Premium
Low sampling rate induces high correlation dimension on electroencephalograms from healthy subjects
Author(s) -
Jing Hongkui,
Takigawa Morikuni
Publication year - 2000
Publication title -
psychiatry and clinical neurosciences
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.609
H-Index - 74
eISSN - 1440-1819
pISSN - 1323-1316
DOI - 10.1046/j.1440-1819.2000.00729.x
Subject(s) - correlation , sampling (signal processing) , correlation dimension , statistics , mathematics , dimension (graph theory) , electroencephalography , aliasing , correlation coefficient , analysis of variance , data set , pattern recognition (psychology) , artificial intelligence , psychology , computer science , neuroscience , combinatorics , mathematical analysis , computer vision , filter (signal processing) , geometry , fractal dimension , undersampling , fractal
The aim of this paper was to elucidate the influence of sampling parameters in the non‐linear analysis of a resting electroencephalogram (EEG) in healthy subjects. Electroencephalograms in 12 healthy volunteers were recorded and the signal digitized at 128, 256, 512 and 1024 Hz, respectively, with the resolution of 8 bits, 12 bits and 16 bits for each sampling rate. Correlation dimension was calculated on each data set. Results were demonstrated on brain maps and examined by analysis of variance ( ANOVA ). Correlation integral functions demonstrated four parts separated by critical points. The data showed that sampling rate significantly affected the estimation, while resolution did not influence the results. The correlation dimensions calculated with the sampling rate at or below 256 Hz were apparently higher than the results obtained at 1024 Hz. The values at 512 Hz and 1024 Hz did not differ. The data revealed that low sampling rate can severely distort the estimation of correlation dimension. The optimal sampling rate for analyzing resting EEG on normal subjects is 512 Hz. Limitation and aliasing phenomenon are discussed in the paper.