Information Technology Reference
In-Depth Information
To find the intersection point (supposing A is nearer to the origin), the following
condition is tested:
X
LBni
(
ʱ
)
X
RAni
(
ʱ
) .
(6)
=>
X
LBni
+
p
LBni
(
ʱ
-
Y
ni
)
X
RAni
-
p
RAni
(
ʱ
-
Y
ni
)
[
Y
ni
,
Y
n
(
i
+
1)
] .
(7)
=> (
ʱ
-
Y
ni
)(
p
LBni
+
p
RAni
)
X
RAni
-
X
LBni
[
Y
ni
,
Y
n
(
i +
1)
] .
(8)
=> (ʱ - Y
ni
)
p
∆
X
[
Y
ni
,
Y
n
(
i
+ 1)
] .
(9)
Equation 9 is tested using the binary-search algorithm, given below.
α
=
;
-
∆Y
=
;
Yp
=
∆Y
*
p
;
while (
∆Y
!= 1 )
{
∆Y
=
∆
;
If (
Yp
≤
∆X
)
{
α
=
α
+
∆Y
;
Yp
=
Yp
+
∆Y
*
p
;
}
else
{
α
=
α
-
∆Y
;
Yp
=
Yp
-
∆Y
*
p
;
}
}
Fig. 5.
Antecedent Unit
Fig. 5 shows the architectural design of the antecedent unit. Each antecedent unit
has 2 memories M
3
and M
4
. The abscissas of the end-points of the linear segments for
the membership functions are stored in these memories. M
3
contains the antecedent
membership functions while M
4
contains the input membership functions. Each
membership function requires 8 memory locations. Thus M
3
has 32 * 8 = 256 words
while M
4
has 2 * 8 = 16 words. Both the memories are 16 bits wide.
