Spoken Indian Language Classification using GMM supervectors and Artificial Neural Networks
Aarti Bakshi, Sunil Kumar Kopparapu · 2019
Indian languages are phonetic in nature; phonetics is branch of linguistics which studies the structure of human language sound. Acoustic phonetic features associated with languages play an important role in spoken language identification. In this paper, Gaussian Mixture Model supervectors is used to capture acoustic phonetic variation in Indian languages. Mel frequency cepstral coefficient (MFCC) with delta coefficients is used to represent the language specific acoustic phonetic information of speech and artificial neural network ANN is used as a classifier for language identification. In the present work, we have conducted extensive experiments for three different datasets created from the news broadcast in different Indian languages from All India Radio. The performance of ANN classifier using GMM supervectors is evaluated on these three datasets.