Speech Features Analysis and Biometric Person Identification in Multilingual Environment

Abstract-Biometric person identification systems are the important in the environments where security must be needed. In this respect, a Biometric person identification systems has been designed which identify the person by determining the authenticity by their voice in multilingual environment. The speech samples are recorded in three Indian languages Hindi, Marathi and Rajasthani for multilingual environment. Pitch, formant frequencies, MFCC and GFCC feature are extracted from the speech signals. For training and testing, neural network using radial basis functions are used. In this experiment accuracy of Biometric person identification 96.52% has been achieved.