Maximum-likelihood estimation of low-rank signals for multiepoch MEG/EEG analysis

摘要: A maximum-likelihood-based algorithm is presented for reducing the effects of spatially colored noise in evoked response magneto- and electro-encephalography data. The repeated component data, or signal interest, modeled as mean, while Kronecker product a spatial temporal covariance matrix. matrix assumed known estimated prior to application algorithm. structure part maximum-likelihood procedure. mean representing interest be low-rank due estimates components are derived order estimate component. relationship between this approach principal analysis (PCA) explored. In contrast prestimulus-based whitening followed by PCA, does not require signal-free data whitening. Consequently, much more effective with nonstationary produces better quality given record length. efficacy demonstrated using simulated real MEG