Language Identification Based On Gaussian Mixture Models

Language identification is a kind of technology of identifying the language of an utterance automatically by using a computer, whose development is based on speech recognition. Being one of the aspects of speech recognition, with the development of the technique of speech recognition, language identification is paid more and more attention for its significance. From the seventies of the last century up to now, though it is just several decades, many kinds of ways of language identification with their own characteristics have already come into being, most of which are not mature. At present, the research into language identification in our country is still in its beginning stage and is less extensive.In spite of some similarities between language identification and traditional speech recognition, there are still discrepancies. Language identification is accomplished under the condition of text-independence and speaker-independence, thus it is necessary for language identification to eliminate the individual information of the signal of speech sound of different languages as far as possible so as to achieve a better result of recognition.In this paper, attention is focused on establishing the Gaussian mixture models by which Mel frequency cepstrum coefficient of the speech of each language in order to distinguish languages. Therefore, discussion about the process of language identification need to be carried out in the following two aspects: On the one hand, we start with acoustic features of speech sound and then make a further analysis of different feature information of different languages, thereby finding the basis for language identification. On the other hand, we discuss the principle and algorithm of Gaussian mixture models technology and so on. Simultaneously, we analyze the problems of Gaussian mixture models emerging during the course of identification and offer the solution.In this paper, we also make a thorough analysis of factors which affect the identification performance of system through specific experiment and then provide a summary. Finally, we point out what should be improved in this field.