M. Hamed
Dalia Wafik
Faculty of Engineering
and
Higher Technological Institute
Port Said 42523, Egypt
10th of Ramadan City, Egypt
AMULTI-SPEAKERSYSTEMFOR
ARABICSPEECHPERCEPTION
ABSTRACT
suitable
for
Arabic tool for training in the field
This paper presents a speech recognition. The of classification of words.
new generalized system multi-speaker signals are This process, even with
for the Arabic speech tested and trained for pattern recognition in the
recognition
based
on recognition. They were case of image processing
3-layer integrated neural classified into three groups will be the most important
networks
consisting
of as male, female and in the next years. Also, the
control
network
and children groups. The effect more generalized system
several sub-networks. The of number of nodes in the will take the advantage
segmentation technique hidden layer of the neural relative to others. So,
is accounted to minimize network on the rate of special experiments for
the selected features, recognition is investigated. the applicability of the
and
consequently, Key words: Arabic, multi- neural network (NN) to
the
computational speaker,
recognition, recognize multi speakers
time effort. The back segmentation,
features. as selected in Table
propagation
and
the
1 where three various
linear predictive coding INTRODUCTION
groups of speakers: male,
bases are considered in However,
the
single female,
and
children
the tested multi-speaker speaker system is a quick are introduced [1, 2].
system.
The
proposed
concept is applied for
words representing the
digits. The recognition rate
is quite accurate, and
so It is recommended to
be used in the software
Research
33

---