西安工业信号检测与估计sde_14 ch10 11 12 贝叶斯估计.ppt

* SDE_15 一般及线性贝叶斯估计 * Comments on Smoothing: IIR Smoother Estimate s[n] based on {…, x[–1], x[0], x[1],…} Time-Invariant Non-Causal IIR Filter The Wiener-Hopf Equations become: Differs From Filter Case Sum over all k Differs From Filter Case Solve for all l H( f ) ≈1 when Pss( f) Pww(f ) H( f) ≈0 when Pss( f) Pww(f ) * SDE_15 一般及线性贝叶斯估计 * Relationship of Prediction to AR Est. Yule-Walker Wiener-Hopf Prediction Equations Symmetric Toeplitz For l=1 we get EXACTLYthe Yule-Walker Eqs used in Ex. 7.18 to solve for the ML estimates of the AR parameters!! FIR Prediction Coefficients are estimated AR parms * SDE_15 一般及线性贝叶斯估计 * Relationship of Prediction to Inverse/Whitening Filter + _ Signal Observed White Noise 1-Step Prediction White Noise Physical Reality Imagination Modeling * SDE_15 一般及线性贝叶斯估计 * Results for 1-Step Prediction: For AR(3) At each kwe predict x[k] using past 3samples Application to Data Compression Smaller Dynamic Range of Error gives More Efficient Binary Coding (e.g., DPCM –Differential Pulse Code Modulation) 西安工业信号检测与估计SDE_14 ch10 11 12 贝叶斯估计教材 谢谢 * * * * * * * * * * * * * * * * * * * * SDE_15 一般及线性贝叶斯估计 * Step #2: Plug-In Step #1 Result for aN scalar scalar where a= [a0a1. . .aN-1]T Only up to N-1 Tasince it is scalar * SDE_15 一般及线性贝叶斯估计 * Thus, expanding out N×N N×1 1×N 1×1 cross-covariance vectors? * SDE_15 一般及线性贝叶斯估计 * Step #3: Minimize w.r.t. a1, a2, … , aN-1 Only up to N-10) This is where the statistical dependence between the data and the parameter is used…via a cross-covariance vectorStep vectorStep #4: Combine Results If Means = 0 So the Optimal LMMSE Estimate is: Note: LMMSE Estimate Only Needs 1stand 2ndMoments… not PDFs!! * SDE_15 一般及线性贝叶斯估计 * Step #5: Find Minimum Bmse Substitute intoBmseresult and simplify: Note: If θ and x are statistically independent then Cθx= 0 Totally based on prior info… the data is useless * SDE_15 一般及线性贝叶斯估计 * Ex. 12.1 DC Level in WGN with Uniform Prior Recall: Uniform prior gave