Universidade de Brasília
Transcrição
Universidade de Brasília
Adaptive & Array Signal Processing AASP Prof. Dr.-Ing. João Paulo C. Lustosa da Costa University of Brasília (UnB) Department of Electrical Engineering (ENE) Laboratory of Array Signal Processing PO Box 4386 Zip Code 70.919-970, Brasília - DF de Brasília Homepage:Universidade http://www.pgea.unb.br/~lasp Laboratório de Processamento de Sinais em Arranjos 1 Mathematical Background: Stationary Processes (1) Stochastic (or random) process Evolution of a statistical phenomenon according to probabilistic laws Before starting the process, it is not possible to define the exactly way it evolves. Infinite number of realizations of the process Strictly stationary Statistical properties are invariant to the time shift If the Probability Density Function (PDF) f(x) is known All the moments can be computed. In practice, the PDF is not known. Therefore, in most cases, only the first and the second moments can be estimated with samples. Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 2 Mathematical Background: Stationary Processes (2) Statistical functions The first moment also known as mean-value function E{ } stands for the expected-value operator (or statistical value operator) and u(n) is the sample at the n-th instant. The autocorrelation function The autocovariance function Note that all the three functions are assumed constant with time. Therefore, they do not depend on n. Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 3 Mathematical Background: Stationary Processes (3) Relation between the statistical functions The relation between mean-value, autocorrelation and autocovariance functions is given by Proof of the relation: If mean is zero the autocovariance and the correlation functions are equal. Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 4 Mathematical Background: Stationary Processes (4) Wide-sense Stationary and mean estimate Note that if If these equations are satisfied, then the process is wide-sense stationary or stationary to the second-order. second-order In practice, only a limited number of samples are available. Therefore, the mean, the autocovariance and the autocorrelation are estimated. Estimate of the mean Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 5 Mathematical Background: Stationary Processes (5) Mean estimate Applying the expected-value operator: Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 6 Mathematical Background: Stationary Processes (6) Mean ergodicity A process is mean ergodic in the mean square-sense error sense if Another way to represent The order doesn’t matter due to the modulus. Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 7 Mathematical Background: Stationary Processes (7) Mean ergodicity Replacing l = n – k, and after some algebra: Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 8 Mathematical Background: Stationary Processes (8) Mean ergodicity If the process is asymptotically uncorrelated, i.e. c(l) → 0 when l increases, then the process is mean ergodic. Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 9 Mathematical Background: Stationary Processes (9) Correlation ergodicity Similarly a process can also be correlation ergodic Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 10 Mathematical Background: Stationary Processes (10) Correlation matrix L by 1 observation vector Correlation matrix The main diagonal contains always real-valued elements. Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 11 Mathematical Background: Stationary Processes (11) Correlation matrix The correlation matrix is Hermitian, i.e. Proof: Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 12 Mathematical Background: Stationary Processes (12) Correlation matrix The correlation matrix is Hermitian, i.e. As a consequence of the Hermitian property Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 13 Mathematical Background: Stationary Processes (13) Correlation matrix The correlation matrix is Hermitian, i.e. The correlation matrix is Toeplitz, i.e. the elements of the main diagonal are equal as well as the elements of each diagonal parallel to the main diagonal. Important: • Wide sense stationary R is Toeplitz • R is Toeplitz Wide sense stationary Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 14 Mathematical Background: Stationary Processes (14) Correlation matrix The correlation matrix is always nonnegative definite and almost always positive definite. We define the scalar , where x is constant, then We know that Nonnegative definite or positive semidefinite Also if If positive definite Also if Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 15 Mathematical Background: Stationary Processes (15) Example of correlation matrix We consider the following data model where u(n) is the received signal, is the signal of interest and v(n) is the zero mean i.i.d. noise component with variance . Note: independent and identically distributed (i.i.d.) Computing the autocorrelation function Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 16 Mathematical Background: Stationary Processes (16) Example of correlation matrix Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 17 Mathematical Background: Stationary Processes (17) Example of correlation matrix Parameter estimation: Given the noise variance and computing r(0), estimate |α|2. Given the estimate of |α|2 and computing r(l), estimate ω. Assuming the case where - All lines and all columns are linearly dependent. - R is rank 1. - Only one eigenvalue is not zero. - The model order is one. Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 18 Mathematical Background: Stochastic Models (1) Model: Hypothesis to explain or to describe the hidden laws that governs or constraints the generation of some physical data. We consider three models Autoregressive (AR): no past values of the input model; Moving average (MA): no past values of the output model; Mixed autoregressive-moving average (ARMA): include both cases. Data model If the data is completely random, then no prediction is possible. However, if there is some dependence of the previous data (AR or MA), then the prediction becomes possible. Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 19 Mathematical Background: Stochastic Models (2) Autoregressive model Applying the z transform Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 20 Mathematical Background: Stochastic Models (3) Autoregressive model Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 21 Mathematical Background: Stochastic Models (4) Autoregressive model Multiplying by and applying In the matrix representation (called Yule-Walker equation) Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 22 Mathematical Background: Stochastic Models (5) Moving average ⇒ Moving average of M input samples ⇒ Pure FIR (Finite Response Filter) – All zero ARMA is a mix of a FIR and IIR generator Systematics Analyser Generator white colored white colored Processo AR IIR (all pole) FIR (all zero) Processo MA FIR (all zero) IIR (all pole) Processo ARMA Mix Mix Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos Mathematical Background: Stochastic Models (6) Example Power Spectral Density 10 3 Φ XX(f) Φ YY(f) Φ (f) Φ ZZ(f) 10 10 2 1 0 0.2 0.4 0.6 0.8 1 f/f p Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 24 Mathematical Background: Stochastic Models (7) Example Pole zero diagram 0.8 0.6 0.4 Imag 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -0.5 0 Real Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 0.5 1 25 Mathematical Background: Stochastic Models (8) Example Frequency response 10 |G(f)| 10 0.2 0.1 10 10 10 H(f) A(f) H(f)*A(f) 0 -0.1 -0.2 0 0.1 0.2 0.3 0.4 0.5 f/f p Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 26 Mathematical Background: Stochastic Models (9) Example Impulse response 1 gh(τ) ga(τ) gh,a(τ g(τ) 0.5 0 -0.5 0 2 4 6 8 10 τ Universidade de Brasília Laboratório de Processamento de Sinais em Arranjos 27