Decision Theoretic Fusion Framework for Actionability Using Data Mining on an Embedded System - Data Mining: Theory, Methodology, Techniques, and Applications - page 102

(eASR), distributed speech recognition (DSR) and embedded TTS (eTTS) are being
introduced for commercial use. In particular, since the demand of Telematics services
is surging, speech interface to interact with human users has become an essential
means of the multi-modal interface. As a Telematics client service interface, the
eASR, DSR, and eTTS combined as a stand-alone unit provides an easy manipulation
interface for command and controlling a car navigation system while the driver can
pay attention to safe driving. In addition, as computer technology is improved, small
sized computers such as AutoPCs has been utilized in various fields. Thus, by using
this embedded system, the user requests and wants to utilize various service applica-
tions that they is used on a desktop PC, even while driving a car.
However, an embedded hardware system has a resource limitation and low per-
formance capability. Actually, this condition is not able to represent huge data. Thus,
a new architectural model is required in an embedded system. One alternative
method is to use the Web. On the server-side, a comprehensive database is first
mined, and then all the discovered patterns are stored in a DBMS. On the client-side,
some abstraction data and indexes are stored. If the user wants to show specific data,
the client obtains that information from a data mining server via the Internet using
abstraction data and indexes. Meanwhile, Multimedia files such as music, moving
picture files can be presented using data streaming method [5]. In our system, we
reduce the memory size by using this concept. Even if a data communication fee per
a packet should be paid, compression techniques for transmission packets could re-
duce the packet size. In addition, this can be resolved according to the policy of ser-
vice usage. On the other hand, sensor network [12][13] is employed. Distributed
sensor network obtain the distinct information by using its own functionality. Finally
fused data provides the reliable information processed from competitive and coopera-
tive terms to the users.
Fig. 5. Embedded system using an AutoPC for car navigation
With the above concepts, we designed a framework for command, control, com-
munication, and intelligence environment based on a software agent on an embedded
car navigation system, and then implemented it on AutoPC environment as shown in
Figure 5. The proposed framework provides the structure to extend the system easily
and integrate with other services. It is possible that the core processing such as