Non-native speaker identity verification based on ICA

Speaker identity verification is an useful biometric recognition approach. Native speaker verification has achieved some important progress. But non-native speaker verification remains a challenging task because of wide varieties of non-native accents. Based on two speech corpus, one native speech corpus and one non-native corpus, by means of speaker adaptation, non-native speaker identity verification experiments are performed. In our experiments, the traditional speaker verification method combines with the non-native speaker´s independent component speech features. The speaker features of ICA are showed to be more effective by the final experimental results.