الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 85 |  |  |
| | | from | 85 | | |
Abstract
The quality of speech is essential in real-time communication systems. There are many
aspects that can affect the speech quality like the signal processing that is included in
the speech coder and the error concealment methods that are required to recover the lost
frames. When an error in frame occurs, this will degrade the quality of the speech and
cause a disturbance on the other side of the call. Enhanced Voice Services (EVS) is one of
the speech Coders that standardized by the 3GPP for LTE network. The EVS codec offers
features and improvements over the standard 3G coders. EVS supports Narrowband,
wideband, super-wideband, and full-band so it covers the full range of human hearing,
offers high quality for mixed content (i.e. speech and music) and provides low latency.
EVS coder has 11 different fixed bit-rates and one variable bit-rate, and also supports
9 AMR-WB bit-rates to interoperable with AMR-WB. EVS has the ability to change
the bit-rate at every 20 ms frame adapting to channel capacity and data content. Many
coders that depend on the concept of ACELP may be affected by the change in the
spoken language or accent or packet loss. EVS coder is working with both ACELP and
MDCT coding techniques and has a seamless switching between the two modes. EVS has
several error concealment techniques to deal with different coding modes achieving higher
robustness against transmission errors. In this thesis, the performance of EVS coder was
evaluated with the clean and interrupted speech compared with other standard coders.
EVS coder was evaluated with the Arabic language (i.e. classical Arabic) compared to
the English language (i.e. American and British accent) using a large number of speech
files about 16,900 files. And EVS-NB was compared with AMR-NB and G.729 coders
using the same databases of speech files. It was shown that for the Arabic database, EVS
coder is better than both G.729 and AMR-NB respectively, in terms of average speech
quality and stability, while EVS coder gives the best performance with the American
English database compared with Arabic and British databases. The other contribution
that the thesis presents is the performance of EVS coder against transmission errors,
EVS-NB was evaluated under error conditions compared with AMR-NB through losing
frames or zero-padding (inserting zeros instead of frame content) using Arabic and English
databases. The results showed that EVS achieves better performance than AMR in case
of zero-padding frames by 8% and 17% where the frames were conducted continuously
and intermittently, respectively. On the other hand, by losing frames, both coders are
relatively close to each other in terms of quality.