Title :
Dimensionality of evoked potentials
Author :
Uijen, G.J.H. ; Stegeman, D.F.
Author_Institution :
Inst. of Neurology, Nijmegen Univ. Hospital, Netherlands
Abstract :
The number of independent signals present in averaged evoked potentials (EP) can be determined by means of singular value decomposition. It is shown that a simple cutoff in the order of singular values by using the noise level as a threshold is not satisfying due to large spatial and temporal correlations in the noise. The method of choice is the use of an information theoretic criterion (MDL) applied to the decorrelated data matrix. The decorrelation matrix is obtained from the single sweeps of data after subtraction of the average
Keywords :
bioelectric potentials; electroencephalography; medical signal processing; singular value decomposition; average subtraction; averaged evoked potentials; decorrelated data matrix; evoked potentials dimensionality; independent signals number; noise level; simple cutoff; single data sweeps; spatial correlations; temporal correlations; Background noise; Cardiology; Covariance matrix; Decorrelation; Eigenvalues and eigenfunctions; Electrodes; Equations; Hospitals; Nervous system; Noise level; Singular value decomposition;
Conference_Titel :
Engineering in Medicine and Biology Society, 1994. Engineering Advances: New Opportunities for Biomedical Engineers. Proceedings of the 16th Annual International Conference of the IEEE
Conference_Location :
Baltimore, MD
Print_ISBN :
0-7803-2050-6
DOI :
10.1109/IEMBS.1994.411816
