Paper Title
Text Dependent Speaker Recognition
Abstract
Now a days biometric person authentication is secure way to authenticate or recognize a person. Speech is one
of the easiest and safest method among all biometric model. Speaker recognition can divided in to text dependent and text
independent recognition. Feature extraction is the first process in any speaker recognition technique. This paper describes
text dependent speaker recognition In this, Mel Frequency Cepstral Coefficient (MFCC) have been used for feature
extraction. Along with this pitch and formants are also extracted. Gaussian Mixture Model (GMM) and Artificial Neural
Network (ANN) are used for modelling and speaker matching process. GMM is parametric model of probability distribution
of features in speaker recognition process. Compared to GMM, ANN is seen to be best for recognizing a person. It is
unaffected by differing shape and style of testing speech. False Acceptance Rate (FAR) in ANN is small compared to ANN