Classification ofTimeSeries using Singular Values andWavelet Subband Analysis withANN andSVM Classifiers
Balizs Beny, Péter Somogyi, Péter Várady · 2005
Oscillation ofthecerebral bloodflow(CBF)isa common featurein severalphysiological or pathophysiological states ofthebrain. Itispromising to identify thedisorders ofthecerebral circulation basedonthe classification ofCBFsignals. Inorder todistinguish between different physiological states, twodifferent classification methods havebeendeveloped; aRadial Basis Function based NeuralNetworkanda Support VectorClassifier with Gaussian kernel. Inordertodescribe thetemporal bloodflow patterns, twofeature extraction procedures wereapplied; spectral matrix andwavelet subband analysis. Thesefeature extraction andclassification methodsare evaluated andtheir efficiencies arecompared. Thecomputations werecarried outwith Mathematica 5.1 and its Wavelet Application.