الفهرس 
chapter 1Only 14 pages are availabe for public view
 |  | from | 16 |  |  |
| | | from | 16 | | |
Abstract
Speech is the main communication method between human beings. However Speech recognition has tremendous growth over the last five decades, speaker and speech recognition are still a challenging and difficult task when the systems are applied in the real world. The two major components of speech/speaker recognition systems are Feature extraction and classification. The most widely used feature extraction techniques in the field of <speech/speaker recognitions are Linear prediction coefficients (LPC), Mel <Frequency Cepstrum Coefficients (MFCC), Perceptual Linear Predictive <(PLP), Relative Spectra (RASTA) with PLP (RASTA-PLP), and Wavelet
Transform (WT) especially Discrete Wavelet Transform (DWT).
Two proposed methods have been implemented in this thesis. For
both methods the feature extraction process is done by combining the
Discrete Wavelet Transform (DWT) and Relative Spectra Perceptual
Linear Predictive (RASTA-PLP) only in the first system, while in the
second system, Radon Transform (RT), which is an image processing
technique, is applied to the feature extracted using DWT and RASTA-
PLP (of the first method which can be considered as an extra feature
extraction step).
The classification or matching process is done in the first system
using a 3 layered feed forward back propagation Neural Network with
1728 input neurons, 50 hidden neurons and 15 output neurons, while in
- iii -