DHMM Based Automatic Language Identification System

This paper focuses on the implementation of automatic language identification system (LID). Automatic language identification is a task to classify unknown utterance of speech into listed languages. LID is implemented using discrete hidden Markov models (DHMM). This system involves two phases and they are training phase and testing phase. In training phase, a common code book of MFCC features is created from huge speech corpus of all listed languages. Language specific DHMMs are created one for each language. In testing phase, MFCC feature are extracted from unknown speech and evaluated against each created DHMM. The language is hypothesized as identified language based on the likelihood value of sequence of feature vectors of unknown speech. The OGI database is used for the study. Even though we have used simple and easy method, the results are very impressive.