The researcher Wisam Hassan Ali from Electrical
Engineering Department at the University of Technology to build a
distinguish patterns of human speech system Visible Speech (VSR) in
order to distinguish the English alphabet letters and number twenty-six
characters from the words of the speakers with a higher rate for the
classification of those characters where they were initially propose a
new method for the selection of the best five images (frame), which
reflect the video clip, which represents the speaker, and then built a
new method to extract the lip clip of the speaker of those images with
high accuracy and without any distortion in the extracted section in
addition, the adoption of two new methods to determine the types of
(features) that represent lip movements, where the first represents the
geometric and calculate method , and the second represents a pictorial
method of binary form of lips.
The researcher, who received PhD proposing six kinds of works to build
several models (VSR) system which showed the simulation results of six
works the highest success rate has been obtained by using LMSNT and it
is 97.05%, followed by NVG RAM, which achieved a success rate of 94.67%
In addition, the researcher was able to conduct the practical
implementation of the work NVG-RAM on a physical device Spartan-3A DSP
3400A using logic gates programmable in situ (FPGA).
|
|