Figure 5From: Speaker-dependent model interpolation for statistical emotional speech synthesis The 5 × 5 Latin square used in the evaluation of naturalness of the synthesized speech. Back to article page