Open AccessText Dependent Speakers Pattern Classification with Back Propagation Neural Network
Author(s) -
N K Kaphungkui,
Gurumayum Robert Michael,
Dr Aditya Bihar Kandali.
Publication year - 2019
Publication title -
international journal of recent technology and engineering
Language(s) - English
Resource type - Journals
ISSN - 2277-3878
DOI - 10.35940/ijrte.d8889.118419
Subject(s) - computer science , artificial neural network , speech recognition , sentence , backpropagation , mel frequency cepstrum , set (abstract data type) , identity (music) , speaker recognition , function (biology) , biometrics , artificial intelligence , pattern recognition (psychology) , natural language processing , feature extraction , acoustics , physics , evolutionary biology , biology , programming language
Speaker Recognition is the procedure of validating a speaker’s claimed identity using his/her speech characteristics which is unique to each individual. The primary objective of all speech recognition system is a man-machine interface which grants access into the system with the voice characteristics. This will served as a highly secure biometric system where security is the primary concern. The primary aim of this paper is to classify each speaker accurately with MFCC and Back Propagation Neural Network. Scaled conjugate gradient training function is used for back propagation neural network. A small database of 10 people is created from a group of five male and five female uttering the same sentence five times repeatedly. The sentence consists of five different words. The numbers of data set for classification is 22182.The accuracy obtained from the classification is 92.1% with small percentage of 7.9% misclassification which is acceptable good. The tool for simulation is MATLAB.