Kurdish Speaker Identification Based On One Dimensional Convolutional Neural Network

Speech recognition is a vital in human- computer interaction area. In this paper, two different models are implemented for speaker identification which are 1 D convolutional neural network (CNN) and feature based model. In the experimental, the three spectral based features including Mel frequency Cepstral Coefficient (MFCC), Linear Prediction Code (LPC) and Local Binary pattern (LBP) are fed to Support Vector Machine (SVM) and K Nearest Neighbor (kNN). results are shown the MFCC is the best feature among the others. MFCC is then extracted for framed signal and used to both proposed models. The result shows that there is not significant different between the accuracy of the CNN and kNN. While SVM is less sensitive compared with others.