Detection of Sensitive Brain Regions in Problem-Solving State Using Support vector mchine

Electroencephalogram (EEG) signals contain a huge amount of human body performance information. Wavelet transform was used to feature extraction of EEG signals during rest state and problem-solving in normal subjects. Several features such as average, median, skewness, entropy, mean, power, norm, variance, kurtosis and standard deviation were calculated in terms of detailed coefficients and the approximation coefficients of the last decomposition level. The EEG signals were recorded during (۱) the demanding problem-solving task and (۲) the rest state. EEGs on ۱۹ channels from all subjects were analyzed. The features were classified by support vector machine(SVM) classifier. The accuracy of higher ۷۰% was highlighted in delta, theta, alpha and beta frequency bands. Results indicated that classification accuracies of ۲ feature sets in ch۲ and ch۱۱ are more than ۷۰% in three bands of theta, theta and beta, in addition, most brain regions are activated in beta frequency.

Keywords:

Electroencephalogram , Feature Extraction , Discrete Wavelet Transform , Support Vector Machine