A feasibility study using Chinese speech as a command/control tool for computer systems.

Abstract

This thesis examined whether American English speech recognition technology can be used by Chinese speakers, in their native tongue, to achieve a reasonable degree of recognition accuracy. Three experiments were completed. The first showed that 88.25% of 4305 trials of Chinese phoneme recognition was correctly recognized. The second showed that 74.67% of 900 trials of simulated speaker independent mode Chinese utterance recognition was correctly recognized. The third showed that 12.44% of 900 trials of speaker dependent mode Chinese utterance recognition was incorrectly recognized on the first attempt. Only 16 utterances required a retraining to eventually obtain a correct recognition.