Automatic Language Identification Using the Frequencies of Occurrence of Phones

Phonetic inventories differ from language to language. Even when languages have identical phones, the frequencies of occurrence of phones differ across languages. It's difficult to introduce new languages when the language identification system used phones label. The frequencies of occurrence of phones were trained by Gaussian mixture model and vector quantization. The method of occurring of phones and three improved methods were compared. The results show that the frequencies of occurrence of phones are effective in language identification. The performance of GMM is better than the performance of VQ. And the third modified method of joint-usefulness is better than the other methods