Advanced Digital Signal Processing(capter one).pptVIP

Solution: (1) Product factorization (spectral factorization theorem) Ricatti Equation (2) Sum factorization (1) Inversion formula method (2) Partial fraction method (3) Compute the system function of the causal IIR filter Another method for computing causal IIR Wiener filter 2.5 Scale Kalman Filter The first prediction: The second prediction: Innovation: The optimal gain (Kalman gain) Prediction error power One Step Prediction error The minimum mean square error Prediction error power Computational steps Initiation 2.6 Vector Kalmin Filter 2.6.1 Signal Vector and Data Vector Case one: Observation Independence Signals Simultaneity Case Two: Estimating A signal of High Order AR Model 2.6.2 Derivation of Vector Kalman Filter Scale arithmetic matrix arithmetic * * 1.3.2 Orthogonal Project Theorem and Gram Schmidt Orthogonalization 1. Orthogonal Project Theorem If is an orthogonal project of a vector x in linear subspace Y, then the distance between and x in Y is minimum. proof Supposing y is a vector within Y, obvious: That is, while , the distance between and x in Y is minimum. 2. Gram-Schmidt Orthogonalization According as orthogonal decomposition function : Gram-Schmidt Orthogonalization is a recursive process：get a set of orthogonal base by non- orthogonal base linear space That is We can regard Gram-Schmidt orthogonalization as the process of extending subspace by appending an orthogonal base to the former subspace. That is , Shown orthogonal project of vector yn in the subspace Yn-1, we have Let Was shown as: 1.3.3 Innovation A vector space Y can be built on both base y and ε, they includes the same information. The difference y from ε is , Show elements noncorrelation in ε. They include different information respectively. It increase new information when a εi is added. So we define random variable εi as innovation. is known as the innovations representation