Home >> List of courses >> Classification Methods in Signal Processing

Classification Methods in Signal Processing

( MKSS, 35479_3I )

Course objective: The main aim is to provide students with a basic overview of the classification problem (testing of hypotheses) that can be found in signal processing domain. Except some basic theoretical background several classification systems commonly used in signal processing applications are outlined, namely: KNN, decision trees, Neural networks, GMM, Genetic algorithms and Evolutional strategies. Furthermore, parameter estimation techniques are presented as well as dimension reducing methods. A successful student should be able to construct simple classification system, be able to test 2 and multi class hypotheses, estimate unknown parameters and have a solid theoretical overview on the topic of classification and estimation.

Concise plan: testing of hypotheses and Bayes classification; ML a MAP estimation; methods reducing dimension: LDA, HLDA, PCA; Gaussian mixture model and EM algorithm; real classification systems- overtraining phenomenon; Neural networks: MLP, RBF; Stochastic optimization: Evolutional strategies, Genetic algorithms, Simulating annealing; decision trees; KNN

Key words: testing of hypotheses, Bayes classification, ML a MAP estimation, LDA, HLDA, PCA, Neural networks: MLP, RBF, Evolutional strategies, Genetic algorithms, GMM and EM algorithm, decision trees, KNN

Nominal yearSemesterCreditsTypeExam typeHours of lectureHours of practise
2. Ing.
compulsory elective course