Source characterization of evoked potentials after SVD truncation using the noise properties in the single sweeps

The noise in evoked electrical brain activity can be substantially reduced by means of singular value decomposition and truncation since the effective rank of the data matrix is small compared to the number of leads. It is shown that this procedure can only be successful when the statistical properties of the noise are taken into account.<
>