Paper Title
Feed Forward Back Propagation Neural Network For Speaker Independent Speech Recognition
Abstract
Speaker recognition is to make sure that the person to be claimed is the correct person or not. Speaker recognition
can be divided into speaker identification and speaker verification. Speaker identification decides whether the speaker is
from the specified group. In speaker verification, a person makes an identity claim. The two main stages in this technique are
feature extraction and feature matching. Feature extraction is the process in which we extract some useful parameters, which
is used to represent the speaker. Feature matching involves identification of the unknown speaker by comparing the features
extracted of the unknown speaker with the enrolled voices of the known speakers. To extract the features of speech signal,
Mel frequency Cepstral Coefficients (MFCC) and for feature matching the multi-layer perceptron method in artificial neural
network (ANN) has been used. A Feed Forward Back Propagation Neural Network (FFBPNN) is used to classify the voices
of various speakers in the learning or training phase. The network is tested with samples from the various speakers. During
the learning phase many parameters are tested with, number of selected coefficients, number of hidden nodes and value of
momentum parameter. In the testing phase the recognition performance is computed for each value of the above
parameters.