% RGK.m Version 1.0 4 February 1995 % The following information is included in the Matlab function rgk.m : %----------------------------------------------------------------------------% % SIGNAL-DEPENDENT TIME-FREQUENCY ANALYSIS USING A RADIAL GAUSSIAN KERNEL % %----------------------------------------------------------------------------% % % This Matlab function implements the ``Optimal Radially Gaussian Kernel % Time-Frequency Representation.'' For details, please consult either % the paper % % R. G. Baraniuk and D. L. Jones, ``Signal-Dependent Time-Frequency % Analysis Using a Radially Gaussian Kernel,'' Signal Processing, % Vol. 32, No. 3, pp. 263-284, June 1993. % % or the thesis % % R. G. Baraniuk, ``Shear Madness: Signal-Dependent and Metaplectic % Time-Frequency Representations,'' Ph.D. Thesis, Department of % Electrical and Computer Engineering, University of Illinois at % Urbana-Champaign, August 1992. Also Coordinated Science Laboratory % Technical Report No. UILU-ENG-92-2226, 1992. See Chapter 6 and % Appendices B and G. % % Equation numbers in the comments below refer to the paper unless % otherwise noted. We have tried to keep as close as possible to the % notation of these documents. % % % FLOW OF THE ALGORITHM: % % Step 1: Compute the rectangularly sampled ambiguity function (AF) % of the signal % % *** Uses the separate function AMBNB *** % (included in this distribution) % % Step 2: Interpolate the AF to polar coordinates % % Step 3: Solve for the optimal kernel spread vector using the % so-called "step-project" algorithm [Eqs. (40)-(42)] % % Step 4: Compute the optimal kernel in polar coordinates % % Step 5: Interpolate the optimal kernel to rectangular coordinates % % Step 6: Inverse FFT the optimal-kernel x AF product to get the % optimal time-frequency representation % % % QUESTIONS? COMMENTS? Drop us a line: % % Paulo Goncalves gpaulo@rice.edu % Richard Baraniuk richb@rice.edu % http://www.dsp.rice.edu % %----------------------------------------------------------------------------% % Help information available in Matlab: %RGK Optimal radially Gaussian kernel time-frequency representation % % Useage: [tfr,Phi,sigma,its] = rgk(s,alpha) % % Input: - s : column or row vector containing the signal to be % analyzed % - alpha : normalized volume of the optimal kernel % reasonable values: 1 < alpha < 5 % alpha = 1 => optimal kernel has same volume as a % spectrogram kernel % % Output: - tfr : optimal radially Gaussian time-frequency representation % - Phi : optimal radially Gaussian kernel % - sigma : spread function parametrized by the radial angle in the % ambiguity domain % - its : number of iterations of the step-projection algorithm % to converge to a (local) maximum % % Example: Two parallel linear chirps % t = (0:127); % s1 = hamming(128)' .* cos(0.2*t + 0.008*t.^2); % s2 = hamming(128)' .* cos(0.6*t + 0.008*t.^2); % s = s1 + s2; % tfr = rgk(s,2); % contour(tfr); xlabel('time'); ylabel('frequency') % % See also: AMBNB % Copyright information: %----------------------------------------------------------------------------% %File Name: rgk.m %Last Modification Date: 1/26/96 18:30:22 %Current Version: rgk.m 1.2 %File Creation Date: Sun Jan 21 16:36:09 1996 %Author: Paulo Goncalves%Extra Verbiage: Richard Baraniuk % %Copyright: All software, documentation, and related files in this distribution % are Copyright (c) 1996 Rice University % %Permission is granted for use and non-profit distribution providing that this %notice be clearly maintained. The right to distribute any portion for profit %or as part of any commercial product is specifically reserved for the author. % %Change History: % %----------------------------------------------------------------------------%