نظام تحويل آلي لنماذج من الكلام العربى المتواصل إلى نص

الموضوعات

علوم تطبيقية وتكنولوجية

روابط وملفات

التعريف والنوع

رقم الوثيقة

efe90eec-46fc-4979-957d-fe50d02d9399

رقم العقد

0

نوع الوسائط

Crawler

نوع المحتوى

الرسائل العلمية

صيغة المصدر

رسائل ماجيستير

نوع الملف

pdf text

أسماء الملفات

1965358_1.pdf

بيانات النشر

ترجمة العنوان

Automatic Conversion System for Samples of Continuous Arabic Speech to Text

ألقاب المؤلفين

[{"name_ar":"بيــــان رضوان ســــــالم","title_ar":"اعداد","title_en":"Preparation"},{"name_ar":"حليم علي","title_ar":"اشراف","title_en":"Supervision"},{"name_ar":"يحيى فريد ","title_ar":"اشراف","title_en":"Supervision"}]

اللغة

Arabic

المصدر والدورية

اسم المصدر

نظام تحويل آلي لنماذج من الكلام العربى المتواصل إلى نص

المحتوى والصفحات

عدد الصفحات

0

ترجمة الملخص

Many researches have focused recently on the speech recognition technique due to its importance in many applications as control electrical devices through voice commands, customer service centers, automated voice-data entry and interpretation, as well as helping people with special needs. Where speech recognition is defined as the conversion of the captured audio signal via the microphone or telephone to a series of written words. Although Arabic is one of the most widely spoken languages across the world, however, The research in the field of continuous automatic speech recognition is rare and the use Of this technique is limited to some large businesses companies, which do not allow the Arabs programmers to take advantage of their systems in the development and construction of private vocal graphical user interfaces. The aim of this research is to build a high performance, speaker-independent Arabic speech recognition system using Hidden Markov Models. We discuss the steps for preparing data needed for training and building this system. We determine the phonetic set for Arabic language and build a phonetic dictionary, then a limited vocal database recorded by 20 persons is built and preprocessed by windowing and extracting the features using Mel Frequency Cepestral Coefficients (MFCC). A3-gram statistical language model is developed .In addition to the stages of the building and the training of hidden Markov model to get a triphone-based acoustic model and then the decoding phase was implemented as well as building a simple application of the actual use of the proposed system. All mentioned above steps are performed using Carnegie Mellon University (CMU) Sphinx tools.

إشراف وإعداد

الإشراف

حليم علي, يحيى فريد

الإعداد

بيــــان رضوان ســــــالم

الاقتباسات الببليوغرافية