Skip to main content
Figure 5 | BioMedical Engineering OnLine

Figure 5

From: An approach to improve the performance of subject-independent BCIs-based on motor imagery allocating subjects by gender

Figure 5

Results of “leave-one-subject-out” method used for validation of SI classifiers. This figure displays the validation results from the total of 30 subjects collected for SI-BCI designs. The performance of the classifiers is validated for the 3 tested conditions, this means, LEFTMI vs. RIGHTMI, LEFTMI vs. REST and RIGHTMI vs. REST. a) ACCs for the LDA classifier designed with data of 10 trials per class from 15 male subjects (BCI-Men). M1 to M15 represent which subject’s data is left out from the classifier training set. b) ACCs for the classifier assembled with data of 10 trials per class from 15 female subjects (BCI-Women). Similarly, F1 to F15 represent which female subject’s data is left out from the classifier training set. c) ACCs for the classifier assembled with data of 10 trials per class from all 30 subjects (BCI-All). Labels 1 to 15 are the male subjects and from labels 16 to 30 are the female subjects. d) Averaged performance for the validation stage of the proposed 3 classifiers (BCI-MEN, BCI-WOMEN and BCI-ALL) for all 3 tested conditions. The dashed line indicates the practical chance level. The asterisks (*p < 0.05 and **p < 0.1) indicate statistically significant difference between gender-specific BCIs and BCI-All.

Back to article page