extremely significantp < 0.001
The results of the paired t -test for the recognition accuracy showed that, with single extraction, SepCNN had an extremely significant improvement compared with all classical algorithms [LDA: 76.34% ( p < 0.001, t = 4.72), SVM: 68.58% ( p < 0.001, t = 7.91), BLDA: 53.23% ( p < 0.001, t = 14.55), SWLDA: 50.62% ( p < 0.001, t = 18.44)].