Abstract: A testing database used for Chinese human-machine communication system is designed to include all Chinese properties, linguistic phenomena of really spoken Chinese. The material used for the database is selected according to the tone model of syllable, word, sentence and text, concerning all initials, finals and their combination as well, that characterizes the tone rules and the variation of acoustic properties of standard Chinese. Both synthesis rules and recognition model at test stage can be extracted from this database. So far the database pronounced by a male speaker and a female speaker, has been recorded in a digital audio recorder.