An Automatic Language Identification Using Audio Features

An automatic Language Identification (LID) is the task of automatically recognizing a language from the given spoken utterance. Language identification is used to identify the language of the particular audio and reduce the complexity of the audio sample. LID systems that rely on multiple language phone recognition language modeling (PRLM) and n-gram language modeling produces the best performance in formal LID evaluations. By contrast, Gaussian mixture model (GMM) systems, which measure acoustic characteristics, are far more computationally efficient but tended to provide inferior levels of performance. We have described here the efficiency of an LID system for two different languages namely English and Hindi. The evaluation of languages is done on the standard recorded databases, from which features are extracted using Mel-frequency cepstral coefficients (MFCC). The language models are done using PRLM and classification is done using Gaussian mixture model (GMM). The obtained results ensure that accuracy of LID is efficient for the chosen languages and the system performance is evaluated on both PRLM and GMM. Keywords-Language Identification, PRLM, GMM, MFCC accuracy.