This may degrade the system performance. For solving this problem, we must explore optimal pronunciation style and train the subjects adopting identical speech manner. Finally, we shall endeavor to make the user interfaces more powerful and friendly.