I. Introduction
Spoken language identification (LID) is a task to identify the language spoken in a given speech utterance. For most of the multilingual speech based applications, the spoken LID system is used as a pre-processing block [1]. The spo-ken LID task can be performed by using acoustic-phonetic and phonotactic approaches. The acoustic-phonetic approach uses spectral information from the speech signal, whereas the phonotactic approach uses the phoneme distribution to model the classifier [1]. Earlier studies shows that the phono-tactic approaches perform better than the acoustic-phonetic approach. But the requirement of the transcribed data (which is difficult to obtain) [2] limits the use spoken LID systems using phonotactic approach. Therefore the research community paid enormous attention to develop spoken LID systems with different acoustic-phonetic based approaches.