Analysis of statistical parametric and unit selection speech synthesis systems applied to emotional speech