Based on the high-dimension chaos characteristics of electroencephalogram(EEG) and the analysis by the multi-variable state space reconstruction,better estimation of the correlative dimension can be obtained by dividing the brain into two parts and using eight EEG channels as the reconstruction samples.Lorenz system was firstly tested by request of the amount of data to test the feasibility of the algorithms.After Comparing the high-dimension data of the EEG of epileptic patients with the results of control subjects,it is indicated that the multi-variable state space reconstruction is applicable in short time noise-containing time sequences to obtain reliable results and can free the researchers from the hard choice of delay time and embedding dimension.