Digital Signal Processing Reference
In-Depth Information
4.2.2 Practical Calculation
4.2.2.1 Error Function
The error function is denoted as erf(
x
) and is defined in the literature in different
ways. We adopt the definition from [SCH75, p. 89]:
ð
x
2
p
e
u
2
d
u:
erf
ðxÞ¼
p
(4.23)
0
This function has some important properties which can be easily verified from
the integral (
4.23
):
erf
ð
0
Þ¼
0
;
(4.24a)
erf
ð1Þ ¼
1
;
(4.24b)
erf
ð1Þ ¼
1
;
(4.24c)
erf
ðxÞ¼
erf
ðxÞ:
(4.24d)
In the following, we show how the normal distribution can be expressed in terms
of the erf function. To this end, in the integral (
4.22
) we introduce the variable
x
m
p
s
u ¼
(4.25)
resulting in:
2
3
ð
ð
ð
u
0
u
1
p
1
2
2
p
2
p
e
u
2
d
u ¼
4
e
u
2
d
u þ
e
u
2
d
u
5
F
X
ðxÞ¼
p
p
p
1
1
0
2
3
ð
1
ð
u
1
2
2
p
2
p
4
e
u
2
d
u þ
e
u
2
d
u
5
:
¼
p
p
(4.26)
0
0
Applying the definition (
4.23
) and property (
4.24c
), the first term in (
4.26
)
becomes:
ð
1
2
p
e
u
2
d
u ¼
erf
ð1Þ ¼ ð
1
Þ¼
1
p
:
(4.27)
0
Search WWH ::
Custom Search