context blind source separation by error-gated Hebbian rule

This paper demonstrates that a neural network using the error-gated Hebbian rule can learn to perform multi-context blind source separation by retaining context memories and extracting common features.

Multi-band component analysis for EEG artifact removal and source reconstruction with application to gamma-band activity

This paper introduces multi-band ICA to improve EEG artifact removal and source reconstruction by increasing component count and SNR, outperforming traditional wide-band ICA.