Speaker Identification in Odiya using Mel Frequency Cepstral Coefficients and Vector Quantisation

Automatic Speaker Identification technology has recently been implemented in several of commercial areas successfully. Speaker identification comes under Speaker recognition and is gaining significance for voice based biometrics. It is used in appliances that understand voice commands, provides security to confidential information,etc. In this paper we have built reference model for each speaker using the acoustic features .Testing has been done by comparing the features of the test sample with the reference model.We have used MFCC technique for feature extraction .For creating the reference model Vector Quantization( VQ) with LGB (Linde, Buzo, and Gray)was used. For testing VQ distortion was used.The system achieved 85% accuracy