Non-native speech recognition based on speaker adaptation

Speaker adaptation is a powerful means of improving the performance of speaker-independent non-native speech recognition system. Based on Yunnan minority Naxi and Lisu accent speech, non-native mandarin speech recognition was implemented applying general speaker adaptation MLLR, MAP and multi-pronunciation dictionary adaptation. In our experiments, the different configuration of feature and methods is discussed. These approaches have been shown to be effective by the experimental results.