Scribd
Upload
14 views

Exerc Session9 v2

This document contains 6 problems related to estimation and detection. The problems cover topics like determining the CRLB, MVU, BLUE, MMSE, LMMSE, MAP, MLE, and Bayesian estimators for various data and parameter models. The data models involve observing noisy signals of unknown parameters, with the noise being Gaussian or uniformly distributed. The parameter distributions include both deterministic and random cases, and both known and unknown prior distributions are considered. The problems aim to find the optimal estimators under different assumptions about the data and parameter distributions.

Uploaded by

zhenyang zhong
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
14 views

Exerc Session9 v2

This document contains 6 problems related to estimation and detection. The problems cover topics like determining the CRLB, MVU, BLUE, MMSE, LMMSE, MAP, MLE, and Bayesian estimators for various data and parameter models. The data models involve observing noisy signals of unknown parameters, with the noise being Gaussian or uniformly distributed. The parameter distributions include both deterministic and random cases, and both known and unknown prior distributions are considered. The problems aim to find the optimal estimators under different assumptions about the data and parameter distributions.

Uploaded by

zhenyang zhong
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Delft University of Technology

Faculty of Electrical Engineering, Mathematics, and Computer Science


Circuits and Systems Group

ET 4386 Estimation and Detection


Autumn 2017

Exercises -Estimation

Problem 1: We observe the data x[n] = Arn + w[n] for n = 0, ..., N − 1. The noise processs
w[n] is uncorrelated Gaussian with variance σ 2 , and r > 0 is known.

Problem 1a: Assume that A is deterministic. Find the CRLB for A.


Problem 1b: Determine the MVU estimator for A and find its variance.
Problem 1c: Determine the BLUE estimator for A.
Problem 1d: What happens with the variance of the MVU estumator if N → ∞?
Problem 1e: Now assume that A is a zero-mean random variable that is Gaussian
2 , independent of w[n]. Calculate the MMSE estimator
distributed with variance σA
of A.
Problem 1f: Determine the minimum Bayesian MSE.
Problem 1g: Find the LMMSE estimator.

Problem 2: We observe the data x[n] for n = 0, ..., N − 1. Each sample has the conditional
pdf (
1
p(x[n]|θ) = θ , 0 ≤ x[n] ≤ θ
0, otherwise.
Conditioned on θ, the observations are independent. The distribution of θ is θ ∼ U (0, β).

Problem 2a: Find the MMSE estimator of θ.


Problem 2b: Consider the situation with little prior knowledge. For what value of β
is the prior non-informative?
Problem 2c: How does the estimator change in the case of little prior knowledge.
Problem 2d: Determine the Bayesian ML estimator.
Problem 2e: Determine the MAP estimator.

Problem 3: We observe data x[n] = x[n] + w[n] for n = 0, ..., N − 1. Both s[n] and w[n]
are zero-mean WSS processes, mutually uncorrelated. Their autocorrelation functions
are given by rss [k] = σs2 [k] and rww [k] = σ 2 [k]

Problem 3a: Determine the LMMSE estimator of s = [s[0], s[1], ..., s[N − 1]]T .
Problem 3b: Determine the minimum Bayesian MSE Bmse(θ̂i ) for this linear esti-
mator.

Problem 4: Given is the data model

x[n] = A + w[n]

for n = 0, ..., N − 1. The unknown parameter A is assumed to have the distribution


(
λ exp [−λA] , A > 0
p(A) =
0, A < 0,

with λ > 0 and w[n] WGN independent of A with variance σ 2 .

Problem 4a: Determine the MAP estimator of A.


Problem 4b: Determine the LMMSE estimator of A.

Problem 5: Given is the a posteriori pdf


(
exp [−(θ − x)] , θ > x
p(θ|x) =
0, θ < x.

Problem 5a: Calculate the MMSE and MAP estimator of θ.

Problem 6: Given is the distribution


(
θ exp [−θx[n]] , x[n] > 0
p(x[n]) =
0, x[n] < 0,

for n = 0, ..., N − 1. The samples x[n] are independent over time.

Problem 6a: Give the CRLB for θ̂.

Problem 6b: Does an unbiased estimator exist that achieves the CRL bound?

Problem 6c: Determine the MLE estimator.

The a priori distribution p(θ) is given by


(
λ exp [−λθ] , θ > 0
p(θ) =
0, θ < 0.

Problem 6d: Calculate the MAP estimator.

Problem 6e: Calculate the MMSE estimator.

You might also like