Design of the Intelligent Agent for Speech Recognition based on Web Application

birthdaytestΤεχνίτη Νοημοσύνη και Ρομποτική

17 Νοε 2013 (πριν από 3 χρόνια και 6 μήνες)

70 εμφανίσεις

Design of the Intelligent
Agent
for Speech Recognition based on Web
Application



Emrul Hamide

bin

Md Saaim
*1
, Ag. Noorajis bin Ag. Nordin
*2
, Abdul Manan bin Ahmad
*3

Speech Recognition Laboratory,

Department of Software Engineering,

Faculty of Computer Sci
ence and Information Systems,

University of Technology Malaysia

81310, Skudai, Johore, Malaysia

Tel: +60
-
7
-
553
-
2201, Fax: +60
-
7
-
556
-
5044



Abstract


This paper tales info account the design aspects of intelligent
agent working with speech recognition base
d on web
applications.
The paper encourages the use of agent technology
in comparison to the traditional client
-
server methodology.

We
highlight clearly the advantages of mobile agent technology
based on web applications

and agent design patterns that we u
se
to develop mobile agent.


1.0

Introductions


With the explosive growth of the Internet technology, both
speech researchers and computer software engineers have been
putting a great deal of effort to integrate speech function into
Internet applications [
Z. Tu and P. Loizou
, 1999

].

Many of web
applications based on internet
-
based still using the old client
-
server paradigm.
For simple applications like recording voice,
ordinary, client
-
server paradigm will be applied. By using client
-
server model, all the
process will be handled completely by
server computer and client computer.


For speech recognition system based on web application, we use
the new mobile agent methodology with more advantages and
make the system more efficiency.
Mobile agent concept is
di
fferent from client server model especially in transmit and
receive the data between the computer through the network.


2.0

Background


This paper presents an intelligent agent, speech recognition
system and web applications.
The description of the compo
nents
can be found in the succeeding portion of the paper.

----------------------------------------------------------


1
email:emrulhamide81@hotmail.com


2
email:
ansanjis@hotmail.com


3
email:manan@fsksm.utm.my















2.1

Mobile Agents

versus client
-
se
rver
.


Mobile agents are programs that can move through a network
under their own control, migrating from host to host and
interacting with other
agents and resources on each
. A concept
of mobile agents always been applied in a network computing
such as ,
distributed systems, information retrieval, network
managem
ent or other else.








Mobile agent

concept

has several advantages if comparing with
client server paradigm.
Internet based speech recognition using
client server paradigm may be it is not a good idea because

of
several problem such as quality of network connection and there
is just one server to handle many client

that can make the server

computer

has a lot of job
.

Beside that, client server paradigm
still using
Remote Procedure C
all
(RPC)
technique to
commun
icate between client and server, which that technique
may not be well suited for
large
internet based applications.

By
using RPC, data

move from the client to a procedure that already
resides on the server (and the client usually remains idle while
the rem
ote procedure still executed)

[
Stefan and Friedemann,
1999].
However, by using mobile agent technology, in an agent
-
based framework the client dispatches an agent, which travels to
the server and performs its task there by interacting locally with
the serv
er’s resources.

Compared to traditional distributed
computing schemes, mobile agents promise (at least in many
cases) to cope more efficiency and elegantly with a dynamic,
heterog
eneous, and open environment which is characteristic for
today’s internet [St
efan and Friedemann, 1999].


Figure 1 shown client
-
server paradigm architecture. Client
-
server paradigm execute synchronously, need
multiple servers to
communicate, sometime have a problem in network bandwidth.



Figure 1:

Client
-
server paradigm


On the o
ther side
mobile agent paradigm, every node is a server.
So, if there is a task, it can be handle not by one server. Beside
that, it executed asynchronously. Therefore, if something happen
to the network, mobile agent still keep the task and continue
again

that task after the network is ready.




Figure 2
: Mobile agent paradigm



Hence, mobile agents are able to emulate remote procedure
calls, but more importantly, they also allow for much more
flexible and dynamics structures than traditional systems base
d
on the client
-
server paradigm

[
Stefan and Friedemann, 1999
]
.
Compare to lower level mechanism such as RPC, the use of
mobile agents for distributed applications often reduces
bandwidth required for the application, and improve
performance. Beside that, i
t also has another potential benefit
s

such as mobile agent concept will execute its task as
asynchronous. Its mean, while the agent acts on behalf of the
client on a remote site, the client may perform other tasks.



2.2

Agent Design Patterns


A mobile age
nt is not a
confined

to the system in which it begins
executions, but it has a unique ability to transport itself from one
system in a network to another [
Aridor and D. Lange
, 1998
].
This ability to travel allows a mobile agent to move to a system
to assig
n a task, to interact with other agent to get information or
data and etc.


For creating mobile agent applications, it has a set of agent
design pattern itself. Generally, the patterns can conceptually
divide into three classes: travelling, task and intera
ction. Every
class has their own purpose and task for every mobile agent
applications. Furthermore, design patterns can increase
productivity, promote reuse and reduce complexity when
developing applications

[
emerson et. al, 2004]
.


2.3

Speech Recognition
System


Generally, speech recognition process contains three steps to
process the
speech, which

is acoustic
processing
, feature
extraction and recognition. In acou
s
tic processing, there are two
tasks. First
ly, speech

that want to be recognized have to

be

d
igitize
d
. Secondly, process analyze the start
-
end point will be
done depend on the speech signal. The second step is feature
extraction and finally after these two steps, the speech will be
classify and recognized.


However, in this paper, internet based s
peech recognition using
mobile agents is presented. Generall
y, speech recognition
system

divided into two general part, speech encoder and speech
decoder. Speech encoder contains two steps from speech
recognition system, which is acoustic processing and fe
ature
extraction. While, speech decoder consists of speech recognize.
Exactly, in internet based speech recognition, speech encoder
will be process at remote computer, which request the
applications. Speech decoder responsible to recognize the
speech and t
he process will be doing at host computer.


3
.0

The Architecture Design of Speech
Recognition

Agent


Figure
3
:
Design of Speech Recognition Agent Architecture


Generally, the design of speech recognition agent architecture is
consists of three main parts
. There is interface layer, speech
recognition system, which divided into two part, speech encoder
and speech decoder and lastly, speech agent. The architecture

of
the system is shown i
n figure 3
.

At the
interface layer

site, there

is the web browser, the
applet and

the voice recording process
.

On the host site, there is the
web server, the speech agent, the
speech
e
n
coder

and the speech decoder
.


I
nterface layer is responsible to get speech input and displaying
the recognized text. However, all tasks

will
be cover up

by web
browse
r. Web browser is responsible for

recording the speech
voice and transmitting

the speech data to the applet. T
he applet
will transmit the speech data to the speech encoder. All steps in
speech encoder such as acoustic processing an
d features
extraction will be process before the result have to transmit to
the speech agent. Speech agent is a part of mobile agent concept.
Main function of speech agent is to control transmission,
receiving speech data form speech
encoder
,
speech decode
r
, and
vice
-
versa. The speech agent also allow to process speech
recognition in asynchronous method, which it is an advantage
for

using
mobile agent concept.

Therefore, speech agents will
handle every request from remote computer safely. Furthermore,
they
also can reduce the network traffics
. Speech decoder is
responsible to recognize the speech data and transmit the
recognized text to speech agent again. Finally, speech agent will
transmit the
outcome

to the interface layer for displaying the
recognized te
xt to the user.



4
.0

Speech Recognition Agent Framework
Model


The whole of speech recognition agent framework mode
l in
detail is shown in figure 4
. It contains several main component,
interface layer, speech encoder, speech decoder, master
-
agent
and slav
e
-
agent.
The description of those components can be
found in the succeeding portion of the paper.





Figure 4
:

The Architecture Design of Speech Recognition
System


4.1

Interface Layer


To record the voice speech, java applet will take the task, which,
d
ownloaded and run by the web browser of the remote
computer. Web browsers such as Netscape Communicator or
Internet Explorer are
recommending

doing

this task. So
,

the

Java
applet will record the user’s voice, transmit speech data to the
speech en
coder wher
e as shown in figure 4
. The applet was
design to be simple, so that it can execute on most remote
computer.


4.2

Speech Agent


Speech agent is responsible to receive the speech data
that
had
been encoded by the speech encoder and send back the output
(in t
ext format) that had been recognized by the speech decoder
to the interface layer.


There are a few techniques in mobile agent concepts, such as
master
-
slave pattern.
Based on figure 4
,
speech agent using
master
-
slave technique to receive and transmit the
speech data
and recognize text.


In master
-
slave patterns, speech
-
agent will divide into two part,
master
-
agent and slave
-
agent

and both of them have their own
task. Master
-
agent is responsible to monitoring
the process
between
slave
-
agent and speech enco
der, speech decoder.
Master
-
agent will monitor the transmitted and received data
between them, such as the name of the

host, what task they
should do and

assigned every slave
-
agent with their own
id
entification
.


Slave
-
agent is responsible to performs it t
asks that given by
master
-
agent. For examples, ones the slave
-
agent get the
assigned task, they moves to speech encoder, speech decoder or
interface layer to performs the task. After that, the slave
-
agent
will returns with the result of the task to the mas
ter
-
agent. Slave
-
agent contains several items such as their identification, the host
identification and the assigned task.


4.3

Speech Encoder, Speech Decoder


T
he task of speech encoder is

to process the speech data
. They
will receive the speech data from

the interface layer. After
the
acoustic processing and features extraction has been done,
speech encoder will request to master
-
agent. Master
-
agent will
send a slave
-
agent to take the result from speech encoder.


Speech decoder will recognize the speech d
ata. Speech decoder
will received the speech data from master
-
agent through slave
-
agent. After speech decoder recognized the speech data, it will
request to the master
-
agent to take the recognized text.


5.0

Conclusion


A mobile agent speech recognition system

based on web
application was presented. Mobile agent with master
-
slave
pattern design approach will be applied to the system. With the
advantages for using mobile agent for web application such as
reduce
the
network l
oad
,

robust and fault
-
tolerant,

execut
e
asynchronously and autonomously and other else. Therefore,
with these advantages,
mobile agents are used to replace the
traditional client
-
server approach.


References



Stefan Funfrocken, Friedemann Mattern, “Mobile agent as an
architectural concept for

internet based distributed
applications”. In: Steinmetz (Ed.): Proc. KiVS'99, pp. 32
-
43,
Springer
-
Verlag, 1999


Ravi Jain, Farooq M. Anjum and Amjad Umar, “A Comparison
of mobile agent and client
-
server paradigms for information
retrieval tasks in virtual

enterprises.” AiWoRC workshop ,
Buffalo NY April 2000


David Kotz and Robert S. Gray. “Mobile Code: The Future of
the Internet.” In Proceedings of the Workshop ``Mobile
Agents in the Context of Competition and Cooperation
(MAC3)'' at Autonomous Agents '9
9, pages 6
-
12, May, 1999.


Danny B. Lange and Mitsuru Oshima, “Mobile Agent with Java:
The Aglet API,” World Wide Web Journal, 1998.


Robert S. Gray and David Kotz and Ronald A. Peterson, Jr. and
Joyce Barton and Daria Chacón and Peter Gerken and Martin
Hofmann and Jeffrey Bradshaw and Maggie Breedy and
Renia Jeffers and Niranjan Suri. “
Mobile
-
Agent versus
Client/Server Performance: Scalability in an Information
-
Retrieval Task”
. In
Proceedings of the Fifth IEEE
International Conference on Mobile Agents
, p
ages 229
-
243,
Atlanta, Georgia, December, 2001.


Z. Tu and P. Loizou (1999).
"Speech recognition over the
Internet using Java,"

IEEE International Conference on
Acoustics, Speech and Signa
l Processing,
vol. 4, pp. 2367
-
2370, Phoenix, AZ.


Aridor and D. Lange, “
Agent

Design

Patterns: Elements of
Agent

Application

Design
,”

in Second International Conference on Autonomous Agents
(Agents’98).


Emerson F. A. Lima, Patrica D. L. Machado, Flavio R.

Sampaio
and Jorge C. A. Figueiredo
, “An Approach to Modelling and
Applying Mobile Agent Design Patterns”,
ACM SIGSOFT
Software Engineering Notes
,

Volume 29 , Issue 3, Pages: 1
-

8, (May 2004)