Pronunciation errors occur when non-native learners are not able to perceive how to pronounce specific phonemes since some phonemes do not exist in their own native language. The pronunciation errors by mistakes as previously mentioned may be found by the absolute threshold. However, the pronunciation errors caused by disparity between his own native language and the target language cannot be easily found by confirming a few sample speech. It is needed to investigate their speech statistically to detect such an inveterate pronunciation error. In this experiment, we aim to detect such kinds of pronunciation errors using the combined thresholds. We prepared task to detect such kinds of pronunciation problems: the P-set consists of eight types of pronunciation patterns that are difficult for non-native learners to pronounce their exact sounds.
