Speaker Recognition Using Marathi (Varhadi) Language

In this paper, we used Mel Frequency Cepstral Coefficients (MFCC) and Linear Prediction Coding (LPC) used for features extractions. We applied these two techniques for identification of a person on two databases such as PHASE-I, & PHASE-II and the distance is calculated between training phonemes and testing phonemes by using Euclidian Distance method. For our experiment we used total 1625 phonemes 25 subjects from PHASE-II speaker database. The database consists of phonemes in Varhadi language spoken in Vidharbha region of Maharashtra. Collection of phoneme is done in Akola district. The phoneme for each subject collected five times i.e. may, nav, hay, and the name are trained using 10 iteration, 40 subjects from PHASE-I The phoneme for each subject collected five times i.e. may, nav, hay, and the names are trained using 10 iteration.