Figure 7
Distributions of KNN-DTW classifier scores. We use Kernel Density Estimation (KDE), a non-parametric smoothing method, to estimate the probability densities based on finite data samples. We also show the threshold values that separate the two classes yielding an optimal F1 score.