Synthesis by rule of voice fundamental frequency contours of spoken Japanese from linguistic information.