3.24 tb782749247
A problem of interest in the analysis of geophysical time series involves a simple model for observed data containing a signal and a reflected version of the signal with unknown amplification factor
a and unknown time delay
δ. For example, the depth of an earthquake is proportional to the time delay
δ for the P wave and its reflected form pP on a seismic record. Assume the signal is white and Gaussian with variance
σ2, and
consider the generating model
xt = st + ast−δ.
(a) Prove the process xt is stationary. If |a| 1, show that
∞
st = ,(
−a)
jxt δj
j=0
is a mean square convergent representation for the signal st, for
t = 1, ±1, ±2,.. ..
(b)
If the time delay
δ is assumed to be known, suggest an approximate computational method for estimating the parameters
a and
σ2 using
maximum likelihood and the Gauss–Newton method.
(c) If the time delay δ is an unknown integer, specify how we could estimate the parameters including δ. Generate a n = 500 point
series with
a =
.9,
σ2
= 1 and δ = 5. Estimate the integer time
delay δ by searching over δ = 3, 4,..., 7.
1 Approved Answer
Solution
A problem of interest in the analysis of geophysical time series involves a simple model for observed data containing a signal and a re?ected version of the signal with unknown ampli?cation factor a and unknown time delay δ. For example, the depth of an earthquake is proportional to the time delay δ for the P wave and its re?ected form P on a seismic record. Assume the signal is white and Gaussian with variance σ2, and consider the generating model x(t) = s(t) + a*s(t−δ).
(a) Prove the process x(t) is stationary. If |a| <1, show that ∞ s(t)=(−a)jx(t -jδ) , j=0 is a mean square convergent representation for the signal s(t), for
t = 1, ±1, ±2,.. ..
As E[s(t)]=0 --> E[x(t)]=0.
Process white gaussian is of 0 mean and sigma2 variance.
(b) If the time delay δ is assumed to be known, suggest an approximate computational method for estimating the parameters a and σ2 using maximum likelihood and the Gauss–Newton method.
Var(X)=(1+a^2)Var(s) --> Var(X)=(1+a^2)(sigma2) --> for a given Z one can find X and Y with Z=(1+X)Y, where X=a^2 and Y=variance of the signal.
(c) If the time delay δ is an unknown integer, specify how we could estimate the parameters including δ. Generate a n = 500 point series with a = .9, σ2= 1 and δ = 5. Estimate the integer time delay δ by searching over δ = 3, 4,..., 7.
Var(X)=(1+0.9^2)*1=1.81 -->s=sqrt(1.81)=1.34
X=rand(500,1,0,1.34)
delta=4