Computationally efficient speaker identification system using AMDF and LPC

Passwords are the first line of protection in any security system, be it a bank account, an email account or a facebook account. But typed passwords can be a liability in most situations as they can be stolen or guessed easily by imposters. The advantage of a biometric security system over typed password system is that the access is granted on the basis of a biologically unique attribute like pitch. The aim of this paper is to present the concept of a faster, more efficient text dependent speaker recognition system based on LPC (Linear Predictive Coding) which is further supplemented by pitch period estimation based on Absolute Magnitude Difference Function (AMDF) and Dynamic Time Warping (DTW).