Abstract:
Automatic speech recognition (ASR) plays an important role in taking technology to the people. There are numerous applications of speech recognition such as direct voice input in aircraft, data entry and speech-to-text processing. The aim of this paper was to develop a voice-learning model for correct Arabic letter pronunciation using machine learning algorithms. The system was designed and implemented through three different phases: signal preprocessing, feature extraction and feature classification. MATLAB platform was used for feature extraction of voice using Mel Frequency Cepstrum Coefficients (MFCC). Matrix of MFCC features was applied to back propagation neural networks for Arabic letter features classification. The overall accuracy obtained from this classification was 65% with an error of 35% for one consonant letter, 87% accuracy and an error of 13% for 10 isolated different letters and 6 vowels each and finally 95% accuracy and an error of 5% for 66 different examples of one letter (vowels, words and sentences) stored in one voice file
