Click here to load whole tree
NITAAI-Veda.nyf > Soul Science God Philosophy > Science and Spiritual Quest > Section 4 Towards a New Biology > MATHEMATICAL TECHNIQUES FOR STUDY OF EEG > 2. Methods > 2.3 Features considered

2.3 Features considered

 

This section describes the features considered for classifying EEG data.

 

Fractal Dimension: Katz's approach is used to calculate the FD of a waveform.

 

Complexity measure: Lempel-Ziv complexity measure C (n) (Zhang XS, Roy RJ and Jension EW, 2001) is used in our work, since it is extremely suited for characterizing the development of spatio-temporal activity patterns in high-dimensionality nonlinear systems.The other features considered are mean, variance and powers in alpha, beta, theta and delta waves.

 

Data handled: The EEG data consists of recordings made from a subject performing meditation. This data was collected during a meditation session where the subject was asked to simply sit without meditation for some time and then asked to start meditation.

 

Features Considered  KNNC  MLP    RBF    SVM

Mean, Variance        84.47  87.58  88.47  89.31

Ractal Dimension (FD), Complexity Measure (CM) 90.54  94.65  97.79  98.18

Mean,   Variance, FD, CM   89.74  98.38  98.29  98.81

Frequency Domain Features 87.28  96.25  96.43  98.29

Mean,   Variance, FD,       CM      & Frequency domain features    89.44  98.81  98.81  99.60

 

Table 1: Testing Accuracy Table for EEG Data (Two-category case)

 

After the end of meditation, recording was continued and post meditation data was collected. We considered 60 segments of premeditation data, 140 segments of meditation data, 140 segments of deep meditation data and 60 segments of post meditation data.

 

Features Considered  KNNC  MLP    RBF    SVM

Mean, Variance        84.47  87.58  88.47  89.31

Fractal       Dimension       (FD), Complexity Measure (CM)  90.54  94.65  97.79  98.18

Mean, Variance, FD, CM      89.74  98.38  98.29  98.81

Frequency Domain Features 87.28  96.25  96.43  98.29

Mean,  Variance,  FD,  CM & Frequency domain features         89.44  98.81  98.81  99.60

Features Considered  KNNC  MLP    RBF    SVM

Mean, Variance        74.40  81.57  86.44  87.30

Fractal      Dimension      (FD), Complexity Measure (CM)    80.55  92.56  97.27  97.65

Mean, Variance, FD, CM      78.50  98.29  97.95  98.81

Frequency Domain Features 77.82  96.25  96.43  97.95

Mean,  Variance,  FD,  CM & Frequency domain features         80.20  98.63  98.29  99.60

Table 2: Testing Accuracy Table for EEG Data (Three-category case).