Feature fusion techniques based training MLP for speaker identification system
Najiya M. Omar, M.E. El-Hawary · 2017
This paper aims to compare the Linear Predictive Cepstral Coefficients (LPCC) method, the Mel-frequency Cepstral Coefficient (MFCC) method, their concatenation (LPCC-MFCC), and a new proposed feature fusion approach based on method involving this concatenation with the respective averages normalization; Linear predictive and Mel-frequency Cepstral Coefficients (LMACC) through applying a multi-layer perceptron (MLP) neural network as classifier for speaker identification system (SIS). The evaluation was made based on classification accuracy. After evaluation, the results of the proposed system LMACC-MLP were verified using Cochlear implant-like spectrally reduced speech (SRS) algorithm proposed in the literature so that the original recorded signal was resynthesized based on an acoustic simulation of the cochlear implant. i.e. human speech perception. Our proposed system demonstrated maximal relative performance in environments on measures of recognition rate, compared with other methods and covering a range of different root- mean - square (RMS) noise amplitudes.