Table 1.

Three-class classification accuracy obtained using different feature extraction techniques on frequency bands.

Method	$\mathit{\delta }$	$\mathit{\theta }$	$\mathit{\alpha }$	$\mathit{\beta }$	$\mathit{\gamma }$	ALLBAND
DE	60.80%	47.41%	57.07%	88.09%	95.09%	88.28%
RASM	53.07%	49.56%	60.49%	88.53%	93.12%	90.62%
DCAU	59.79%	55.15%	64.02%	91.31%	95.12%	94.70%
DASM	57.44%	52.54%	63.58%	91.41%	95.87%	94.34%
AsMap+CNN	62.18%	56.20%	69.56%	93.99%	97.10%	96.25%

The window size was set to 3 s.