الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 170 |  |  |
| | | from | 170 | | |
Abstract
In today’s all modern communications, speech coding plays a very vital role. Beside its basic technical functions it works as the economy room in the operation of voice transmission. All mobile generations, VoIP as well as all other communication technologies try to make higher rates of transmission and efficient use of the available BW. The speech coding is a basic element in the efficient use of BW issue. Simply, while one type of waveform coding can use about 1 Mbps for only 3 to 4 voice calls, using the LPC (Linear Prediction Coding), we can reach from about 50 to 60 voice calls for the same BW with approximately the same quality. This great benefit forces many standard bodies to generate a series of Speech coders with different rates, quality and delay combinations. However the choice between these CODECs depends on those relations for a certain network. All these CODECs built initially for 7 languages not including the Arabic or Cairo accented Arabic. So what will be the performance when the language or the accent is different from English? In this thesis an extensive testing were made on one of the most used CODEC on modern communication which is the ACELP ITU G.723.1 and compared with the wave form CODEC 𝜇_law ITU G711. The test is done for more than 200 voice samples (English, Cairo accented Arabic and Arabic) from different speakers (male and female) and the assessment were developed using the PESQ algorithm which will be converted to MOS scale.
We conclude that the quality will be degraded slightly when using other language or accent rather than English language. The main point is not only the slight degradation but the higher standard deviation of the language and
Digital Speech Communication Coding for Arabic Language IV
[Type the company name] 2013
accents compared with English language, i.e. the more stability with English. Additional investigation on the problem was done to answer the question of what will happen if the speech is compressed more and at the same time the spoken language changed. An extensive testing is done on ITU G.723.1 (5.3 kbps) and 3GPP Adaptive Multi Rate “AMR” (12.2 kbps). It has been proved that the speech quality dramatically decreases as the compression ratio increases, which will be rather lower and unstable for Arabic and Cairo accent over that of the English language. With more efforts it has been found that the basic problem in the LP vector quantization CodeBook. We tried and succeeded to put an initial step in the development of a new CodeBook using LBG and K-Means algorithms that helps in the improvement of the voice quality in Cairo Arabic accent.