Special Topics: Data Science

Size: px
Start display at page:

Download "Special Topics: Data Science"

Transcription

1 Special Topics: Data Science L6b: Wiener Filter Victor Solo School of Electrical Engineering University of New South Wales Sydney, AUSTRALIA

2 Topics Norbert Wiener MIT Professor (Frequency Domain) Wiener Filters 1 Signal Extraction: Wiener optimum Filter MSE Example Deconvolution 3 Spectral Identification- Estimation Signal Extraction occurs in many disciplines e.g. estimation of: long term temperature trends; employment; signals from seismic arrays. Deconvolution - where the signal of interest has been filtered before being recorded - also occurs widely eg medical imaging; oceanography; geological prospecting; remote sensing. It is closely associated with ill-conditioned inverse problems. It is possible to obtain the (frequency domain) Wiener filter by taking limits in the finite data Wiener filter. But it is simpler to derive the frequency domain Wiener filter from scratch. Prof. V. Solo (UNSW) / 10

3 Signal Extraction I Model y t = s t + n t, s t n u for all t, u s t is stationary with spectrum F s (ω) n t is stationary with spectrum F n (ω). So y t is stationary with spectrum (ω) = F s (ω) + F n (ω). Proof I ŝ t = Σ W t u y u = (W y) t s t ŝ t = s t (W (s + n)) t = ((δ W ) s) t (W n) t E(s t ŝ t ) = E((δ W ) s) t + E(W n) t Result III: The MMSE linear infinite data estimator is: ŝ t = W (z 1 )y t W (e jω ) = Fs (ω) (ω) = 1 1+F n(ω)/f s (ω) Note that W (z 1 ) is a two-sided filter. W (z 1 ) = Σ w r z r All signals have zero mean. W = W (ω) = W (e jω ) F (ω) = γ(e jω ) = Σ γr e jωtr By Parseval s Theorem E((δ W ) s) t = π π F ((δ W ) s(ω) dω π By the spectral-filtering results F ((δ W ) s (ω) = 1 W F s (ω) Similarly E(W n) t = π π W F n (ω) dω π mse = π π ( 1 W F s + W F n ) dω π This is quadratic in W and can be optimized by completing the square. Prof. V. Solo (UNSW) 3 / 10

4 Signal Extraction II Proof II Set W o = which we add and subtract inside the integrand. mse(ω) = 1 W o ( W W o ) F s + W o + ( W W o ) F n = 1 W o F s + W W o F s M s Re( W W o )(1 W o )F s + W o F n + W W o F n M n +Re( W W o ) W o F n = M s + cp + M n = mse o + cp cp = Re( W W o )[ W o F n (1 W o )F s ] = Re( W W o )[ W o (F n + F s ) F s ] = 0 Fs (ω) (ω) = Fs So mse(ω) = mse o (ω) + W W o mse o (ω) with equality iff W = W o. Optimum MSE mse o (ω) = 1 W o F s + W o F n = 1 Fs F s +F n F s + Fs F n = F n F F y s + F s F F y n = F n F s (F n + F s )/Fy = F n F s / = W o F n = F n F s /(F n + F s ) = Fs 1+λ ω λ ω = Fs F n is a frequency domain vsnr. So mse o = π F s dω π 1+λ ω π Optimum relative mse is then rmse = mse o var(s = Fs dω 1+λω t) dωπ Fs π < 1 Prof. V. Solo (UNSW) 4 / 10

5 Signal Extraction Example: AR(1)+WN AR(1) in white noise σ F s = s = σ 1 φe jω s A & F n = σn = σs / A + σn Fs = σ s +σ n A A = σ s A A N = σ s N = N A Now look for a spectral factorization N = B = V 1 αe jω. V (1 + α αcos(ω)) = σ s + σ n(1 + φ φcos(ω)) V (1 + α ) = σ s + σ n(1 + φ ) αv = σ nφ Divide nd equation by 1st to get α 1+α = σ n φ σ s +σ n (1+φ ) = φ λ+1+φ = 1 M where λ = vsnr = σ s σ n So solve: α Mα + 1 = 0 α = 1 (M ± M 4) So need M (λ φ )/φ λ φ φ λ + (1 φ) 0 which holds. Stable solution α = 1 (M M 4) (why?) Next sgn(φ) = sgn(m) = sgn(α) V = σn φ α = σ φ n α Prof. V. Solo (UNSW) 5 / 10

6 AR(1)+WN II W o W o = Fs = = σ s N σ s V 1 αe jω = λ α φ = σ s α σ n φ 1 1 αe jω 1 σ = 1 αe jω 1 αe jω Filter Properties quadratic equation α 1+α = φ λ+1+φ < φ α φ 1+α = λ+1+φ 1+φ Check that, for x > 0 x 1+x is an increasing function of x But LHS<RHS α < φ V > σn quadratic as λ 0 then α φ quadratic as λ then α 0. Optimum mse 1st var(s t ) = σ s 1 φ then mse o = dω W o F n π = σ σ 1 αe jω n dω π = σ σ n 1 α So relative mse is σ σ n 1 α 1 φ σs 1 φ λ 1 α = α 1 α / φ 1 φ = λ α 1 φ Check that for 0 < x < 1, x 1 x is an increasing function of x. It follows that, as expected, rmse < 1. as λ 0 then rmse 1 as λ then rmse 0 Prof. V. Solo (UNSW) 6 / 10

7 AR(1)+WN III W o = Forwards-Backwards Filter W o = Two-sided Filter The last term is an AR(1) spectrum so = γ o Σ α r e jωr σ 1 αe jω where γ o = σ 1 α = λ α 1 φ 1 α. So the weights are w r = γ o α r So the Wiener signal estimator in the time domain is: ŝ t = γ o Σ α r y t r = γ o (w y) t In practice the sum is truncated once the weights are very small. σ We rewrite W = 1 αe jω in z-transform notation as σ 1 αz σ 1 αz 1. W (z 1 ) = Then ŝ t = W (z 1 )y t = σ σ 1 αz 1 αz y 1 t. This suggests ŝ t can be computed in two stages: Forward or Causal filtering ξ t = σ 1 αz 1 y t Followed by Backwards or one-sided anti-causal filtering ŝ t = σ 1 αz ξ t Prof. V. Solo (UNSW) 7 / 10

8 General Approach to Computation of Wiener Filter Numerous approaches possible. We use direct MA spectral factorisation via Wilson s algorithm. It is compact, fast and reliable. Suppose both F s, F n are described by ARMA models Bs F s = A s and F n = Bn A n Then W = Fs Bs Bs = A s /( A s + Bn A n ) = B s A n B s A n + B na s Use Wilson s algorithm to get: B s A n + B n A s = θ where θ(z 1 ) is one-sided. Then W o = W o (z 1 ) = Bs An θ Wilson s Newton Algorithm Given MA covariances cr, 0 r m form the vector c = [c0, c 1,, c m] Introduce the MA parameter vector θ = [θ 0, θ,, θ m ] and corresponding covariances c r (θ) = Σ l=m r l=0 θ l θ l+r and vector c(θ) Wilson s iteration is: θ (k+1) = T 1 (θ k )(c(θ (k) ) + c ) where T (θ) = T L (θ) + T R (θ) T L (θ) = Bs (z)an(z) B s (z 1 )A n(z 1 ) θ(z) θ(z 1 ).. = backwards filter forwards θ 0 filter Prof. V. Solo (UNSW) 8 / 10 T R (θ) = θ 0 θ 1 θ m θ 1 θ θ m 0.. θ m 0 0 θ 0 θ 1 θ m 0 θ 0 θ 1 θ m 1 θ m

9 Topics Deconvolution 1 Wiener Filter Optimum MSE 3 Computation Prof. V. Solo (UNSW) /5

10 Wiener Deconvolution Filter Deconvolution generalizes the signal extraction problem. The model for the recorded data is y t = s t + n t But the signal of interest has been filtered to generate s t s t = k x t = k(z 1 )x t where k(z 1 ) could be two-sided. The problem is to estimate: x t from infinite data y, We consider a linear estimator ˆx t = (W y) t and show W o = k Fx = k Fx k F x +F n The mse is E(x t ˆx t ) = E(et ) = dω F e π Then F e = F x + Fˆx ReF x ˆx = F x + W Re(F xy W ) Set W o = Fxy and complete the square to show this is the optimal filter. Continuing W o = F x k k F x +F n Fx k = k Fx k F x +F n F s F s +F n W o = 1 k = 1 k Thus the WF functions by first estimating s and then inverse filtering that to get ˆx t. But k may have zeros so this implementation is not reliable. The 1st formula does not have that problem. Prof. V. Solo (UNSW) 3 / 5

11 Optimum MSE MSE We have mse(ω) = F e and F e = F x + W o Re(F xy W o ) = F x + W o Re(F x k W o ) = F x + k F x F y = F x k F x = F x (1 k F x F = F n x = F x F n = k F x +F n F x 1+ k λ ω F x F x k ) natural extension of signal extraction where λ ω = Fx F n which is also a natural extension of the signal extraction formula. Relative MSE The relative mse is then rmse = Fx Fn Fy dω π Fx dω π < 1 as λ ω 0 then rmse 1 as λ ω then rmse 0 Prof. V. Solo (UNSW) 4 / 5

12 Computation via Wilson s Algorithm Suppose F x, F n are ARMA spectra and k is a proper rational transfer function F x = Nx D x, F n = Nn D n, k = N k D k W o = k = N k D k = N k D k Fx k F x +F n N x / D x Nx N k Dx D k + Nn Dn N x D n D k N x N k D n + D k N n D k Here we do the one-sided case. If k is two-sided we write it as a sum of a causal filter + anti-causal filter. Then the aproach here is easily modified. Now apply Wilson s spectral factorization algorithm to factor the denominator as θ where θ(z 1 ) is one-sided. Then W o = N k N Dk θ where N = N x D n D k is one-sided. So W o (z 1 ) = N k (z) D k (z) = backwards filter backwards filter N(z) N(z 1 ) θ(z) θ(z 1 ) forwwards filter The two-sided weights can also be computed similarly to the signal extraction case but will now not be symmetric in lag. Prof. V. Solo (UNSW) 5 / 5

13 Appendix A: Backwards or Anti-causal Filtering Let H(z) = Σ 0 h r z r be an anti-causal filter. We show s t = H(z)u t can be implemented by a causal filtering of the reversed input sequence. Given a fixed time interval [0, T ] introduce the reversed sequences: ū t = u T t and s t = s T t We illustrate with a single pole filter H(z) = 1 1 φz. Then s t = 1 1 φz u t (1 φz)s t = u t s t = φs t+1 + u t. Consider the pairs of equations. s T 1 = φs T + u T 1 s 1 = φ s 0 + ū 1 s T = φs T 1 + u T s = φ s 1 + ū.. s T t = φs T t+1 + u T t s t = φ s t 1 + ū t The LHS is the anti-causal filtering; the RHS is the causal equivalent on the reversed input sequence. It generates the reverse output sequence. Reversing that gives the output sequence. Prof. V. Solo (UNSW) 9 / 10

14 Appendix B: Wilson s Newton Spectral Factorisation Newton s Algorithm Solve φ(x (m+1) 1 ) = f (m+1) 1 (x) a. Let x o be a start value. Do a Taylor series around x o φ(x) φ(x o ) + dφ (x x dxo T o ) Us this to find x 1 so that φ(x 1 ) 0. 0 = φ(x o ) + dφ (x dxo T 1 x o ) x 1 x o = [ dφ ] 1 (f (x dxo T o ) a) Iterate this update. There is no general proof of convergence. Results can be found under various conditions. In this case Wilson supplied a proof under the minimal condition that the MA covariance generating function is positive definite. Wilson s Algorithm Here we have f r = c r cr = Σ l=m r l=0 θ l θ l+r cr, r = 0, 1,, m fr θ s =< θ s+r > + < θ s r > where { θi if i = 0,, m < θ i >= 0 if otherwise Next check that T L (θ)θ = c(θ) = T R (θ)θ [ fr θ s ] = T L (θ) + T R (θ) = T (θ) Applying the Newton update gives the iteration θ (k+1) = T 1 (θ (k) )(c(θ (k) ) + c ) Prof. V. Solo (UNSW) 10 / 10

V. Adaptive filtering Widrow-Hopf Learning Rule LMS and Adaline

V. Adaptive filtering Widrow-Hopf Learning Rule LMS and Adaline V. Adaptive filtering Widrow-Hopf Learning Rule LMS and Adaline Goals Introduce Wiener-Hopf (WH) equations Introduce application of the steepest descent method to the WH problem Approximation to the Least

More information

EE482: Digital Signal Processing Applications

EE482: Digital Signal Processing Applications Professor Brendan Morris, SEB 3216, brendan.morris@unlv.edu EE482: Digital Signal Processing Applications Spring 2014 TTh 14:30-15:45 CBC C222 Lecture 11 Adaptive Filtering 14/03/04 http://www.ee.unlv.edu/~b1morris/ee482/

More information

Nonparametric and Parametric Defined This text distinguishes between systems and the sequences (processes) that result when a WN input is applied

Nonparametric and Parametric Defined This text distinguishes between systems and the sequences (processes) that result when a WN input is applied Linear Signal Models Overview Introduction Linear nonparametric vs. parametric models Equivalent representations Spectral flatness measure PZ vs. ARMA models Wold decomposition Introduction Many researchers

More information

Gradient Based Optimization Methods

Gradient Based Optimization Methods Gradient Based Optimization Methods Antony Jameson, Department of Aeronautics and Astronautics Stanford University, Stanford, CA 94305-4035 1 Introduction Consider the minimization of a function J(x) where

More information

Forecasting with ARMA

Forecasting with ARMA Forecasting with ARMA Eduardo Rossi University of Pavia October 2013 Rossi Forecasting Financial Econometrics - 2013 1 / 32 Mean Squared Error Linear Projection Forecast of Y t+1 based on a set of variables

More information

Adaptive Filter Theory

Adaptive Filter Theory 0 Adaptive Filter heory Sung Ho Cho Hanyang University Seoul, Korea (Office) +8--0-0390 (Mobile) +8-10-541-5178 dragon@hanyang.ac.kr able of Contents 1 Wiener Filters Gradient Search by Steepest Descent

More information

Part III Example Sheet 1 - Solutions YC/Lent 2015 Comments and corrections should be ed to

Part III Example Sheet 1 - Solutions YC/Lent 2015 Comments and corrections should be  ed to TIME SERIES Part III Example Sheet 1 - Solutions YC/Lent 2015 Comments and corrections should be emailed to Y.Chen@statslab.cam.ac.uk. 1. Let {X t } be a weakly stationary process with mean zero and let

More information

STAT 443 Final Exam Review. 1 Basic Definitions. 2 Statistical Tests. L A TEXer: W. Kong

STAT 443 Final Exam Review. 1 Basic Definitions. 2 Statistical Tests. L A TEXer: W. Kong STAT 443 Final Exam Review L A TEXer: W Kong 1 Basic Definitions Definition 11 The time series {X t } with E[X 2 t ] < is said to be weakly stationary if: 1 µ X (t) = E[X t ] is independent of t 2 γ X

More information

III.C - Linear Transformations: Optimal Filtering

III.C - Linear Transformations: Optimal Filtering 1 III.C - Linear Transformations: Optimal Filtering FIR Wiener Filter [p. 3] Mean square signal estimation principles [p. 4] Orthogonality principle [p. 7] FIR Wiener filtering concepts [p. 8] Filter coefficients

More information

Wiener Filtering. EE264: Lecture 12

Wiener Filtering. EE264: Lecture 12 EE264: Lecture 2 Wiener Filtering In this lecture we will take a different view of filtering. Previously, we have depended on frequency-domain specifications to make some sort of LP/ BP/ HP/ BS filter,

More information

EE482: Digital Signal Processing Applications

EE482: Digital Signal Processing Applications Professor Brendan Morris, SEB 3216, brendan.morris@unlv.edu EE482: Digital Signal Processing Applications Spring 2014 TTh 14:30-15:45 CBC C222 Lecture 11 Adaptive Filtering 14/03/04 http://www.ee.unlv.edu/~b1morris/ee482/

More information

Time Series Analysis. Solutions to problems in Chapter 5 IMM

Time Series Analysis. Solutions to problems in Chapter 5 IMM Time Series Analysis Solutions to problems in Chapter 5 IMM Solution 5.1 Question 1. [ ] V [X t ] = V [ǫ t + c(ǫ t 1 + ǫ t + )] = 1 + c 1 σǫ = The variance of {X t } is not limited and therefore {X t }

More information

Some notes about signals, orthogonal polynomials and linear algebra

Some notes about signals, orthogonal polynomials and linear algebra Some notes about signals, orthogonal polynomials and linear algebra Adhemar Bultheel Report TW 180, November 1992 Revised February 1993 n Katholieke Universiteit Leuven Department of Computer Science Celestijnenlaan

More information

X (z) = n= 1. Ã! X (z) x [n] X (z) = Z fx [n]g x [n] = Z 1 fx (z)g. r n x [n] ª e jnω

X (z) = n= 1. Ã! X (z) x [n] X (z) = Z fx [n]g x [n] = Z 1 fx (z)g. r n x [n] ª e jnω 3 The z-transform ² Two advantages with the z-transform:. The z-transform is a generalization of the Fourier transform for discrete-time signals; which encompasses a broader class of sequences. The z-transform

More information

Parametric Signal Modeling and Linear Prediction Theory 1. Discrete-time Stochastic Processes (cont d)

Parametric Signal Modeling and Linear Prediction Theory 1. Discrete-time Stochastic Processes (cont d) Parametric Signal Modeling and Linear Prediction Theory 1. Discrete-time Stochastic Processes (cont d) Electrical & Computer Engineering North Carolina State University Acknowledgment: ECE792-41 slides

More information

Università di Pavia. Forecasting. Eduardo Rossi

Università di Pavia. Forecasting. Eduardo Rossi Università di Pavia Forecasting Eduardo Rossi Mean Squared Error Forecast of Y t+1 based on a set of variables observed at date t, X t : Yt+1 t. The loss function MSE(Y t+1 t ) = E[Y t+1 Y t+1 t ]2 The

More information

Stabilization with Disturbance Attenuation over a Gaussian Channel

Stabilization with Disturbance Attenuation over a Gaussian Channel Proceedings of the 46th IEEE Conference on Decision and Control New Orleans, LA, USA, Dec. 1-14, 007 Stabilization with Disturbance Attenuation over a Gaussian Channel J. S. Freudenberg, R. H. Middleton,

More information

CHAPTER 2 RANDOM PROCESSES IN DISCRETE TIME

CHAPTER 2 RANDOM PROCESSES IN DISCRETE TIME CHAPTER 2 RANDOM PROCESSES IN DISCRETE TIME Shri Mata Vaishno Devi University, (SMVDU), 2013 Page 13 CHAPTER 2 RANDOM PROCESSES IN DISCRETE TIME When characterizing or modeling a random variable, estimates

More information

The goal of the Wiener filter is to filter out noise that has corrupted a signal. It is based on a statistical approach.

The goal of the Wiener filter is to filter out noise that has corrupted a signal. It is based on a statistical approach. Wiener filter From Wikipedia, the free encyclopedia In signal processing, the Wiener filter is a filter proposed by Norbert Wiener during the 1940s and published in 1949. [1] Its purpose is to reduce the

More information

5 Transfer function modelling

5 Transfer function modelling MSc Further Time Series Analysis 5 Transfer function modelling 5.1 The model Consider the construction of a model for a time series (Y t ) whose values are influenced by the earlier values of a series

More information

Nonparametric Function Estimation with Infinite-Order Kernels

Nonparametric Function Estimation with Infinite-Order Kernels Nonparametric Function Estimation with Infinite-Order Kernels Arthur Berg Department of Statistics, University of Florida March 15, 2008 Kernel Density Estimation (IID Case) Let X 1,..., X n iid density

More information

Time Series Examples Sheet

Time Series Examples Sheet Lent Term 2001 Richard Weber Time Series Examples Sheet This is the examples sheet for the M. Phil. course in Time Series. A copy can be found at: http://www.statslab.cam.ac.uk/~rrw1/timeseries/ Throughout,

More information

6.3 Forecasting ARMA processes

6.3 Forecasting ARMA processes 6.3. FORECASTING ARMA PROCESSES 123 6.3 Forecasting ARMA processes The purpose of forecasting is to predict future values of a TS based on the data collected to the present. In this section we will discuss

More information

STAD57 Time Series Analysis. Lecture 8

STAD57 Time Series Analysis. Lecture 8 STAD57 Time Series Analysis Lecture 8 1 ARMA Model Will be using ARMA models to describe times series dynamics: ( B) X ( B) W X X X W W W t 1 t1 p t p t 1 t1 q tq Model must be causal (i.e. stationary)

More information

3. ARMA Modeling. Now: Important class of stationary processes

3. ARMA Modeling. Now: Important class of stationary processes 3. ARMA Modeling Now: Important class of stationary processes Definition 3.1: (ARMA(p, q) process) Let {ɛ t } t Z WN(0, σ 2 ) be a white noise process. The process {X t } t Z is called AutoRegressive-Moving-Average

More information

INTRODUCTION Noise is present in many situations of daily life for ex: Microphones will record noise and speech. Goal: Reconstruct original signal Wie

INTRODUCTION Noise is present in many situations of daily life for ex: Microphones will record noise and speech. Goal: Reconstruct original signal Wie WIENER FILTERING Presented by N.Srikanth(Y8104060), M.Manikanta PhaniKumar(Y8104031). INDIAN INSTITUTE OF TECHNOLOGY KANPUR Electrical Engineering dept. INTRODUCTION Noise is present in many situations

More information

4 Derivations of the Discrete-Time Kalman Filter

4 Derivations of the Discrete-Time Kalman Filter Technion Israel Institute of Technology, Department of Electrical Engineering Estimation and Identification in Dynamical Systems (048825) Lecture Notes, Fall 2009, Prof N Shimkin 4 Derivations of the Discrete-Time

More information

1. Nonlinear Equations. This lecture note excerpted parts from Michael Heath and Max Gunzburger. f(x) = 0

1. Nonlinear Equations. This lecture note excerpted parts from Michael Heath and Max Gunzburger. f(x) = 0 Numerical Analysis 1 1. Nonlinear Equations This lecture note excerpted parts from Michael Heath and Max Gunzburger. Given function f, we seek value x for which where f : D R n R n is nonlinear. f(x) =

More information

Adaptive Beamforming Algorithms

Adaptive Beamforming Algorithms S. R. Zinka srinivasa_zinka@daiict.ac.in October 29, 2014 Outline 1 Least Mean Squares 2 Sample Matrix Inversion 3 Recursive Least Squares 4 Accelerated Gradient Approach 5 Conjugate Gradient Method Outline

More information

Levinson Durbin Recursions: I

Levinson Durbin Recursions: I Levinson Durbin Recursions: I note: B&D and S&S say Durbin Levinson but Levinson Durbin is more commonly used (Levinson, 1947, and Durbin, 1960, are source articles sometimes just Levinson is used) recursions

More information

f (1 0.5)/n Z =

f (1 0.5)/n Z = Math 466/566 - Homework 4. We want to test a hypothesis involving a population proportion. The unknown population proportion is p. The null hypothesis is p = / and the alternative hypothesis is p > /.

More information

The Kalman Filter ImPr Talk

The Kalman Filter ImPr Talk The Kalman Filter ImPr Talk Ged Ridgway Centre for Medical Image Computing November, 2006 Outline What is the Kalman Filter? State Space Models Kalman Filter Overview Bayesian Updating of Estimates Kalman

More information

Levinson Durbin Recursions: I

Levinson Durbin Recursions: I Levinson Durbin Recursions: I note: B&D and S&S say Durbin Levinson but Levinson Durbin is more commonly used (Levinson, 1947, and Durbin, 1960, are source articles sometimes just Levinson is used) recursions

More information

EC402: Serial Correlation. Danny Quah Economics Department, LSE Lent 2015

EC402: Serial Correlation. Danny Quah Economics Department, LSE Lent 2015 EC402: Serial Correlation Danny Quah Economics Department, LSE Lent 2015 OUTLINE 1. Stationarity 1.1 Covariance stationarity 1.2 Explicit Models. Special cases: ARMA processes 2. Some complex numbers.

More information

SIMON FRASER UNIVERSITY School of Engineering Science

SIMON FRASER UNIVERSITY School of Engineering Science SIMON FRASER UNIVERSITY School of Engineering Science Course Outline ENSC 810-3 Digital Signal Processing Calendar Description This course covers advanced digital signal processing techniques. The main

More information

ECE 8440 Unit 13 Sec0on Effects of Round- Off Noise in Digital Filters

ECE 8440 Unit 13 Sec0on Effects of Round- Off Noise in Digital Filters ECE 8440 Unit 13 Sec0on 6.9 - Effects of Round- Off Noise in Digital Filters 1 We have already seen that if a wide- sense staonary random signal x(n) is applied as input to a LTI system, the power density

More information

Advanced Digital Signal Processing -Introduction

Advanced Digital Signal Processing -Introduction Advanced Digital Signal Processing -Introduction LECTURE-2 1 AP9211- ADVANCED DIGITAL SIGNAL PROCESSING UNIT I DISCRETE RANDOM SIGNAL PROCESSING Discrete Random Processes- Ensemble Averages, Stationary

More information

Multivariate ARMA Processes

Multivariate ARMA Processes LECTURE 8 Multivariate ARMA Processes A vector y(t) of n elements is said to follow an n-variate ARMA process of orders p and q if it satisfies the equation (1) A 0 y(t) + A 1 y(t 1) + + A p y(t p) = M

More information

Adaptive Filtering. Squares. Alexander D. Poularikas. Fundamentals of. Least Mean. with MATLABR. University of Alabama, Huntsville, AL.

Adaptive Filtering. Squares. Alexander D. Poularikas. Fundamentals of. Least Mean. with MATLABR. University of Alabama, Huntsville, AL. Adaptive Filtering Fundamentals of Least Mean Squares with MATLABR Alexander D. Poularikas University of Alabama, Huntsville, AL CRC Press Taylor & Francis Croup Boca Raton London New York CRC Press is

More information

Calculation of ACVF for ARMA Process: I consider causal ARMA(p, q) defined by

Calculation of ACVF for ARMA Process: I consider causal ARMA(p, q) defined by Calculation of ACVF for ARMA Process: I consider causal ARMA(p, q) defined by φ(b)x t = θ(b)z t, {Z t } WN(0, σ 2 ) want to determine ACVF {γ(h)} for this process, which can be done using four complementary

More information

MATH 1231 MATHEMATICS 1B Calculus Section 4.4: Taylor & Power series.

MATH 1231 MATHEMATICS 1B Calculus Section 4.4: Taylor & Power series. MATH 1231 MATHEMATICS 1B 2010. For use in Dr Chris Tisdell s lectures. Calculus Section 4.4: Taylor & Power series. 1. What is a Taylor series? 2. Convergence of Taylor series 3. Common Maclaurin series

More information

Spectral analysis of two doubly infinite Jacobi operators

Spectral analysis of two doubly infinite Jacobi operators Spectral analysis of two doubly infinite Jacobi operators František Štampach jointly with Mourad E. H. Ismail Stockholm University Spectral Theory and Applications conference in memory of Boris Pavlov

More information

Each problem is worth 25 points, and you may solve the problems in any order.

Each problem is worth 25 points, and you may solve the problems in any order. EE 120: Signals & Systems Department of Electrical Engineering and Computer Sciences University of California, Berkeley Midterm Exam #2 April 11, 2016, 2:10-4:00pm Instructions: There are four questions

More information

ECE531 Lecture 11: Dynamic Parameter Estimation: Kalman-Bucy Filter

ECE531 Lecture 11: Dynamic Parameter Estimation: Kalman-Bucy Filter ECE531 Lecture 11: Dynamic Parameter Estimation: Kalman-Bucy Filter D. Richard Brown III Worcester Polytechnic Institute 09-Apr-2009 Worcester Polytechnic Institute D. Richard Brown III 09-Apr-2009 1 /

More information

NAME: ht () 1 2π. Hj0 ( ) dω Find the value of BW for the system having the following impulse response.

NAME: ht () 1 2π. Hj0 ( ) dω Find the value of BW for the system having the following impulse response. University of California at Berkeley Department of Electrical Engineering and Computer Sciences Professor J. M. Kahn, EECS 120, Fall 1998 Final Examination, Wednesday, December 16, 1998, 5-8 pm NAME: 1.

More information

Lecture 4 - Spectral Estimation

Lecture 4 - Spectral Estimation Lecture 4 - Spectral Estimation The Discrete Fourier Transform The Discrete Fourier Transform (DFT) is the equivalent of the continuous Fourier Transform for signals known only at N instants separated

More information

Lesson 9: Autoregressive-Moving Average (ARMA) models

Lesson 9: Autoregressive-Moving Average (ARMA) models Lesson 9: Autoregressive-Moving Average (ARMA) models Dipartimento di Ingegneria e Scienze dell Informazione e Matematica Università dell Aquila, umberto.triacca@ec.univaq.it Introduction We have seen

More information

Stability. X(s) Y(s) = (s + 2) 2 (s 2) System has 2 poles: points where Y(s) -> at s = +2 and s = -2. Y(s) 8X(s) G 1 G 2

Stability. X(s) Y(s) = (s + 2) 2 (s 2) System has 2 poles: points where Y(s) -> at s = +2 and s = -2. Y(s) 8X(s) G 1 G 2 Stability 8X(s) X(s) Y(s) = (s 2) 2 (s 2) System has 2 poles: points where Y(s) -> at s = 2 and s = -2 If all poles are in region where s < 0, system is stable in Fourier language s = jω G 0 - x3 x7 Y(s)

More information

Statistical and Adaptive Signal Processing

Statistical and Adaptive Signal Processing r Statistical and Adaptive Signal Processing Spectral Estimation, Signal Modeling, Adaptive Filtering and Array Processing Dimitris G. Manolakis Massachusetts Institute of Technology Lincoln Laboratory

More information

Parametric Signal Modeling and Linear Prediction Theory 4. The Levinson-Durbin Recursion

Parametric Signal Modeling and Linear Prediction Theory 4. The Levinson-Durbin Recursion Parametric Signal Modeling and Linear Prediction Theory 4. The Levinson-Durbin Recursion Electrical & Computer Engineering North Carolina State University Acknowledgment: ECE792-41 slides were adapted

More information

EE538 Final Exam Fall :20 pm -5:20 pm PHYS 223 Dec. 17, Cover Sheet

EE538 Final Exam Fall :20 pm -5:20 pm PHYS 223 Dec. 17, Cover Sheet EE538 Final Exam Fall 005 3:0 pm -5:0 pm PHYS 3 Dec. 17, 005 Cover Sheet Test Duration: 10 minutes. Open Book but Closed Notes. Calculators ARE allowed!! This test contains five problems. Each of the five

More information

C. A. Bouman: Digital Image Processing - January 29, Image Restortation

C. A. Bouman: Digital Image Processing - January 29, Image Restortation C. A. Bouman: Digital Image Processing - January 29, 2013 1 Image Restortation Problem: You want to know some image X. But you only have a corrupted version Y. How do you determine X from Y? Corruption

More information

Review of Discrete-Time System

Review of Discrete-Time System Review of Discrete-Time System Electrical & Computer Engineering University of Maryland, College Park Acknowledgment: ENEE630 slides were based on class notes developed by Profs. K.J. Ray Liu and Min Wu.

More information

Methods of Integration

Methods of Integration Methods of Integration Professor D. Olles January 8, 04 Substitution The derivative of a composition of functions can be found using the chain rule form d dx [f (g(x))] f (g(x)) g (x) Rewriting the derivative

More information

Applied Mathematics Masters Examination Fall 2016, August 18, 1 4 pm.

Applied Mathematics Masters Examination Fall 2016, August 18, 1 4 pm. Applied Mathematics Masters Examination Fall 16, August 18, 1 4 pm. Each of the fifteen numbered questions is worth points. All questions will be graded, but your score for the examination will be the

More information

Chapter 9: Forecasting

Chapter 9: Forecasting Chapter 9: Forecasting One of the critical goals of time series analysis is to forecast (predict) the values of the time series at times in the future. When forecasting, we ideally should evaluate the

More information

Signal Analysis, Systems, Transforms

Signal Analysis, Systems, Transforms Michael J. Corinthios Signal Analysis, Systems, Transforms Engineering Book (English) August 29, 2007 Springer Contents Discrete-Time Signals and Systems......................... Introduction.............................................2

More information

LEAST SQUARES APPROXIMATION

LEAST SQUARES APPROXIMATION LEAST SQUARES APPROXIMATION One more approach to approximating a function f (x) on an interval a x b is to seek an approximation p(x) with a small average error over the interval of approximation. A convenient

More information

Forecasting. This optimal forecast is referred to as the Minimum Mean Square Error Forecast. This optimal forecast is unbiased because

Forecasting. This optimal forecast is referred to as the Minimum Mean Square Error Forecast. This optimal forecast is unbiased because Forecasting 1. Optimal Forecast Criterion - Minimum Mean Square Error Forecast We have now considered how to determine which ARIMA model we should fit to our data, we have also examined how to estimate

More information

Signals and Spectra - Review

Signals and Spectra - Review Signals and Spectra - Review SIGNALS DETERMINISTIC No uncertainty w.r.t. the value of a signal at any time Modeled by mathematical epressions RANDOM some degree of uncertainty before the signal occurs

More information

Lecture 2 OKAN UNIVERSITY FACULTY OF ENGINEERING AND ARCHITECTURE

Lecture 2 OKAN UNIVERSITY FACULTY OF ENGINEERING AND ARCHITECTURE OKAN UNIVERSITY FACULTY OF ENGINEERING AND ARCHITECTURE EEE 43 DIGITAL SIGNAL PROCESSING (DSP) 2 DIFFERENCE EQUATIONS AND THE Z- TRANSFORM FALL 22 Yrd. Doç. Dr. Didem Kivanc Tureli didemk@ieee.org didem.kivanc@okan.edu.tr

More information

Adaptive MMSE Equalizer with Optimum Tap-length and Decision Delay

Adaptive MMSE Equalizer with Optimum Tap-length and Decision Delay Adaptive MMSE Equalizer with Optimum Tap-length and Decision Delay Yu Gong, Xia Hong and Khalid F. Abu-Salim School of Systems Engineering The University of Reading, Reading RG6 6AY, UK E-mail: {y.gong,x.hong,k.f.abusalem}@reading.ac.uk

More information

Topic 4: The Z Transform

Topic 4: The Z Transform ELEN E480: Digital Signal Processing Topic 4: The Z Transform. The Z Transform 2. Inverse Z Transform . The Z Transform Powerful tool for analyzing & designing DT systems Generalization of the DTFT: G(z)

More information

Q3. Derive the Wiener filter (non-causal) for a stationary process with given spectral characteristics

Q3. Derive the Wiener filter (non-causal) for a stationary process with given spectral characteristics Q3. Derive the Wiener filter (non-causal) for a stationary process with given spectral characteristics McMaster University 1 Background x(t) z(t) WF h(t) x(t) n(t) Figure 1: Signal Flow Diagram Goal: filter

More information

Interactions of Information Theory and Estimation in Single- and Multi-user Communications

Interactions of Information Theory and Estimation in Single- and Multi-user Communications Interactions of Information Theory and Estimation in Single- and Multi-user Communications Dongning Guo Department of Electrical Engineering Princeton University March 8, 2004 p 1 Dongning Guo Communications

More information

Power series solutions for 2nd order linear ODE s (not necessarily with constant coefficients) a n z n. n=0

Power series solutions for 2nd order linear ODE s (not necessarily with constant coefficients) a n z n. n=0 Lecture 22 Power series solutions for 2nd order linear ODE s (not necessarily with constant coefficients) Recall a few facts about power series: a n z n This series in z is centered at z 0. Here z can

More information

EE 225a Digital Signal Processing Supplementary Material

EE 225a Digital Signal Processing Supplementary Material EE 225A DIGITAL SIGAL PROCESSIG SUPPLEMETARY MATERIAL EE 225a Digital Signal Processing Supplementary Material. Allpass Sequences A sequence h a ( n) ote that this is true iff which in turn is true iff

More information

Math 471. Numerical methods Introduction

Math 471. Numerical methods Introduction Math 471. Numerical methods Introduction Section 1.1 1.4 of Bradie 1.1 Algorithms Here is an analogy between Numerical Methods and Gastronomy: Calculus, Lin Alg., Diff. eq. Ingredients Algorithm Recipe

More information

How might we evaluate this? Suppose that, by some good luck, we knew that. x 2 5. x 2 dx 5

How might we evaluate this? Suppose that, by some good luck, we knew that. x 2 5. x 2 dx 5 8.4 1 8.4 Partial Fractions Consider the following integral. 13 2x (1) x 2 x 2 dx How might we evaluate this? Suppose that, by some good luck, we knew that 13 2x (2) x 2 x 2 = 3 x 2 5 x + 1 We could then

More information

For a stochastic process {Y t : t = 0, ±1, ±2, ±3, }, the mean function is defined by (2.2.1) ± 2..., γ t,

For a stochastic process {Y t : t = 0, ±1, ±2, ±3, }, the mean function is defined by (2.2.1) ± 2..., γ t, CHAPTER 2 FUNDAMENTAL CONCEPTS This chapter describes the fundamental concepts in the theory of time series models. In particular, we introduce the concepts of stochastic processes, mean and covariance

More information

Statistical signal processing

Statistical signal processing Statistical signal processing Short overview of the fundamentals Outline Random variables Random processes Stationarity Ergodicity Spectral analysis Random variable and processes Intuition: A random variable

More information

, B = (b) Use induction to prove that, for all positive integers n, f(n) is divisible by 4. (a) Use differentiation to find f (x).

, B = (b) Use induction to prove that, for all positive integers n, f(n) is divisible by 4. (a) Use differentiation to find f (x). Edexcel FP1 FP1 Practice Practice Papers A and B Papers A and B PRACTICE PAPER A 1. A = 2 1, B = 4 3 3 1, I = 4 2 1 0. 0 1 (a) Show that AB = ci, stating the value of the constant c. (b) Hence, or otherwise,

More information

UCSD ECE 153 Handout #46 Prof. Young-Han Kim Thursday, June 5, Solutions to Homework Set #8 (Prepared by TA Fatemeh Arbabjolfaei)

UCSD ECE 153 Handout #46 Prof. Young-Han Kim Thursday, June 5, Solutions to Homework Set #8 (Prepared by TA Fatemeh Arbabjolfaei) UCSD ECE 53 Handout #46 Prof. Young-Han Kim Thursday, June 5, 04 Solutions to Homework Set #8 (Prepared by TA Fatemeh Arbabjolfaei). Discrete-time Wiener process. Let Z n, n 0 be a discrete time white

More information

9740/01 October/November MATHEMATICS (H2) Paper 1 Suggested Solutions. (ii)

9740/01 October/November MATHEMATICS (H2) Paper 1 Suggested Solutions. (ii) GCE A Level October/November 9 Suggested Solutions Mathematics H (97/) version. MATHEMATICS (H) Paper Suggested Solutions. Topic: Matrices (i) Given that u n is a quadratic polynomial in n, Let u n an

More information

Implicit 3-D depth migration by wavefield extrapolation with helical boundary conditions

Implicit 3-D depth migration by wavefield extrapolation with helical boundary conditions Stanford Exploration Project, Report 97, July 8, 1998, pages 1 13 Implicit 3-D depth migration by wavefield extrapolation with helical boundary conditions James Rickett, Jon Claerbout, and Sergey Fomel

More information

a( i), where N i N ) is accurately d( i), where N i N. The derivative at ( ) ( ) ( )

a( i), where N i N ) is accurately d( i), where N i N. The derivative at ( ) ( ) ( ) More on implementing the derivative filter John C. Bancroft ABSTRACT More on implementing the derivative filter An update is presented on finding faster and or more accurate implementations of differentiator

More information

Modeling and testing long memory in random fields

Modeling and testing long memory in random fields Modeling and testing long memory in random fields Frédéric Lavancier lavancier@math.univ-lille1.fr Université Lille 1 LS-CREST Paris 24 janvier 6 1 Introduction Long memory random fields Motivations Previous

More information

Taylor and Laurent Series

Taylor and Laurent Series Chapter 4 Taylor and Laurent Series 4.. Taylor Series 4... Taylor Series for Holomorphic Functions. In Real Analysis, the Taylor series of a given function f : R R is given by: f (x + f (x (x x + f (x

More information

Today. ESE 531: Digital Signal Processing. IIR Filter Design. Impulse Invariance. Impulse Invariance. Impulse Invariance. ω < π.

Today. ESE 531: Digital Signal Processing. IIR Filter Design. Impulse Invariance. Impulse Invariance. Impulse Invariance. ω < π. Today ESE 53: Digital Signal Processing! IIR Filter Design " Lec 8: March 30, 207 IIR Filters and Adaptive Filters " Bilinear Transformation! Transformation of DT Filters! Adaptive Filters! LMS Algorithm

More information

Open Economy Macroeconomics: Theory, methods and applications

Open Economy Macroeconomics: Theory, methods and applications Open Economy Macroeconomics: Theory, methods and applications Lecture 4: The state space representation and the Kalman Filter Hernán D. Seoane UC3M January, 2016 Today s lecture State space representation

More information

Proof by induction ME 8

Proof by induction ME 8 Proof by induction ME 8 n Let f ( n) 9, where n. f () 9 8, which is divisible by 8. f ( n) is divisible by 8 when n =. Assume that for n =, f ( ) 9 is divisible by 8 for. f ( ) 9 9.9 9(9 ) f ( ) f ( )

More information

Lecture 8: Policy Gradient

Lecture 8: Policy Gradient Lecture 8: Policy Gradient Hado van Hasselt Outline 1 Introduction 2 Finite Difference Policy Gradient 3 Monte-Carlo Policy Gradient 4 Actor-Critic Policy Gradient Introduction Vapnik s rule Never solve

More information

Chapter 6: Rational Expr., Eq., and Functions Lecture notes Math 1010

Chapter 6: Rational Expr., Eq., and Functions Lecture notes Math 1010 Section 6.1: Rational Expressions and Functions Definition of a rational expression Let u and v be polynomials. The algebraic expression u v is a rational expression. The domain of this rational expression

More information

EE 4372 Tomography. Carlos E. Davila, Dept. of Electrical Engineering Southern Methodist University

EE 4372 Tomography. Carlos E. Davila, Dept. of Electrical Engineering Southern Methodist University EE 4372 Tomography Carlos E. Davila, Dept. of Electrical Engineering Southern Methodist University EE 4372, SMU Department of Electrical Engineering 86 Tomography: Background 1-D Fourier Transform: F(

More information

On Input Design for System Identification

On Input Design for System Identification On Input Design for System Identification Input Design Using Markov Chains CHIARA BRIGHENTI Masters Degree Project Stockholm, Sweden March 2009 XR-EE-RT 2009:002 Abstract When system identification methods

More information

Gaussian vectors and central limit theorem

Gaussian vectors and central limit theorem Gaussian vectors and central limit theorem Samy Tindel Purdue University Probability Theory 2 - MA 539 Samy T. Gaussian vectors & CLT Probability Theory 1 / 86 Outline 1 Real Gaussian random variables

More information

... Econometric Methods for the Analysis of Dynamic General Equilibrium Models

... Econometric Methods for the Analysis of Dynamic General Equilibrium Models ... Econometric Methods for the Analysis of Dynamic General Equilibrium Models 1 Overview Multiple Equation Methods State space-observer form Three Examples of Versatility of state space-observer form:

More information

Ch. 15 Forecasting. 1.1 Forecasts Based on Conditional Expectations

Ch. 15 Forecasting. 1.1 Forecasts Based on Conditional Expectations Ch 15 Forecasting Having considered in Chapter 14 some of the properties of ARMA models, we now show how they may be used to forecast future values of an observed time series For the present we proceed

More information

Probability Background

Probability Background Probability Background Namrata Vaswani, Iowa State University August 24, 2015 Probability recap 1: EE 322 notes Quick test of concepts: Given random variables X 1, X 2,... X n. Compute the PDF of the second

More information

REVERSE CHAIN RULE CALCULUS 7. Dr Adrian Jannetta MIMA CMath FRAS INU0115/515 (MATHS 2) Reverse Chain Rule 1/12 Adrian Jannetta

REVERSE CHAIN RULE CALCULUS 7. Dr Adrian Jannetta MIMA CMath FRAS INU0115/515 (MATHS 2) Reverse Chain Rule 1/12 Adrian Jannetta REVERSE CHAIN RULE CALCULUS 7 INU05/55 (MATHS 2) Dr Adrian Jannetta MIMA CMath FRAS Reverse Chain Rule /2 Adrian Jannetta Reversing the chain rule In differentiation the chain rule is used to get the derivative

More information

An example of Bayesian reasoning Consider the one-dimensional deconvolution problem with various degrees of prior information.

An example of Bayesian reasoning Consider the one-dimensional deconvolution problem with various degrees of prior information. An example of Bayesian reasoning Consider the one-dimensional deconvolution problem with various degrees of prior information. Model: where g(t) = a(t s)f(s)ds + e(t), a(t) t = (rapidly). The problem,

More information

Lecture 9 Infinite Impulse Response Filters

Lecture 9 Infinite Impulse Response Filters Lecture 9 Infinite Impulse Response Filters Outline 9 Infinite Impulse Response Filters 9 First-Order Low-Pass Filter 93 IIR Filter Design 5 93 CT Butterworth filter design 5 93 Bilinear transform 7 9

More information

A. Poulimenos, M. Spiridonakos, and S. Fassois

A. Poulimenos, M. Spiridonakos, and S. Fassois PARAMETRIC TIME-DOMAIN METHODS FOR NON-STATIONARY RANDOM VIBRATION IDENTIFICATION AND ANALYSIS: AN OVERVIEW AND COMPARISON A. Poulimenos, M. Spiridonakos, and S. Fassois DEPARTMENT OF MECHANICAL & AERONAUTICAL

More information

x n cos 2x dx. dx = nx n 1 and v = 1 2 sin(2x). Andreas Fring (City University London) AS1051 Lecture Autumn / 36

x n cos 2x dx. dx = nx n 1 and v = 1 2 sin(2x). Andreas Fring (City University London) AS1051 Lecture Autumn / 36 We saw in Example 5.4. that we sometimes need to apply integration by parts several times in the course of a single calculation. Example 5.4.4: For n let S n = x n cos x dx. Find an expression for S n

More information

Math 56 Homework 1 Michael Downs. ne n 10 + ne n (1)

Math 56 Homework 1 Michael Downs. ne n 10 + ne n (1) . Problem (a) Yes. The following equation: ne n + ne n () holds for all n R but, since we re only concerned with the asymptotic behavior as n, let us only consider n >. Dividing both sides by n( + ne n

More information

MA8109 Stochastic Processes in Systems Theory Autumn 2013

MA8109 Stochastic Processes in Systems Theory Autumn 2013 Norwegian University of Science and Technology Department of Mathematical Sciences MA819 Stochastic Processes in Systems Theory Autumn 213 1 MA819 Exam 23, problem 3b This is a linear equation of the form

More information

STA205 Probability: Week 8 R. Wolpert

STA205 Probability: Week 8 R. Wolpert INFINITE COIN-TOSS AND THE LAWS OF LARGE NUMBERS The traditional interpretation of the probability of an event E is its asymptotic frequency: the limit as n of the fraction of n repeated, similar, and

More information

OCR Maths FP1. Topic Questions from Papers. Summation of Series. Answers

OCR Maths FP1. Topic Questions from Papers. Summation of Series. Answers OCR Maths FP Topic Questions from Papers Summation of Series Answers PhysicsAndMathsTutor.com . Σr +Σr + Σ Σr = n(n +)(n +) Σr = n(n +) Σ = n PhysicsAndMathsTutor.com Consider the sum of three separate

More information

Long-range dependence

Long-range dependence Long-range dependence Kechagias Stefanos University of North Carolina at Chapel Hill May 23, 2013 Kechagias Stefanos (UNC) Long-range dependence May 23, 2013 1 / 45 Outline 1 Introduction to time series

More information

TIME SERIES ANALYSIS. Forecasting and Control. Wiley. Fifth Edition GWILYM M. JENKINS GEORGE E. P. BOX GREGORY C. REINSEL GRETA M.

TIME SERIES ANALYSIS. Forecasting and Control. Wiley. Fifth Edition GWILYM M. JENKINS GEORGE E. P. BOX GREGORY C. REINSEL GRETA M. TIME SERIES ANALYSIS Forecasting and Control Fifth Edition GEORGE E. P. BOX GWILYM M. JENKINS GREGORY C. REINSEL GRETA M. LJUNG Wiley CONTENTS PREFACE TO THE FIFTH EDITION PREFACE TO THE FOURTH EDITION

More information