EC5.
406 – Signal Detection and Estimation Theory – Assignment 1
Due Date: 21st August, 2025
Instructions:
• Submission on Moodle: Upload PDF file containing the handwritten assignment solutions. Compile
all your pages in the right order into a single PDF file.
• Late submission: A 10% penalty per day will be applicable (accepted up to at most 3 days after the
deadline).
• Discussions are allowed, but copying and plagiarism will not be tolerated and will attract strict penalties.
• Clearly show the steps used to arrive at the solutions.
1. A sinusoidal signal is observed in presence of white Gaussian noise:
x[n] = sin (2πf n) + w[n] , n = 0, 1, . . . , N − 1,
where w[n] are i.i.d. Gaussian random variables with mean 0 and variance σ 2 (known). The pdf of the
observation x[n] depends on the frequency parameter f and is denoted by p(x[n]; f ).
(a) For f = 41 , plot the pdf of x[0], x[1], x[2], and x[3].
(b) For x[0] = 1, plot the likelihood function L(f ) = p(x[0]; f ) for f ∈ [−1, 1].
1
(c) If N = 3, compute the CRLB for estimating f when f = 4 and f = 21 . Give simplified answers.
(d) Using MATLAB or otherwise, plot the CRLB as function of f ∈ [−1, 1] when N = 3, 6, 9.
2. We wish to estimate the parameter A which influences the observations as
x[n] = A(−1)n + w[n] , n = 0, 1, . . . , N − 1,
where w[n] are i.i.d. Gaussian random variables with mean 0 and variance σ 2 .
(a) Verify that regularity conditions are satisfied when N observations are available.
(b) Find the Fisher information for the parameter A.
(c) Find the Cramer-Rao lower bound for A.
(d) Find MVU estimator for A when N = 2. Is this an efficient estimator? Justify.
3. Two sets of measurements x[n] and y[n] are available for estimating a constant signal A,
x[n] = A + w[n] , n = 0, 1, 2,
y[n] = A + v[n] , n = 0, 1, 2.
The noises are independent zero-mean Gaussians but with varying variances. Specifically, w[0], w[1] and,
w[2] have variance σ 2 /2, σ 2 and 2σ 2 respectively whereas v[0], v[1] and, v[2] all have a variance of σ 2 .
Only one of the two sets of data could be used to estimate A.
Statisticians are arguing over which data set would lead to a better estimator. To resolve the issue, for
each set, identify the MVU estimator and the minimum variance it can achieve. Are these estimators
efficient? Justify.
4. Problem 2.2 from Chapter 2 of Volume 1 (Estimation Theory).
5. Problem 2.9 from Chapter 2 of Volume 1 (Estimation Theory).
6. Problem 3.9 from Chapter 3 of Volume 1 (Estimation Theory).
