الفهرس 
الفصل الاولOnly 14 pages are availabe for public view
 |  | from | 97 |  |  |
| | | from | 97 | | |
Abstract
Automatic speech recognition (ASR) is a technology that allows a computer to recognize the words that a person speaks into a microphone or telephone. It has a wide application area. It can be used in command recognition (voice control interface with computer) and dictation. It can help handicapped people to deal with society. It can help in making life easier and very promising.
Therefore it is very important to enhance the recognition accuracy of the Arabic spoken letters. The accuracy of recognition system is affected by the feature extraction and the used classifier. Robust and an effective method is introduced to evaluate speech feature to improve the performance the recognition accuracy. This thesis introduces applying the mel frequency cepstral coefficient to extract the speech features to enhance the speech recognition accuracy. Hidden Markov model is used as a classifier. The objective of the proposed system is to enhance the performance by introducing different systems with different features. Three systems are proposed to recognize the spoken Arabic letters. This is done in three phases. The first phase is based on neural networks. The second is based on hidden Markov model. While third phase based on combination between neural networks and hidden Markov models. A comparison between different feature extractions is given. The accuracy of neural network is found to be 42% with MFCC for 84 spoken letters. It is found to be 46.33%, 31.4% and 84.31%for cepestrum, LPC and energy per frame respectively. This is done for 28 Arabic spoken letters. But with grouping the spoken letters, the accuracy of neural network reached to 99% with energy per frame for 28 spoken letters. The hidden Markov model based is on probabilities. Its performance is found to be 98.5%. But for combination system based on neural network and hidden Markov models, the accuracy of 99.25% is obtained.