New features for language recognition from speech signal

In this paper, we derive new feature vectors for identifying the language from short utterance of speech of an unknown person. By applying window technique on speech signal, Mel-frequency cepstral coefficients (MFCC) and formants of speech are extracted. With these two kinds of features, we derive a new feature set using cluster based computation. Later a classifier is designed one for each language using the new features vectors and applied on recognition output with a specific apriori knowledge. We use OGI database to perform the experiments and achieved good recognition performance.