Fig. 6

The results of the on-line evaluation. a) Screen-shots of two examples of the on-line implementation showing an activation of the prosthetic device during the second phase. The graph presents the confidence of the majority vote with the elbow’s angular velocity profile. b) the average time and standard deviations until the confidence threshold was reached for the correctly classified trials. The pattern recognition system exceeds the confidence threshold of 0.5 significantly faster (p=0.019) when trained including all the phases of the motion. c) the classification accuracy of the testing phase of the on-line evaluation. The pattern recognition system presents better accuracy when trained including all the phases of the motion