FIGURE 6.

(A) Sensitivity and specificity obtained from intra-subject RF classifiers; ‘full’ refers to models trained with all features—mean and variance of $a_x$ , $a_y$ , $a_z$ , mean, variance, maximum, minimum and entropy of ${‖a‖}_2$ (11 features), ‘mean of $a_x$ ’ are models trained with only the one feature, ‘all mean’ refers to models trained with mean of $a_x$ , $a_y$ , $a_z$ (3 features), ‘all mean and variance’ are models trained with mean and variance of $a_x$ , $a_y$ , $a_z$ (6 features), (B) Sensitivity and specificity obtained when certain tasks were removed from the train and test sets. Changes in sensitivity were found to be statistically insignificant in all tasks except in intra-subject models for opening bottle task for the left hand (F = 6.05, p = 0.0005). Task labels ‘openbottle’ corresponds to ‘Open a bottle’, ‘drinkcup’ corresponds to ‘Drinking from a teacup with handles’ and ‘walk’ corresponds to ‘Walk 25 m’, ‘Hit a light switch’ and ‘Open a door’.