Algorithm 2: A-SVM with ORICA-CSP.

Step 1: Input features vectors ${\mathrm{f}}_{\mathrm{v}}=\log ({\mathrm{V}}_{\mathrm{m}}/{\displaystyle \sum }_{\mathrm{t}=1}{\mathrm{V}}_{\mathrm{m}})$ $//{\mathrm{V}}_{\mathrm{m}}$, variance matrix of EEG signal projection Step 2: Determine the class label with function $\mathrm{f}({\mathrm{s}}_{\mathrm{i}}$$)=(\mathrm{w}, {\mathrm{s}}_{\mathrm{i}}$) + b $//{\mathrm{s}}_{\mathrm{i}}$ ∈$\mathrm{Rn}, \mathrm{with} \mathrm{N} \mathrm{samples} \{I$$,\dots , {\mathrm{s}}_{\mathrm{N}}$ } Step 3: Classify $\mathrm{the} \mathrm{new} \mathrm{sample} \mathrm{by} {\mathrm{u}}_{\mathrm{j}}$ $\mathrm{and} {\mathrm{v}}_{\mathrm{j}}$ //${\mathrm{u}}_{\mathrm{j}}$$=\mathrm{sign}\left(\mathrm{f}\right({\mathrm{s}}_{\mathrm{i}}$$\left)\right) \mathrm{and} \mathrm{the} \mathrm{posterior} \mathrm{class} \mathrm{probability}, {\mathrm{v}}_{\mathrm{j}}$$=\mathrm{Pb}({\mathrm{v}}_{\mathrm{j}}$$={\mathrm{v}}_{\mathrm{j}}$$|{\mathrm{s}}_{\mathrm{i}}$) is calculated using Platt’s probabilistic output. $//\mathrm{Classifier} \mathrm{is} \mathrm{adapted} \mathrm{with} {\mathrm{u}}_{\mathrm{j}}$ $\mathrm{and} {\mathrm{v}}_{\mathrm{j}}$ Step 4: Define threshold th Step 5:$\mathbf{if} {\mathrm{v}}_{\mathrm{j}}$ $> \mathrm{th} \mathrm{holIs}, {\mathrm{s}}_{\mathrm{i}}$ is introduced to the dataset for training T Step 6: Update whenever new samples are included in the solution