This paper describes the mechanical design and the speech planning of an anthropomorphic talking robot WT-1R (W__-aseda T__-alker-No. 1__-R__-efined) for the production of consonant sounds. In 2000, we mechanically produced Japanese vowels (/a/, /i/, /u/, /e/, /o/) using a first robot WT-1 (W__-aseda T__-alker-No.1__-). In 2001, we developed a second robot WT-1R, which was improvement over WT-1's mechanisms for producing natural vowels and consonant sounds. WT-1R has articulators (the tongue, lips, teeth, nasal cavity and soft palate) and vocal organs (the lungs and vocal cords). The total DOF (degrees of freedom) is 15. Furthermore, we proposed the speech planning of WT-1R by considering the complicated phenomena of consonant sound as three parts (steady consonant sound, transient consonant sound and vowel). WT-1R could mechanically produce Japanese vowels (/a/, /i/, /u/, /e/, /o/), some consonant sounds (/s/, /h/, /m/, /p/) and continuous sound /waseda/.
