Digital Signal Processing Reference
In-Depth Information
2.8.3 The role of linear dependence as spuriousness in de-
termining K within the state space-based perspective
of signal processing
The investigated FID from its geometric progression c
n
=
K
k=1
d
k
z
k
can also
be viewed from a state spacebased perspective in signal processing. This can
be done by introducing a K−dimensional linear vector spaceKwith the sym
metric inner product (2.3) and a basis{Υ
k
}(1≤k≤K) which is a complete
set of the solutions of the stationary Schrodinger equation (2.6). The cor
responding nonstationary solution|Φ(t)) of the timedependent Schrodinger
equation (2.1) is given by (2.4). Its discretized version is|Φ
n
) = U
n
|Φ
0
) where
U is the time evolution operator U≡U(τ) = exp (−i
ˆ
τ). Here, the continu
ous time variable t is discretized via t = nτ (0≤n≤N−1), where τ is the
sampling rate. The two equivalent representations Φ and Υ can be related to
each other. This is done by using the spectral decomposition (2.16) of the time
evolution operator in Φ
n
to arrive at the expansion Φ
n
=
K
k=1
d
1/2
u
k
|Υ
k
),
where u
k
= exp (−iω
k
τ) and d
k
is the amplitude/residue from (2.19), i.e.,
d
k
= (Φ
0
|Υ
k
)
2
. To cohere with the notation from the other subsections in
this section, we shall switch from the u−to the z−variable, in which case Φ
n
will be replaced by Φ
′
n
where|Φ
′
n
) = exp (in
ˆ
τ)|Φ
0
) and
k
K
d
1/2
k
Φ
′
n
=
z
k
d
k
= (Φ
′
0
|Υ
k
)
2
.
|Υ
k
)
(2.215)
k=1
This K−dimensional state vector Φ
′
n
is defined through its K components or
“coordinates”
Φ
′
0
={d
1/2
,d
1/2
2
,d
1/2
3
,,d
1/2
K
}
1
Φ
′
1
={d
1/2
z
1
,d
1/2
z
2
,d
1/2
z
3
,,d
1/
K
z
K
}
1
2
3
Φ
′
2
={d
1/
1
z
1
,d
1/
2
z
2
,d
1/
3
z
3
,d
1/2
K
z
2
K
}
.
Φ
′
n
={d
1/2
z
1
,d
1/2
z
2
,d
1/2
z
3
,,d
1/
K
z
K
}.
(2.216)
1
2
3
}
k=1
In other words, since{Υ
k
is a basis in the spaceK, any function from
this space including Φ
′
n
∈Kcan be expanded in terms of the elements Υ
k
such that the k th component/coordinate of Φ
′
n
is d
1/2
k
u
k
as in (2.215), or
equivalently
Φ
′
n
= d
1/
1
z
1
Υ
1
+ d
1/
2
z
2
Υ
2
+ d
1/
3
z
3
Υ
3
++ d
1/2
K
z
K
Υ
K
. (2.217)
Taking the scalar product between|Φ
′
n
) and (Φ
′
m
|Φ
′
n
), setting the
normalization of Υ
k
to unity through||Υ
k
||≡1 and using the orthogonaliza
tion (Υ
k
′
|Υ
k
) = δ
k,k
′
, it follows from (2.216) and (2.217)
|via (Φ
′
m
K
S
n,m
≡(Φ
′
m
|Φ
′
n
) =
d
k
z
n+m
.
(2.218)
k
k=1
Search WWH ::
Custom Search