Skip to main content
. Author manuscript; available in PMC: 2014 Mar 13.
Published in final edited form as: J Neural Eng. 2012 Jun 19;9(4):046002. doi: 10.1088/1741-2560/9/4/046002

Table 2. Extracted features.

Groups Feature list
Frequency domain

  • Total power (0-12Hz)

  • Peak frequency of spectrum

  • Spectral edge frequency (80%, 90%, 95%)

  • Power in 2Hz width sub-bands (0-2Hz, 1-3Hz, …10-12Hz)

  • Normalised power in sub-bands

  • Wavelet energy (the EEG is decomposed into 8 coefficients using the Daubechy 4 wavelet, the energy in the 5th coefficient corresponding to 1-2Hz is used as a feature)
Time domain

  • Curve length

  • Number of maxima and minima

  • Root mean squared amplitude

  • Hjorth parameters

  • Zero crossings (raw epoch, Δ, ΔΔ)

  • Autoregressive modelling error (model order 1-9)

  • Skewness

  • Kurtosis

  • Nonlinear energy

  • Variance (Δ, ΔΔ)
Information theory

  • Shannon entropy

  • Singular value decomposition entropy

  • Fisher information

  • Spectral entropy