Font Size: a A A

Study On Multivariate Analysis Of Electroencephalogram Data

Posted on:2010-06-25Degree:MasterType:Thesis
Country:ChinaCandidate:J N LiuFull Text:PDF
GTID:2178360275485631Subject:Computer technology
Abstract/Summary:PDF Full Text Request
Multivariate statistics develop rapidly in the pass serval decades. Clustering analysis and discriminant analysis which are branches in multivariate statistics are used to Electroencephalogram Data classification more and more. Clustering analysis performe classifing according to EEG features,and in order to make sure the accuracy rate and then discriminant analysis is used to forecast. We analysis 21 derivation electrodes'EEG data collected from six healthy subjects by Multivariate statistics in order to find the change of EEG after drinking.We have performed clustering analysis for 21 derivation electrodes'EEG data collected from six healthy subjects.EEG data was goten from six objects of closing eyes and taking in 7.2ml alcohol in every 20 minutes for five times.Result of clustering to 21 channels under relative power of six events confirms that four normal clusters have gotten by frontal poles, front and center, occiput and temporal area. The frontal poles have been clustered in a cluster separately when subject close eyes in the beginning. But they have been clustered with the nearby electrodes when subjects take in alcohol. With the volume increases, they have been departed again. Such phenomenon takes place again and again during the six events. About the occiput, except a subject's right and left occiput come into being a cluster, all the other subjects'most occiput electrodes have been clustered in a cluster,and only a few have been clustered with other electrodes and even become the only member in a cluster. There are fourteen clusters have been clustered with the occiput electrodes and most of the rest have been clustered with front and temporal electrodes. The temporal electrodes changes randomly in events and even become the single member in a cluster. Reslut of clustering to EEG power in six events indices that when take in 36ml alcohol, frontal pole and the front except central points has decreased 15.74% and 10.76% separately.Front and central emporals has decreased 5.25% and left posterior temporal increased 18.34%.The parietal and occipital except left posterior temporal and central points has increased 16.50% and 14% separately. EEG activity has prominent reaction to absorbing alcohol.The regular four clusters when close eyes has been disrupted when subjects take in alcohol.After taking in alcohol, the frontal pole, forehead and center are in a cluster and with the volume increasing frontal poles are in a cluster again.That repeats in six events in turn randomly. The center has usually been clustered with electrodes nearby. The occipital electrodes are in a cluster separately in alcohol events. Tempora changes randomly when take in alcohol and even come into being a single electrode. The high-frequency wave increases in the beginning and decreases later during taking in alcohol events.About three discriminant analysis Mahalanobis,Fisher discriminant and Bayes discriminant analysis of six subjects, the electrode classification accuracy rates are 64.4%, 72.3%,22.7% seprately. We have taken Fisher discriminant analysis to the EEG data of taking in alcohol. The electrodes classification is reasonable when haven't drinking. After drinking 200ml beer, the number of electrodes which the center being discriminated to the occiput increases, the front and being discriminated to the center increases. After drinking 400ml beer, the number of electrodes which the center being discriminated to the back reduce, the front and temporal being discriminated to the central reduced. After drinking 600ml beer, the number of electrodes which the center being discriminated to the back reduce, the front being discriminated to the central increase, and the temporal being discriminated to the front increase. After drinking 800ml beer, the number of electrodes which the front being discriminated to the center reduce, the occiput being discriminated to the center increased.After drinking 1000m beerl, the number of electrodes which the front and occiput being discriminated to the center increase.After taking multivariate anlysis to EEG of normal and drinking events,here is the conclusion: Four normal clusters have gotten by frontal poles, front and center, occiput and temporal area when close eyes and no drinking.Once drink, high-frequency wave of electrode area increas obviously and center and temporal increas a little but occiput even have no obvious change.With the drinking going on, high-frequency wave decrease so that such change tend to renew like before.So we know clustering analysis and discriminant analysis togather have sloved the question of EEG change after taking drinking generally.
Keywords/Search Tags:Electroencephalogram, Clustering, Absolute power, Relative power, Discriminant Analysis
PDF Full Text Request
Related items