Figure 7
From: Silent speech command word recognition using stepped frequency continuous wave radar
Confusion matrices for all recognition experiments. (a) multi-session, subject 1. (b) multi-session, subject 2. Rows with perfect classification results are omitted here for clarity and brevity. (c) inter-session, subject 1 and 2 combined (upper and lower triangles, respectively) across all three splits with the respective feature set that achieved the highest mean inter-session test accuracy, i.e., feature set 5 for subject 1 and feature set 6 for subject 2. Reported are the results where the validation set is a subset of the training set, as discussed in the text. Non-empty cells are shaded from white (zero occurrences) to black (maximal number of occurrences).
