TSKS15 Detection and Estimation of Signals
TSKS15 Detection and Estimation of Signals treats statistical signal processing, specifically parameter estimation and detection of signals. The purpose of the course is to provide a solid foundation in algorithms, models, methods and theory for the extraction of information from noisy signals. Applications are found within radar systems, communications systems, positioning systems and image analysis.
- Problems in radar, communications and source localization systems.
- Classical versus Bayesian approaches.
- Hypothesis testing. Binary and M-ary tests. Bayes cost and error probability. Neyman-Pearson theorem.
- Classical estimation. Maximum-likelihood, Fisher information, Cramer-Rao bound.
- Bayesian estimation theory. MMSE and LMMSE.
- Composite hypothesis testning. GLRT. Marginalization. Model selection.
- Linear and nonlinear models with Gaussian noise. Slepian-Bang formula. Noise whitening.
- Detection of signals in continuous time.
- Performance and variance analysis. Asymptotic properties of estimates.
- Complex-valued data and noise. Circularly symmetric noise.
- Applications: amplitude, frequency, phase, time-delay and angle estimation.
The course consists of a lecture series and two computer lab projects:
- Time-of-arrival (range) estimation. The purpose of this project is to implement a maximum-likelihood time-of-arrival (range) estimation algorithm, perform Monte-Carlo simulation of its performance, and compare with theoretical bounds.
- Music recognition. The purpose of this project is to design a classifier algorithm for the recognition of music, implement this algorithm, and evaluate its performance and robustness to model assumptions using Monte-Carlo simulation.
Erik G. Larsson
Last updated: 2020 08 26 15:38