Figure 4From: Speaker-dependent model interpolation for statistical emotional speech synthesis The 5 × 5 Graeco-Latin square used in the evaluation of emotional expressiveness of the synthesized speech. Back to article page