Technische Universität Kaiserslautern WS 2016/17 Fachbereich Mathematik Prof. Dr. J. Franke 19. January 2017

Size: px
Start display at page:

Download "Technische Universität Kaiserslautern WS 2016/17 Fachbereich Mathematik Prof. Dr. J. Franke 19. January 2017"

Transcription

1 Technische Universität Kaiserslautern WS 2016/17 Fachbereich Mathematik Prof. Dr. J. Franke 19. January 2017 Exercises for Computational Modelling with Statistics IV This exercise is about the spectral analysis of signals as a tool to discover hidden periodicities or quasiperiodicities in data. Exercise 1 (Spectral analysis of periodic signals) The original spectral analysis developped by Schuster (1898) was meant to discover hidden periodicities in signals which are observed with measurement noise. We develop this approach step by step. In practice one usually has larger sample sizes, but we focus on sample size N=128 such that the signals are clearly visible in the plots. Note that N = 2 7 ; we use it as the FFT-algorithm is most efficient for powers of 2. In practice, if one observes a continuous signal (like a EEG recording) and if one can choose the rate of discretization, then it is optimal for the spectral analysis, to choose it such that the resulting sample size is a power of 2. We ll look at various examples and store the relevant results in the following matrices whose size we now fix: FREQ=zeros(6,3); AMP=zeros(6,3); a) i) Generate a pure cosine signal with period l = 16, amplitude A = 1, angular frequency α = 2π l : N=128; alpha = 2*pi/16; t=1:1:n; X = cos(alpha*t); Have a look at the signal data plot(t,x, * ) and at the interpolated continuous signal ii) We use the fast Fourier transform (FFT) to calculate the discrete Fourier transform (DFT) of the 1 data. We use the scaling factor N as in the lecture. FX is complex-valued. To look at a plot, we calculate the periodogram which is the square of the absolute value of the DFT: Just have a look at it: plot(px) PX is a symmetric function. Therefore, we know it completely if we only look at the first half of it. Also, for interpretability, we plot it against angular frequency, more precisely against the equidistant Fourier frequencies ω j = 2πj N, j = 0,..., N 2. omega = 2*pi*(0:1:N/2)/N; ; To recover the frequency from the signal, we have to look where the periodogram assumes its maximum. We are also interested in the value of the maximum. The MATLAB command [M,I] = max(px) determines the value M of the maximum and the index I of the vector PX where the maximum is located. To relate I to angular frequency, frequency and period of the signal component we have to look at the corresponding value of the vector omega: angfreq = omega(i); freq = angfreq/(2*pi) If everything is ok, you should recover exactly the period l. This only works so nicely as the corresponding 1

2 angular frequency is a Fourier frequency, or, equivalently, N is an integer multiple of l, as N = 8l. iii) Please store M and freq: AMP(1,1) = M; FREQ(1,1) = freq; iv) To check the influence of a different amplitude A and a phase shift φ repeat step i) and ii) for the following signal: A=2; phi = pi/4; X = A*cos(alpha*t+phi); [M,I] = max(px) angfreq = omega(i); freq = angfreq/(2*pi) Please store M and freq: AMP(2,1)=M; FREQ(2,1)=freq; Note that the maximum value M of the periodogram is quadratic in A, and M and I are not influenced by the phase shift φ. The phase information is lost by squaring the DFT. You can play around with different values of A and φ (e.g. A=3; phi=0.123;). b) Now we look at a cosine function with an angular frequency α which is not a Fourier frequency, but lies between two of them alpha = 2*pi*7.5/128; X = cos(alpha*t); [M,I] = max(px) From the plot, the periodogram value at I-1 is almost as large, so we also store it: M2=PX(I-1); The true frequency lies somewhere between the two maxima. We just interpolate and set: angfreq = (omega(i)+omega(i-1))/2; freq = angfreq/(2*pi) Compare this with the true frequency f = α 2π alpha/(2*pi) Please store M, M2 and freq: AMP(3,1)=M2; AMP(3,2)=M; FREQ(3,1)=freq; c) We now look at a superposition of 3 cosine functions (the sine is just a cosine with phase shift): alpha1 = 2*pi/32; alpha2 = 2*pi/16; alpha3 = 2*pi/8; X = cos(alpha1*t + pi/4) + 2*cos(alpha2*t) + 0.5*sin(alpha3*t); plot(t,x, * ) [M1,I] = max(px) angfreq = omega(i); freq1 = angfreq/(2*pi) 1 To get the next largest maximum, we set the value of PX at I to 0 and repeat the procedure: P=PX; P(I)=0; [M2,I] = max(p) angfreq = omega(i); freq2 = angfreq/(2*pi) 2 And finally for the third local maximum: 2

3 P(I)=0; [M3,I] = max(p) angfreq = omega(i); freq3 = angfreq/(2*pi) 3 As all 3 angular frequencies were Fourier frequencies you should recover the periods exactly in the order of amplitude size, i.e. 16, 32 and 8. Please store M1, M2, M3 and freq1, freq2, freq3: AMP(4,1)=M1; AMP(4,2)=M2; AMP(4,3)=M3; FREQ(4,1)=freq1; FREQ(4,2)=freq2; FREQ(4,3)=freq3; d) We add some Gaussian noise to the previous function: X = X + normrnd(0,1,1,n); plot(t,x, * ) We now repeat the steps from c): [M1,I] = max(px) angfreq = omega(i); freq1 = angfreq/(2*pi) 1 P=PX; P(I)=0; [M2,I] = max(p) angfreq = omega(i); freq2 = angfreq/(2*pi) 2 P(I)=0; [M3,I] = max(p) angfreq = omega(i); freq3 = angfreq/(2*pi) 3 We still see the two dominating frequencies corresponding to periods 16 and 32. The periodogram at the frequency corresponding to period 8 is still larger then the rest, but in another simulation it was only marginally larger than the next largest periodogram value at a completely different frequency. So it is not guaranteed that we discover a periodic component with such a small amplitude from only 128 observations. The chance to discover a particular periodic component is related to its amplitude, the noise standard deviation (in our case 1) and the sample size N. Please store M1, M2, M3 and freq1, freq2, freq3: AMP(5,1)=M1; AMP(5,2)=M2; AMP(5,3)=M3; FREQ(5,1)=freq1; FREQ(5,2)=freq2; FREQ(5,3)=freq3; e) Finally, we look at the function with non-fourier frequency α from b) with added random noise: alpha = 2*pi*7.5/128; X = cos(alpha*t) + normrnd(0,1,1,n); [M1,I1] = max(px); From the plot, the periodogram value at a neighbouring frequency is almost as large: P=PX; P(I)=0; [M2,I2] = max(p); The true maximum lies somewhere between I1 and I2, so we interpolate again: angfreq = (omega(i1)+omega(i2))/2; freq = angfreq/(2*pi) Please store M1, M2 and freq: AMP(6,1)=M1; AMP(6,2)=M2; FREQ(6,1)=freq; 3

4 Exercise 2 (The effect of the mean on spectral analysis) The signals of Exercise 1 have approximately mean 0 if we average over time t = 1,..., N. This is due to the symmetric fluctuation of the cosine around 0 and due to the law of large numbers for the noise component which we have chosen with expectation 0. In practice, signals frequently have a nonvanishing mean (compare the sunspot numbers example in the lecture). a) As in Exercise 1 a), generate a pure cosine signal with period l = 16, amplitude A = 1, angular frequency α = 2π l, but with a nonvanishing mean 1: N=128; alpha = 2*pi/16; t=1:1:n; X = 1 + cos(alpha*t); Then, calculate the periodogram: omega = 2*pi*(0:1:N/2)/N; Now the same with the mean-corrected data: X=X-mean(X); b) We now generate Gaussian white noise, i.e. a sequence of i.i.d. normal random variables, first with mean 0, and calculate the periodogram: X=normrnd(0,1,1,N); The periodogram fluctuates around the true spectral density which, in case of white noise, is a constant function. Now we change the mean from 0 to 1: X=1+X; and the same with mean 10: X=normrnd(10,1,1,N); Even for moderately large mean, the peak of the periodogram is so extreme that due to the scaling of the figure we cannot see the structure of the periodogram at the other frequencies. Therefore, mean correction is necessary: X=X-mean(X); c) We now consider a so-called autoregressive process of order 1 or AR(1)-process with Gaussian innovations, i.e. a signal satisfying X t = ax t 1 + Z t, t = 2,..., N, where Z 1,..., Z N are i.i.d. N (0, σ 2 ) random variables. To get a stationary signal, we need a < 1. To get a positive correlation between neighbouring values, we need a > 0. We choose a = 0.95, σ = 0.1. We first generate a realization of the AR(1)-process with mean 0: X=zeros(1,N); Z=normrnd(0,0.1,1,N); for j=2:n; X(j) = 0.95*X(j-1) + Z(j); end; Then, we change the mean to 1: X=1+X; We calculate the periodogram of the raw signal: 4

5 As before, the large peak at 0 obscures the remaining structure of the periodogram such that we have to subtract the mean Y=X-mean(X); FY = fft(y)/sqrt(n); PY = abs(fy). 2; PY = PY(1:(N/2+1));. plot(omega,py, * ) We see that the spectrum is large for small frequencies and then decreases for larger frequencies. This is typical for AR(1)-processes with a > 0, where the true unknown spectral density has a maximum at 0 and then decreases monotonically. Note also, that due to the mean correction, the periodogram at frequency 0 becomes 0. Otherwise, the periodogram is unchanged. To see this, plot the periodogram of the original signal X and that of the mean-corrected signal Y together omitting the first values at frequency 0: plot(omega(2:end),px(2:end), *,omega(2:end),py(2:end), o ) Save this figure, but not - as in the previous exercises - as MATLAB.fig file, but in the JPEG format which you can insert easily into text files (e.g. MS-Word, LaTeX, etc.). For this, in the plot window, click on File, then Save As (not Save!). The default option is MATLAB Figure (*.fig), but you can click on it, and then you get the other options for the figure format. Choose JPEG image (*.jpg). Then save the figure as AR1pgram.jpg. Exercise 3 (Smoothing the periodogram) On the average, if there are no strictly periodic components in the signal like in Exercise 1, the periodogram is equal to the spectral density. However, it is wildly fluctuating, and it becomes worse with increasing sample size. Therefore, to get a clear picture of the spectral structure of the signal, we have to smooth the periodogram, i.e. to calculate weighted averages of neighbouring periodogram values. We first generate data from a Gaussian autoregressive process of order 2 or AR(2)-process: X t = a 1 X t 1 + a 2 X t 2 + Z t, t = 2,..., N, with Z 1,..., Z N as in Exercise 1c). We choose a 1 = 0.95, a 2 = 0.8, σ = 0.1 and now N = 256. N=256; t=1:1:n; X=zeros(1,N); Z=normrnd(0,0.1,1,N); X(1)=Z(1); X(2)=0.95*X(1)+Z(2); for j=3:n; X(j) = 0.95*X(j-1) - 0.8*X(j-2) + Z(j); end; a) Now, we look at the periodogram. We can skip the mean correction as the simulated data have mean 0, but for real data a mean correction usually is necessary (compare Exercise 2). omega = 2*pi*(0:1:N/2)/N; We see that this has a peak at an angular frequency slightly larger than 1, i.e. a frequency slightly larger than 1 2π 0.16 and a period slightly smaller than 2π, i.e. approximately 6. However, the periodogram is rather random. b) To get a better impression about the form of the true spectrum, we smooth the periodogram P : SP (ω k ) = b w j P (ω k+j ), (1) j= b where w b,..., w b 0 are weights which sum up to 1 and are symmetric: b w j = 1, w j = w j, j = 0,..., b. j= b 5

6 The larger the b, the smoother the smoothed periodogram will be (depending also, but not so strongly, on the particular form of the weights). The signal analysis toolbox of MATLAB has fixed routines for all of this, but as we do not have it in the students licence, we have to do it in the pedestrian way. From (1) we see, that for e.g. calculating SP at frequency 0, we also need values P (ω j ), j = b,..., 1. The same problem appears at the right-hand boundary. So, we need an extended version of the periodogram, making use of the symmetry and periodicity of this function. P = abs(fx). 2; P=[P((N/2+1):N) P]; plot(p, * ) i) We define the vector (w b,..., w b ) of weights, where b = 4: W = [ ]; W=W/sum(W); The last command guarantees that the weights sum to 1. The operation (1) is a so-called convolution of the vector P and the vector W. In MATLAB, we can easily do it using the function conv: SP=conv(P,W, same ); We are only interested in the central part (as it already contains all the relevant information by symmetry and periodicity): SP=SP((N/2+1):(N+1)); This is the smoothed periodogram, and we can look at it: plot(omega,sp) This looks quite reasonable as it looks smooth, but not too smooth (showing yet a bit of small variability, such that we do not run into the danger of obscuring essential features of the spectrum by smoothing). We can also plot the smoothed periodogram and the raw periodogram together: plot(omega,sp,omega,px, * ) Please, save this figure as Spgram.jpg. ii) To check the influence of the weights, repeat the previous calculations with constant weights, but the same b: W = ones(1,9); W=W/sum(W); SP2=conv(P,W, same ); SP2=SP2((N/2+1):(N+1)); We compare the two smoothed periodograms: plot(omega,sp,omega,sp2) Please, save this figure as Spgram2.jpg. SP2 is a bit rougher. Weight vector with a maximum in the center and then decreasing to the left and right are preferable. Usually, they are generated as values of specific functions, the so-called kernels. Popular and also good kernels are the Gaussian kernel (= density of the standard normal distribution) or the Bartlett-Priestley kernel (quadratic function with a maximum at 0). ii) Finally, let us consider the effect of the choice of b by choosing a weight vector with b = 2: W = [ ]; W=W/sum(W); SP3=conv(P,W, same ); SP3=SP3((N/2+1):(N+1)); plot(omega,sp,omega,sp3) Please, save this figure as Spgram3.jpg. Note that this estimate of the true spectrum is still rather rough. Send me a mail (franke@mathematik.uni-kl.de) with the header Exercise 4. In the mail, you should list the names of the students who are members of the group handing in the solutions. Then, attach to the mail the file ex4work.mat and the 2 figure files AR1pgram.jpg, Spgram.jpg, Spgram2.jpg, Spgram3.jpg. Deadline: Sunday, February 5,

Statistics of Stochastic Processes

Statistics of Stochastic Processes Prof. Dr. J. Franke All of Statistics 4.1 Statistics of Stochastic Processes discrete time: sequence of r.v...., X 1, X 0, X 1, X 2,... X t R d in general. Here: d = 1. continuous time: random function

More information

Centre for Mathematical Sciences HT 2017 Mathematical Statistics

Centre for Mathematical Sciences HT 2017 Mathematical Statistics Lund University Stationary stochastic processes Centre for Mathematical Sciences HT 2017 Mathematical Statistics Computer exercise 3 in Stationary stochastic processes, HT 17. The purpose of this exercise

More information

Computer Exercise 0 Simulation of ARMA-processes

Computer Exercise 0 Simulation of ARMA-processes Lund University Time Series Analysis Mathematical Statistics Fall 2018 Centre for Mathematical Sciences Computer Exercise 0 Simulation of ARMA-processes The purpose of this computer exercise is to illustrate

More information

18.085: Summer 2016 Due: 3 August 2016 (in class) Problem Set 8

18.085: Summer 2016 Due: 3 August 2016 (in class) Problem Set 8 Problem Set 8 Unless otherwise specified, you may use MATLAB to assist with computations. provide a print-out of the code used and its output with your assignment. Please 1. More on relation between Fourier

More information

BSM510 Numerical Analysis

BSM510 Numerical Analysis BSM510 Numerical Analysis Polynomial Interpolation Prof. Manar Mohaisen Department of EEC Engineering Review of Precedent Lecture Polynomial Regression Multiple Linear Regression Nonlinear Regression Lecture

More information

Lab 4: Quantization, Oversampling, and Noise Shaping

Lab 4: Quantization, Oversampling, and Noise Shaping Lab 4: Quantization, Oversampling, and Noise Shaping Due Friday 04/21/17 Overview: This assignment should be completed with your assigned lab partner(s). Each group must turn in a report composed using

More information

The Discrete Fourier Transform (DFT) Properties of the DFT DFT-Specic Properties Power spectrum estimate. Alex Sheremet.

The Discrete Fourier Transform (DFT) Properties of the DFT DFT-Specic Properties Power spectrum estimate. Alex Sheremet. 4. April 2, 27 -order sequences Measurements produce sequences of numbers Measurement purpose: characterize a stochastic process. Example: Process: water surface elevation as a function of time Parameters:

More information

Time Series. Anthony Davison. c

Time Series. Anthony Davison. c Series Anthony Davison c 2008 http://stat.epfl.ch Periodogram 76 Motivation............................................................ 77 Lutenizing hormone data..................................................

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

SEISMIC WAVE PROPAGATION. Lecture 2: Fourier Analysis

SEISMIC WAVE PROPAGATION. Lecture 2: Fourier Analysis SEISMIC WAVE PROPAGATION Lecture 2: Fourier Analysis Fourier Series & Fourier Transforms Fourier Series Review of trigonometric identities Analysing the square wave Fourier Transform Transforms of some

More information

Problem Set 8 - Solution

Problem Set 8 - Solution Problem Set 8 - Solution Jonasz Słomka Unless otherwise specified, you may use MATLAB to assist with computations. provide a print-out of the code used and its output with your assignment. Please 1. More

More information

Centre for Mathematical Sciences HT 2018 Mathematical Statistics

Centre for Mathematical Sciences HT 2018 Mathematical Statistics Lund University Stationary stochastic processes Centre for Mathematical Sciences HT 2018 Mathematical Statistics Computer exercise 1 in Stationary stochastic processes, HT 18. The purpose of this computer

More information

Lectures 9-10: Polynomial and piecewise polynomial interpolation

Lectures 9-10: Polynomial and piecewise polynomial interpolation Lectures 9-1: Polynomial and piecewise polynomial interpolation Let f be a function, which is only known at the nodes x 1, x,, x n, ie, all we know about the function f are its values y j = f(x j ), j

More information

User Guide for Hermir version 0.9: Toolbox for Synthesis of Multivariate Stationary Gaussian and non-gaussian Series

User Guide for Hermir version 0.9: Toolbox for Synthesis of Multivariate Stationary Gaussian and non-gaussian Series User Guide for Hermir version 0.9: Toolbox for Synthesis of Multivariate Stationary Gaussian and non-gaussian Series Hannes Helgason, Vladas Pipiras, and Patrice Abry June 2, 2011 Contents 1 Organization

More information

SF2943: TIME SERIES ANALYSIS COMMENTS ON SPECTRAL DENSITIES

SF2943: TIME SERIES ANALYSIS COMMENTS ON SPECTRAL DENSITIES SF2943: TIME SERIES ANALYSIS COMMENTS ON SPECTRAL DENSITIES This document is meant as a complement to Chapter 4 in the textbook, the aim being to get a basic understanding of spectral densities through

More information

BME 50500: Image and Signal Processing in Biomedicine. Lecture 5: Correlation and Power-Spectrum CCNY

BME 50500: Image and Signal Processing in Biomedicine. Lecture 5: Correlation and Power-Spectrum CCNY 1 BME 50500: Image and Signal Processing in Biomedicine Lecture 5: Correlation and Power-Spectrum Lucas C. Parra Biomedical Engineering Department CCNY http://bme.ccny.cuny.edu/faculty/parra/teaching/signal-and-image/

More information

Centre for Mathematical Sciences HT 2017 Mathematical Statistics. Study chapters 6.1, 6.2 and in the course book.

Centre for Mathematical Sciences HT 2017 Mathematical Statistics. Study chapters 6.1, 6.2 and in the course book. Lund University Stationary stochastic processes Centre for Mathematical Sciences HT 2017 Mathematical Statistics Computer exercise 2 in Stationary stochastic processes, HT 17. The purpose with this computer

More information

MATLAB and Mathematical Introduction Will be discussed in class on 1/24/11

MATLAB and Mathematical Introduction Will be discussed in class on 1/24/11 MATLAB and Mathematical Introduction Will be discussed in class on 1/24/11 GEOP 523; Theoretical Seismology January 17, 2011 Much of our work in this class will be done using MATLAB. The goal of this exercise

More information

Problem Set 2 Solution Sketches Time Series Analysis Spring 2010

Problem Set 2 Solution Sketches Time Series Analysis Spring 2010 Problem Set 2 Solution Sketches Time Series Analysis Spring 2010 Forecasting 1. Let X and Y be two random variables such that E(X 2 ) < and E(Y 2 )

More information

Fourier Analysis of Stationary and Non-Stationary Time Series

Fourier Analysis of Stationary and Non-Stationary Time Series Fourier Analysis of Stationary and Non-Stationary Time Series September 6, 2012 A time series is a stochastic process indexed at discrete points in time i.e X t for t = 0, 1, 2, 3,... The mean is defined

More information

Engineering Geodesy I. Exercise 3: Monitoring of a Bridge Determination of the Eigenfrequencies

Engineering Geodesy I. Exercise 3: Monitoring of a Bridge Determination of the Eigenfrequencies Engineering Geodesy I Exercise 3: Monitoring of a Bridge Determination of the Eigenfrequencies Prof. Dr. H. Ingensand Geodetic Metrology and Engineering Geodesy 29.11.2010 Model Analyses of the Bridge

More information

lightcurve Data Processing program v1.0

lightcurve Data Processing program v1.0 lightcurve Data Processing program v1.0 Build 8/22/2012 Using Lightcurve... 1 Lightcurve Command Line Parameters... 2 Lightcurve Outputs and Files... 3 Lightcurve Examples... 3 Data File Formats... 4 Lightcurve

More information

Lecture 6: Discrete Fourier Transform

Lecture 6: Discrete Fourier Transform Lecture 6: Discrete Fourier Transform In the previous lecture we introduced the discrete Fourier transform as given either by summations or as a matrix vector product The discrete Fourier transform of

More information

TAKEHOME FINAL EXAM e iω e 2iω e iω e 2iω

TAKEHOME FINAL EXAM e iω e 2iω e iω e 2iω ECO 513 Spring 2015 TAKEHOME FINAL EXAM (1) Suppose the univariate stochastic process y is ARMA(2,2) of the following form: y t = 1.6974y t 1.9604y t 2 + ε t 1.6628ε t 1 +.9216ε t 2, (1) where ε is i.i.d.

More information

Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification

Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification Page 1 of 16 Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification The use of the Lomb-Scargle Periodogram (LSP) for the analysis of biological signal rhythms

More information

Digital Signal Processing: Signal Transforms

Digital Signal Processing: Signal Transforms Digital Signal Processing: Signal Transforms Aishy Amer, Mohammed Ghazal January 19, 1 Instructions: 1. This tutorial introduces frequency analysis in Matlab using the Fourier and z transforms.. More Matlab

More information

Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science : Discrete-Time Signal Processing

Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science : Discrete-Time Signal Processing Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science 6.34: Discrete-Time Signal Processing OpenCourseWare 006 ecture 8 Periodogram Reading: Sections 0.6 and 0.7

More information

Communication Engineering Prof. Surendra Prasad Department of Electrical Engineering Indian Institute of Technology, Delhi

Communication Engineering Prof. Surendra Prasad Department of Electrical Engineering Indian Institute of Technology, Delhi Communication Engineering Prof. Surendra Prasad Department of Electrical Engineering Indian Institute of Technology, Delhi Lecture - 3 Brief Review of Signals and Systems My subject for today s discussion

More information

(Refer Slide Time: 01:30)

(Refer Slide Time: 01:30) Networks and Systems Prof V.G K.Murti Department of Electrical Engineering Indian Institute of Technology, Madras Lecture - 11 Fourier Series (5) Continuing our discussion of Fourier series today, we will

More information

Lecture 7 Random Signal Analysis

Lecture 7 Random Signal Analysis Lecture 7 Random Signal Analysis 7. Introduction to Probability 7. Amplitude Distributions 7.3 Uniform, Gaussian, and Other Distributions 7.4 Power and Power Density Spectra 7.5 Properties of the Power

More information

8.2 Harmonic Regression and the Periodogram

8.2 Harmonic Regression and the Periodogram Chapter 8 Spectral Methods 8.1 Introduction Spectral methods are based on thining of a time series as a superposition of sinusoidal fluctuations of various frequencies the analogue for a random process

More information

Image Filtering, Edges and Image Representation

Image Filtering, Edges and Image Representation Image Filtering, Edges and Image Representation Capturing what s important Req reading: Chapter 7, 9 F&P Adelson, Simoncelli and Freeman (handout online) Opt reading: Horn 7 & 8 FP 8 February 19, 8 A nice

More information

Wavelets and Multiresolution Processing

Wavelets and Multiresolution Processing Wavelets and Multiresolution Processing Wavelets Fourier transform has it basis functions in sinusoids Wavelets based on small waves of varying frequency and limited duration In addition to frequency,

More information

NCSS Statistical Software. Harmonic Regression. This section provides the technical details of the model that is fit by this procedure.

NCSS Statistical Software. Harmonic Regression. This section provides the technical details of the model that is fit by this procedure. Chapter 460 Introduction This program calculates the harmonic regression of a time series. That is, it fits designated harmonics (sinusoidal terms of different wavelengths) using our nonlinear regression

More information

Laboratory Project 1: Introduction to Random Processes

Laboratory Project 1: Introduction to Random Processes Laboratory Project 1: Introduction to Random Processes Random Processes With Applications (MVE 135) Mats Viberg Department of Signals and Systems Chalmers University of Technology 412 96 Gteborg, Sweden

More information

LQR, Kalman Filter, and LQG. Postgraduate Course, M.Sc. Electrical Engineering Department College of Engineering University of Salahaddin

LQR, Kalman Filter, and LQG. Postgraduate Course, M.Sc. Electrical Engineering Department College of Engineering University of Salahaddin LQR, Kalman Filter, and LQG Postgraduate Course, M.Sc. Electrical Engineering Department College of Engineering University of Salahaddin May 2015 Linear Quadratic Regulator (LQR) Consider a linear system

More information

Transform methods. and its inverse can be used to analyze certain time-dependent PDEs. f(x) sin(sxπ/(n + 1))

Transform methods. and its inverse can be used to analyze certain time-dependent PDEs. f(x) sin(sxπ/(n + 1)) AMSC/CMSC 661 Scientific Computing II Spring 2010 Transforms and Wavelets Dianne P. O Leary c 2005,2010 Some motivations: Transform methods The Fourier transform Fv(ξ) = ˆv(ξ) = v(x)e ix ξ dx, R d and

More information

Lecture 11: Spectral Analysis

Lecture 11: Spectral Analysis Lecture 11: Spectral Analysis Methods For Estimating The Spectrum Walid Sharabati Purdue University Latest Update October 27, 2016 Professor Sharabati (Purdue University) Time Series Analysis October 27,

More information

CS S Lecture 5 January 29, 2019

CS S Lecture 5 January 29, 2019 CS 6363.005.19S Lecture 5 January 29, 2019 Main topics are #divide-and-conquer with #fast_fourier_transforms. Prelude Homework 1 is due Tuesday, February 5th. I hope you ve at least looked at it by now!

More information

Linear Prediction 1 / 41

Linear Prediction 1 / 41 Linear Prediction 1 / 41 A map of speech signal processing Natural signals Models Artificial signals Inference Speech synthesis Hidden Markov Inference Homomorphic processing Dereverberation, Deconvolution

More information

Multimedia & Computer Visualization. Exercise #5. JPEG compression

Multimedia & Computer Visualization. Exercise #5. JPEG compression dr inż. Jacek Jarnicki, dr inż. Marek Woda Institute of Computer Engineering, Control and Robotics Wroclaw University of Technology {jacek.jarnicki, marek.woda}@pwr.wroc.pl Exercise #5 JPEG compression

More information

Today s lecture. The Fourier transform. Sampling, aliasing, interpolation The Fast Fourier Transform (FFT) algorithm

Today s lecture. The Fourier transform. Sampling, aliasing, interpolation The Fast Fourier Transform (FFT) algorithm Today s lecture The Fourier transform What is it? What is it useful for? What are its properties? Sampling, aliasing, interpolation The Fast Fourier Transform (FFT) algorithm Jean Baptiste Joseph Fourier

More information

9. Image filtering in the spatial and frequency domains

9. Image filtering in the spatial and frequency domains Image Processing - Laboratory 9: Image filtering in the spatial and frequency domains 9. Image filtering in the spatial and frequency domains 9.. Introduction In this laboratory the convolution operator

More information

Discrete-Time Gaussian Fourier Transform Pair, and Generating a Random Process with Gaussian PDF and Power Spectrum Mark A.

Discrete-Time Gaussian Fourier Transform Pair, and Generating a Random Process with Gaussian PDF and Power Spectrum Mark A. Discrete-Time Gaussian Fourier Transform Pair, and Generating a Random Process with Gaussian PDF and Power Spectrum Mark A. Richards October 3, 6 Updated April 5, Gaussian Transform Pair in Continuous

More information

E2.5 Signals & Linear Systems. Tutorial Sheet 1 Introduction to Signals & Systems (Lectures 1 & 2)

E2.5 Signals & Linear Systems. Tutorial Sheet 1 Introduction to Signals & Systems (Lectures 1 & 2) E.5 Signals & Linear Systems Tutorial Sheet 1 Introduction to Signals & Systems (Lectures 1 & ) 1. Sketch each of the following continuous-time signals, specify if the signal is periodic/non-periodic,

More information

STAD57 Time Series Analysis. Lecture 23

STAD57 Time Series Analysis. Lecture 23 STAD57 Time Series Analysis Lecture 23 1 Spectral Representation Spectral representation of stationary {X t } is: 12 i2t Xt e du 12 1/2 1/2 for U( ) a stochastic process with independent increments du(ω)=

More information

covariance function, 174 probability structure of; Yule-Walker equations, 174 Moving average process, fluctuations, 5-6, 175 probability structure of

covariance function, 174 probability structure of; Yule-Walker equations, 174 Moving average process, fluctuations, 5-6, 175 probability structure of Index* The Statistical Analysis of Time Series by T. W. Anderson Copyright 1971 John Wiley & Sons, Inc. Aliasing, 387-388 Autoregressive {continued) Amplitude, 4, 94 case of first-order, 174 Associated

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

DAMOCO: MATLAB toolbox for multivariate data analysis, based on coupled oscillators approach Version 1.0

DAMOCO: MATLAB toolbox for multivariate data analysis, based on coupled oscillators approach Version 1.0 DAMOCO: MATLAB toolbox for multivariate data analysis, based on coupled oscillators approach Version 1.0 Björn Kralemann 1, Michael Rosenblum 2, Arkady Pikovsky 2 1 Institut für Pädagogik, Christian-Albrechts-Universität

More information

6.435, System Identification

6.435, System Identification System Identification 6.435 SET 3 Nonparametric Identification Munther A. Dahleh 1 Nonparametric Methods for System ID Time domain methods Impulse response Step response Correlation analysis / time Frequency

More information

0 for all other times. Suppose further that we sample the signal with a sampling frequency f ;. We can write x # n =

0 for all other times. Suppose further that we sample the signal with a sampling frequency f ;. We can write x # n = PSD, Autocorrelation, and Noise in MATLAB Aaron Scher Energy and power of a signal Consider a continuous time deterministic signal x t. We are specifically interested in analyzing the characteristics of

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

TIME SERIES ANALYSIS

TIME SERIES ANALYSIS 2 WE ARE DEALING WITH THE TOUGHEST CASES: TIME SERIES OF UNEQUALLY SPACED AND GAPPED ASTRONOMICAL DATA 3 A PERIODIC SIGNAL Dots: periodic signal with frequency f = 0.123456789 d -1. Dotted line: fit for

More information

Nuclear Physics Fundamental and Application Prof. H. C. Verma Department of Physics Indian Institute of Technology, Kanpur

Nuclear Physics Fundamental and Application Prof. H. C. Verma Department of Physics Indian Institute of Technology, Kanpur Nuclear Physics Fundamental and Application Prof. H. C. Verma Department of Physics Indian Institute of Technology, Kanpur Lecture - 5 Semi empirical Mass Formula So, nuclear radius size we talked and

More information

Distortion Analysis T

Distortion Analysis T EE 435 Lecture 32 Spectral Performance Windowing Spectral Performance of Data Converters - Time Quantization - Amplitude Quantization Quantization Noise . Review from last lecture. Distortion Analysis

More information

Introduction to Signal Analysis Parts I and II

Introduction to Signal Analysis Parts I and II 41614 Dynamics of Machinery 23/03/2005 IFS Introduction to Signal Analysis Parts I and II Contents 1 Topics of the Lecture 11/03/2005 (Part I) 2 2 Fourier Analysis Fourier Series, Integral and Complex

More information

Course Name: Digital Signal Processing Course Code: EE 605A Credit: 3

Course Name: Digital Signal Processing Course Code: EE 605A Credit: 3 Course Name: Digital Signal Processing Course Code: EE 605A Credit: 3 Prerequisites: Sl. No. Subject Description Level of Study 01 Mathematics Fourier Transform, Laplace Transform 1 st Sem, 2 nd Sem 02

More information

Fundamentals of the DFT (fft) Algorithms

Fundamentals of the DFT (fft) Algorithms Fundamentals of the DFT (fft) Algorithms D. Sundararajan November 6, 9 Contents 1 The PM DIF DFT Algorithm 1.1 Half-wave symmetry of periodic waveforms.............. 1. The DFT definition and the half-wave

More information

Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification

Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification Page 1 of 13 Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification The use of the Lomb-Scargle Periodogram (LSP) for the analysis of biological signal rhythms

More information

Frequency estimation by DFT interpolation: A comparison of methods

Frequency estimation by DFT interpolation: A comparison of methods Frequency estimation by DFT interpolation: A comparison of methods Bernd Bischl, Uwe Ligges, Claus Weihs March 5, 009 Abstract This article comments on a frequency estimator which was proposed by [6] and

More information

Designing Information Devices and Systems I Spring 2017 Babak Ayazifar, Vladimir Stojanovic Homework 2

Designing Information Devices and Systems I Spring 2017 Babak Ayazifar, Vladimir Stojanovic Homework 2 EECS 16A Designing Information Devices and Systems I Spring 2017 Babak Ayazifar, Vladimir Stojanovic Homework 2 This homework is due February 6, 2017, at 23:59. Self-grades are due February 9, 2017, at

More information

BAYESIAN DECISION THEORY

BAYESIAN DECISION THEORY Last updated: September 17, 2012 BAYESIAN DECISION THEORY Problems 2 The following problems from the textbook are relevant: 2.1 2.9, 2.11, 2.17 For this week, please at least solve Problem 2.3. We will

More information

Electrical Engineering Written PhD Qualifier Exam Spring 2014

Electrical Engineering Written PhD Qualifier Exam Spring 2014 Electrical Engineering Written PhD Qualifier Exam Spring 2014 Friday, February 7 th 2014 Please do not write your name on this page or any other page you submit with your work. Instead use the student

More information

G13CCF NAG Fortran Library Routine Document

G13CCF NAG Fortran Library Routine Document G13 Time Series Analysis G13CCF NAG Fortran Library Routine Document Note. Before using this routine, please read the Users Note for your implementation to check the interpretation of bold italicised terms

More information

Random signals II. ÚPGM FIT VUT Brno,

Random signals II. ÚPGM FIT VUT Brno, Random signals II. Jan Černocký ÚPGM FIT VUT Brno, cernocky@fit.vutbr.cz 1 Temporal estimate of autocorrelation coefficients for ergodic discrete-time random process. ˆR[k] = 1 N N 1 n=0 x[n]x[n + k],

More information

CS711008Z Algorithm Design and Analysis

CS711008Z Algorithm Design and Analysis CS711008Z Algorithm Design and Analysis Lecture 5 FFT and Divide and Conquer Dongbo Bu Institute of Computing Technology Chinese Academy of Sciences, Beijing, China 1 / 56 Outline DFT: evaluate a polynomial

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

Random Signal Transformations and Quantization

Random Signal Transformations and Quantization York University Department of Electrical Engineering and Computer Science EECS 4214 Lab #3 Random Signal Transformations and Quantization 1 Purpose In this lab, you will be introduced to transformations

More information

Tutorial 2 - Learning about the Discrete Fourier Transform

Tutorial 2 - Learning about the Discrete Fourier Transform Tutorial - Learning about the Discrete Fourier Transform This tutorial will be about the Discrete Fourier Transform basis, or the DFT basis in short. What is a basis? If we google define basis, we get:

More information

!Sketch f(t) over one period. Show that the Fourier Series for f(t) is as given below. What is θ 1?

!Sketch f(t) over one period. Show that the Fourier Series for f(t) is as given below. What is θ 1? Second Year Engineering Mathematics Laboratory Michaelmas Term 998 -M L G Oldfield 30 September, 999 Exercise : Fourier Series & Transforms Revision 4 Answer all parts of Section A and B which are marked

More information

Numerical Linear Algebra

Numerical Linear Algebra Math 45 David Arnold David-Arnold@Eureka.redwoods.cc.ca.us Abstract In this activity you will learn how to solve systems of linear equations using LU decomposition, with both forward and substitution.

More information

Introduction to Time Series Analysis. Lecture 18.

Introduction to Time Series Analysis. Lecture 18. Introduction to Time Series Analysis. Lecture 18. 1. Review: Spectral density, rational spectra, linear filters. 2. Frequency response of linear filters. 3. Spectral estimation 4. Sample autocovariance

More information

Signal Modeling Techniques in Speech Recognition. Hassan A. Kingravi

Signal Modeling Techniques in Speech Recognition. Hassan A. Kingravi Signal Modeling Techniques in Speech Recognition Hassan A. Kingravi Outline Introduction Spectral Shaping Spectral Analysis Parameter Transforms Statistical Modeling Discussion Conclusions 1: Introduction

More information

The roots are found with the following two statements. We have denoted the polynomial as p1, and the roots as roots_ p1.

The roots are found with the following two statements. We have denoted the polynomial as p1, and the roots as roots_ p1. Part II Lesson 10 Numerical Analysis Finding roots of a polynomial In MATLAB, a polynomial is expressed as a row vector of the form [an an 1 a2 a1 a0]. The elements ai of this vector are the coefficients

More information

Practical Spectral Estimation

Practical Spectral Estimation Digital Signal Processing/F.G. Meyer Lecture 4 Copyright 2015 François G. Meyer. All Rights Reserved. Practical Spectral Estimation 1 Introduction The goal of spectral estimation is to estimate how the

More information

Digital Signal Processing Module 1 Analysis of Discrete time Linear Time - Invariant Systems

Digital Signal Processing Module 1 Analysis of Discrete time Linear Time - Invariant Systems Digital Signal Processing Module 1 Analysis of Discrete time Linear Time - Invariant Systems Objective: 1. To understand the representation of Discrete time signals 2. To analyze the causality and stability

More information

Winter 2019 Math 106 Topics in Applied Mathematics. Lecture 9: Markov Chain Monte Carlo

Winter 2019 Math 106 Topics in Applied Mathematics. Lecture 9: Markov Chain Monte Carlo Winter 2019 Math 106 Topics in Applied Mathematics Data-driven Uncertainty Quantification Yoonsang Lee (yoonsang.lee@dartmouth.edu) Lecture 9: Markov Chain Monte Carlo 9.1 Markov Chain A Markov Chain Monte

More information

The Spectrum of Broadway: A SAS

The Spectrum of Broadway: A SAS The Spectrum of Broadway: A SAS PROC SPECTRA Inquiry James D. Ryan and Joseph Earley Emporia State University and Loyola Marymount University Abstract This paper describes how to use the sophisticated

More information

Numerical Linear Algebra

Numerical Linear Algebra Introduction Numerical Linear Algebra Math 45 Linear Algebra David Arnold David-Arnold@Eureka.redwoods.cc.ca.us Abstract In this activity you will learn how to solve systems of linear equations using LU

More information

Notes on Random Processes

Notes on Random Processes otes on Random Processes Brian Borchers and Rick Aster October 27, 2008 A Brief Review of Probability In this section of the course, we will work with random variables which are denoted by capital letters,

More information

Temperature measurement

Temperature measurement Luleå University of Technology Johan Carlson Last revision: July 22, 2009 Measurement Technology and Uncertainty Analysis - E7021E Lab 3 Temperature measurement Introduction In this lab you are given a

More information

SHOCK AND VIBRATION RESPONSE SPECTRA COURSE Unit 6B. Notes on the Fourier Transform Magnitude

SHOCK AND VIBRATION RESPONSE SPECTRA COURSE Unit 6B. Notes on the Fourier Transform Magnitude SHOCK AND VIBRATION RESPONSE SPECTRA COURSE Unit 6B. Notes on the Fourier Transform Magnitude By Tom Irvine Introduction Fourier transforms, which were introduced in Unit 6A, have a number of potential

More information

Transforms and Orthogonal Bases

Transforms and Orthogonal Bases Orthogonal Bases Transforms and Orthogonal Bases We now turn back to linear algebra to understand transforms, which map signals between different domains Recall that signals can be interpreted as vectors

More information

Assignment #10: Diagonalization of Symmetric Matrices, Quadratic Forms, Optimization, Singular Value Decomposition. Name:

Assignment #10: Diagonalization of Symmetric Matrices, Quadratic Forms, Optimization, Singular Value Decomposition. Name: Assignment #10: Diagonalization of Symmetric Matrices, Quadratic Forms, Optimization, Singular Value Decomposition Due date: Friday, May 4, 2018 (1:35pm) Name: Section Number Assignment #10: Diagonalization

More information

4 damped (modified) Newton methods

4 damped (modified) Newton methods 4 damped (modified) Newton methods 4.1 damped Newton method Exercise 4.1 Determine with the damped Newton method the unique real zero x of the real valued function of one variable f(x) = x 3 +x 2 using

More information

3. Lecture. Fourier Transformation Sampling

3. Lecture. Fourier Transformation Sampling 3. Lecture Fourier Transformation Sampling Some slides taken from Digital Image Processing: An Algorithmic Introduction using Java, Wilhelm Burger and Mark James Burge Separability ² The 2D DFT can be

More information

From Stationary Methods to Krylov Subspaces

From Stationary Methods to Krylov Subspaces Week 6: Wednesday, Mar 7 From Stationary Methods to Krylov Subspaces Last time, we discussed stationary methods for the iterative solution of linear systems of equations, which can generally be written

More information

Properties of a Taylor Polynomial

Properties of a Taylor Polynomial 3.4.4: Still Better Approximations: Taylor Polynomials Properties of a Taylor Polynomial Constant: f (x) f (a) Linear: f (x) f (a) + f (a)(x a) Quadratic: f (x) f (a) + f (a)(x a) + 1 2 f (a)(x a) 2 3.4.4:

More information

IMPROVEMENTS IN MODAL PARAMETER EXTRACTION THROUGH POST-PROCESSING FREQUENCY RESPONSE FUNCTION ESTIMATES

IMPROVEMENTS IN MODAL PARAMETER EXTRACTION THROUGH POST-PROCESSING FREQUENCY RESPONSE FUNCTION ESTIMATES IMPROVEMENTS IN MODAL PARAMETER EXTRACTION THROUGH POST-PROCESSING FREQUENCY RESPONSE FUNCTION ESTIMATES Bere M. Gur Prof. Christopher Niezreci Prof. Peter Avitabile Structural Dynamics and Acoustic Systems

More information

CHAPTER 4. Interpolation

CHAPTER 4. Interpolation CHAPTER 4 Interpolation 4.1. Introduction We will cover sections 4.1 through 4.12 in the book. Read section 4.1 in the book on your own. The basic problem of one-dimensional interpolation is this: Given

More information

LABORATORY 1 DISCRETE-TIME SIGNALS

LABORATORY 1 DISCRETE-TIME SIGNALS LABORATORY DISCRETE-TIME SIGNALS.. Introduction A discrete-time signal is represented as a sequence of numbers, called samples. A sample value of a typical discrete-time signal or sequence is denoted as:

More information

Computer Intensive Methods in Mathematical Statistics

Computer Intensive Methods in Mathematical Statistics Computer Intensive Methods in Mathematical Statistics Department of mathematics johawes@kth.se Lecture 5 Sequential Monte Carlo methods I 31 March 2017 Computer Intensive Methods (1) Plan of today s lecture

More information

STA 4273H: Statistical Machine Learning

STA 4273H: Statistical Machine Learning STA 4273H: Statistical Machine Learning Russ Salakhutdinov Department of Statistics! rsalakhu@utstat.toronto.edu! http://www.utstat.utoronto.ca/~rsalakhu/ Sidney Smith Hall, Room 6002 Lecture 11 Project

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 14:305:45 CBC C222 Lecture 8 Frequency Analysis 14/02/18 http://www.ee.unlv.edu/~b1morris/ee482/

More information

Raman Spectroscopy of Liquids

Raman Spectroscopy of Liquids Chemistry 357 Spring 2013 Raman Spectroscopy of Liquids Lab TA: Paul Dent pwdent@syr.edu PURPOSE: You will investigate Raman light scattering of several different molecular liquids. You will also determine

More information

UCSD SIO221c: (Gille) 1

UCSD SIO221c: (Gille) 1 UCSD SIO221c: (Gille) 1 Edge effects in spectra: detrending, windowing and pre-whitening One of the big challenges in computing spectra is to figure out how to deal with edge effects. Edge effects matter

More information

SECTION 7: CURVE FITTING. MAE 4020/5020 Numerical Methods with MATLAB

SECTION 7: CURVE FITTING. MAE 4020/5020 Numerical Methods with MATLAB SECTION 7: CURVE FITTING MAE 4020/5020 Numerical Methods with MATLAB 2 Introduction Curve Fitting 3 Often have data,, that is a function of some independent variable,, but the underlying relationship is

More information

Semi-analytical Fourier transform and its application to physical-optics modelling

Semi-analytical Fourier transform and its application to physical-optics modelling SPIE Optical System Design 2018, Paper 10694-24 Semi-analytical Fourier transform and its application to physical-optics modelling Zongzhao Wang 1,2, Site Zhang 1,3, Olga Baladron-Zorito 1,2 and Frank

More information

Multiplying huge integers using Fourier transforms

Multiplying huge integers using Fourier transforms Fourier transforms October 25, 2007 820348901038490238478324 1739423249728934932894??? integers occurs in many fields of Computational Science: Cryptography Number theory... Traditional approaches to

More information