Table 3.

Feature definitions

	$10{ \mathbf{\text{log}} }_{10}\left(\mathit{\mu }(·)\right)$	$10{ \mathbf{\text{log}} }_{10}\left(\mathit{\sigma }(·)\right)$
${\mathit{x}}^{(i)}$		$f_5^{(i)}=10 { \text{log} }_{10}\left(\sigma \left({\mathit{x}}^{(i)}\right)\right)$
$\mathit{R}\left({\mathit{y}}^{\left(i\right)}\right)$	$f_1^{(i)}=10 { \text{log} }_{10}\left(\mu \left(R\left({\mathit{y}}^{\left(i\right)}\right)\right)\right)$	$f_6^{(i)}=10 { \text{log} }_{10}\left(\sigma \left(D\left({\mathit{y}}^{(i)}\right)\right)\right)$
$\mathit{L}\left({\mathit{y}}^{\left(i\right)}\right)$	$f_2^{(i)}=10 { \text{log} }_{10}\left(\mu \left(L\left({\mathit{y}}^{\left(i\right)}\right)\right)\right)$	$f_7^{(i)}=10 { \text{log} }_{10}\left(\sigma \left(L\left({\mathit{y}}^{(i)}\right)\right)\right)$
$\mathit{C}\left({\mathit{y}}^{\left(i\right)}\right)$	$f_3^{(i)}=10 { \text{log} }_{10}\left(\mu \left(C\left({\mathit{y}}^{(i)}\right)\right)\right)$	$f_8^{(i)}=10 { \text{log} }_{10}\left(\sigma \left(C\left({\mathit{y}}^{(i)}\right)\right)\right)$
$\mathit{T}\left({\mathit{y}}^{\left(i\right)}\right)$	$f_4^{(i)}=10 { \text{log} }_{10}\left(\mu \left(T\left({\mathit{y}}^{(i)}\right)\right)\right)$	$f_9^{(i)}=10 { \text{log} }_{10}\left(\sigma \left(T\left({\mathit{y}}^{(i)}\right)\right)\right)$
	${ \text{tan} }^{-1}\left(\text{skew}(·)\right)$	$10{ \text{log} }_{10}\left(\text{kurt}(·)\right)$
${\mathit{y}}^{(i)}$	$f_{10}^{(i)}={ \text{tan} }^{-1}\left(\text{skew}\left({\mathit{y}}^{(i)}\right)\right)$	$f_{15}^{\left(i\right)}=10 { \text{log} }_{10}\left(\text{kurt}\left({\mathit{y}}^{\left(i\right)}\right)\right)$
$\mathit{R}\left({\mathit{y}}^{\left(i\right)}\right)$	$f_{11}^{(i)}={ \text{tan} }^{-1}\left(\text{skew}\left(R\left({\mathit{y}}^{(i)}\right)\right)\right)$	$f_{16}^{(i)}=10 { \text{log} }_{10}\left(\text{kurt}\left(D\left({\mathit{y}}^{(i)}\right)\right)\right)$
$\mathit{L}\left({\mathit{y}}^{\left(i\right)}\right)$	$f_{12}^{(i)}={ \text{tan} }^{-1}\left(\text{skew}\left(L\left({\mathit{y}}^{(i)}\right)\right)\right)$	$f_{17}^{(i)}=10 { \text{log} }_{10}\left(\text{kurt}\left(L\left({\mathit{y}}^{(i)}\right)\right)\right)$
$\mathit{C}\left({\mathit{y}}^{\left(i\right)}\right)$	$f_{13}^{(i)}={ \text{tan} }^{-1}\left(\text{skew}\left(C\left({\mathit{y}}^{(i)}\right)\right)\right)$	$f_{18}^{(i)}=10 { \text{log} }_{10}\left(\text{kurt}\left(C\left({\mathit{y}}^{(i)}\right)\right)\right)$
$\mathit{T}\left({\mathit{y}}^{\left(i\right)}\right)$	$f_{14}^{(i)}={ \text{tan} }^{-1}\left(\text{skew}\left(T\left({\mathit{y}}^{(i)}\right)\right)\right)$	$f_{19}^{(i)}={ \text{log} }^{-1}\left(\text{kurt}\left(T\left({\mathit{y}}^{(i)}\right)\right)\right)$

Features are computed by starting with either the normalized (${\mathit{y}}^{(i)}$) or non-normalized (${\mathit{x}}^{(i)}$) data per for each epoch/channel combination i; c.f. Equations 1 and 2. Next, a transformation is applied (none, rectification R, line-length L, curvature C, or Teager-Kaiser Energy T; c.f. Equations 3–6), followed by a statistical measure (mean, variance, skewness or kurtosis). Lastly, a scaling transformation (arctangent or conversion to decibels) is applied. Rows in the table represent specific transforms to the data, and columns represent specific combinations of statistical measure and scaling transformation.