gusucode.com > 信号处理工具箱 - signal源码程序 > signal\signal\signal\Readme.m
% README file for the Signal Processing Toolbox.
% Version 4.2 (R11) 10-Jul-1998
%
%
% List of changes:
%
% NOTE: Items marked with '*' have changed in a way which might affect
% your code.
%
% FIXES
% ~~~~~
% latc2tf
% tf2latc
% - Now convert back and forth correctly.
% pburg
% - Now works correctly in the complex case.
% sos2zp
% - Now handles delays in filters correctly. For example:
% sos=[0 1 0 1 2 0]; [z,p,k]=sos2zp(sos); previously returned:
% z=0, p=[0;-2], k=0; and now returns z=[], p=-2, k=1.
% xcorr
% - Now no longer returns a small imaginary part for the zero lag
% autocorrelation of complex data.
% zp2sos
% - Now handles delays in filters correctly. For example:
% b=[0 2]; a=[1 1/2]; [z,p,k]=tf2zp(b,a); sos=zp2sos(z,p,k);
% previously returned: [2.0000 0 0 1.0000 0.5000 0] and now
% returns: [0 2.0000 0 1.0000 0.5000 0].
%
%
% ENHANCEMENTS
% ~~~~~~~~~~~~
% detrend
% Now ships with MATLAB; in the toolbox/matlab/datafun directory.
% firrcos
% - Now allows you to specify either a bandwidth or a roll-off factor.
% - Now allows the design of either a normal or a square root raised
% cosine filter.
% - Now allows a user setable variable delay of the impulse response.
% - Now accepts a window parameter in the filter design.
% levinson
% - Now is a CMEX function.
% pburg
% pmtm
% pmusic
% pyulear
% * Now, when no Fs is specified, these functions return the PSD estimate,
% Pxx(w), as a function of normalized angular frequency,
% w=2*pi*f/Fs[rads/samp]. If Fs is specified, they return the psd
% estimate as a function of physical frequency, f [Hz]. Fs defaults to
% 1 Hz. These functions now correctly scale the psd by the sampling
% frequency, for linear frequency, or 2*Pi, for normalized, angular
% frequency. In addition they now return the single-sided PSD for real
% signals and the double-sided PSD for complex signals.
%
% NOTE: The old versions of these files (pburg, pmtm, pmusic, and pyulear)
% are available from the MathWorks's ftp site (ftp.mathworks.com)in
% the Technical Support area.
%
% Note that the new functions pcov, pmcov and pwelch also adhere to
% the specifications listed above.
% poly2rc
% - Now also returns the zero lag autocorrelation, when called with an
% optional second input argument, the final prediction error.
% rc2poly
% * Now returns a column vector.
% - Now also returns the final prediction error, when called with an
% optional second input argument, the zero lag autocorrelation.
% sos2ss
% sos2tf
% sos2zp
% - Now allow for a optional second input argument, the gain returned
% by the functions that convert to SOS (ss2sos, tf2sos and zp2sos).
% ss2sos
% zp2sos
% * Now have an extra output argument corresponding to the gain of the
% second-order sections structure. Furthermore, an extra input
% argument can be given to specify the desired scaling of the
% structure. Scaling choices are: infinity-norm, 2-norm and none.
%
% NEW FUNCTIONS
% ~~~~~~~~~~~~~
% ac2poly
% - Autocorrelation sequence to prediction polynomial conversion.
% ac2rc
% - Autocorrelation sequence to reflection coefficients conversion.
% arburg
% - AR parametric modeling via Burg's method.
% arcov
% - AR parametric modeling via covariance method.
% armcov
% - AR parametric modeling via modified covariance method.
% aryule
% - AR parametric modeling via the Yule-Walker method.
% buffer
% - Buffer a signal vector into a matrix of data frames.
% is2rc
% - Inverse sine parameters to reflection coefficients conversion.
% lar2rc
% - Log area ratios to reflection coefficients conversion.
% lsf2poly
% - Line spectral frequencies to prediction polynomial conversion.
% pcov
% - Power Spectrum estimate via Covariance method.
% peig
% - Power Spectrum estimate via the Eigenvector method.
% pmcov
% - Power Spectrum estimate via the Modified Covariance method.
% poly2ac
% - Prediction polynomial to autocorrelation sequence conversion.
% poly2lsf
% - Prediction polynomial to line spectral frequencies conversion.
% pwelch
% - Power Spectrum estimate via the Welch's modified periodogram method.
% rc2ac
% - Reflection coefficients to autocorrelation sequence conversion.
% rc2is
% - Reflection coefficients to inverse sine parameters conversion.
% rc2lar
% - Reflection coefficients to log area ratios conversion.
% rlevinson
% - Reverse Levinson-Durbin Recursion.
% seqperiod
% - Find minimum-length repeating sequence in a vector.
% sgolay
% - Design a Savitzky-Golay smoothing filter.
% sgolaydemo
% - Demonstrates Savitzky-Golay filtering.
% sgolayfilt
% - Filter a signal with a Savitzky-Golay smoothing filter.
% sosfilt
% - Filter a signal using second-order sections (biquad).
% tf2sos
% - Transfer Function to second-order sections conversion.
% wavplay
% - Play sound using Windows audio output device.
% wavrecord
% - Record sound using Windows audio input device.
%
% GRAPHICAL USER INTERFACE
% ~~~~~~~~~~~~~~~~~~~~~~~~
% SPTOOL - Now loads a default session upon starting. Loading a default
% session is an option that is set via SPTool's preferences.
% SIGNAL BROWSER - Printing with preview is now possible.
% FILTER DESIGNER - Added a Pole/Zero Editor as a new filter design method.
% SPECTRUM VIEWER - Added the Covariance Method and the Modified Covariance Method.
% * Removed the Maximum Entropy Method (MEM).
% * Welch's method now calls PWELCH instead of PSD. Since PWELCH
% does not provide a detrending option and it scales the PSD
% magnitude by 1/Fs, the scale by 1/Fs option has been
% removed along with the detrending option.
% * The BURG and YULE AR methods now scale the PSD magnitude by 1/Fs.
% * The option to specify an autocorrelation matrix to the Yule AR
% method has been removed since it's not part of the Yule AR
% method definition.
% - Printing with preview is now possible.
% -----------------------------------------------------------------------------
% README file for the Signal Processing Toolbox.
% Version 4.1 21-Nov-1997
%
% List of changes:
%
% * NOTE: Items marked with * have changed in a way which might affect your
% code!
%
% FIXES
% ~~~~~
% cremez
% * Results structure fixes: RES.H is now the actual frequency
% response on the frequency grid returned by RES.fgrid. RES.fextr
% was wrong and has been corrected. RES.fgrid and RES.fextr are
% now normalized correctly with 1 corresponding to half the
% sampling frequency.
% - The problem where cremez would return a complex filter when it
% was supposed to return a real filter has been fixed.
%
% invfreqz
% invfreqs
% - Now both work for complex filters.
%
% levinson
% lpc
% xcorr
% * Incorrectly returned conjugate for complex cases.
%
%
% ENHANCEMENTS
% ~~~~~~~~~~~~
% hamming
% hanning
% blackman
% - now accept two new parameters to describe window sampling:
% 'symmetric' or 'periodic'; passing in an empty matrix now returns
% an empty matrix; passing in one point returns unity.
%
%
% NEW FUNCTIONS
% ~~~~~~~~~~~~~
% pburg
% - power spectrum estimate using Burg's method.
% pyulear
% - power spectrum estimate using Yule-Walker AR method.
%
% GRAPHICAL USER INTERFACE
% ~~~~~~~~~~~~~~~~~~~~~~~~
% SPTOOL - Support for importing component structures from the
% MATLAB command line was added.
% FILTER DESIGNER - The Filter Designer was completely redesigned. It has
% a better interface and now it's extensible.
% Measurements of the filter design can be viewed as
% the filter is designed. It also allows the overlay
% of spectra.
% FILTER VIEWER - Now supports the viewing (overlaying) of multiple
% filters. Measurement rulers were added.
% SPECTRUM VIEWER - The following new PSD methods were added: Burg, FFT
% and Yule-Walker AR.
%
% -----------------------------------------------------------------------------
% README file for the Signal Processing Toolbox.
% Version 4.0.1 04-Apr-1997
%
% This version contains fixes to bugs and a few enhancements in the GUI and
% functions. The full Readme file for version 4.0 is included below.
%
% List of changes:
%
% * NOTE: Items marked with * have changed in a way which might affect your
% code!
%
% chebwin
% - Supports even length windows.
% - Gives more accurate side-lobe heights especially when R is small
% (< ~20 dB).
% * This improvement will cause your results to change where you use the
% Chebyshev window (especially when R is small).
% cremez
% - Allows for LGRID grid density input to improve exactness of the
% filter design in some cases.
% - Returns a few more results in the RES structure output.
% dpss
% * Always computes Slepian sequences directly, returning more accurate
% (and slightly different) results for large N.
% - Uses MEX-file based algorithm which is much faster than in ver 4.0.
% - Can return any range of the N sequences, not just the first 2*NW.
% impinvar
% - Now works for multiple poles.
% lpc
% * Now calculates the correct gain G based on a biased autocorrelation.
% The gain factor is now 1/sqrt(length(X)) times the previous gain
% factor.
% pmem
% - The default for Fs is changed to 2, to be consistent with other
% spectral estimation routines.
% * This change will affect your plots if you use the second output
% argument to this function without specifying Fs on input.
% prony
% * Now works correctly for complex inputs.
% remez
% - Is now a "function-function", which allows you to write a function
% that defines the desired frequency response. This feature is
% completely backwards compatible but allows greater flexibility in
% designing filters with arbitrary frequency responses. See the
% remez.m M-file for details about how to do this.
% - Now takes an LGRID grid density input to improve exactness of the
% filter design. By increasing this parameter your filter may be more
% exactly equiripple but will take longer to design.
% - A bug in filters which have very short bands in relation to the
% filter length is fixed.
% - Now optionally returns the maximum error, extremal frequencies,
% frequency grid, and other results in a RES structure (like cremez).
% resample
% - For all combinations of signal length, P, Q, and filter length, the
% output length is now exactly ceil(N*P/Q) where N is the input signal
% length. For some short signals and filters the length was too
% small.
% sptool
% - Now works with 0 and 1 length signals and filters.
% - Allows for non-evenly spaced power spectrum data (imported only).
% - Minor appearance / layout improvements to buttons, popupmenus, etc.
% - Now remembers last location of save, export, and import from disk
% operations.
% - Fills in '.mat' when you type in a MAT-file name with no extension
% when importing from disk into SPTool.
% - Limits the number of popupmenu items to 24 in the "Selection" area of
% Signal Browser and Spectrum Viewer.
% - Saves Preferences on disk only at end of SPTool session.
%
% ----------------------------------------------------------------------------
% README file for the Signal Processing Toolbox.
% Version 4.0 15-Nov-1996
%
% The README FILE
% This file contains a list of bug fixes, enhancements, and new features in
% the Signal Processing Toolbox since version 3.0. There is also an important
% section highlighting changes which might affect the behavior of any m-files
% that you have which use the Toolbox.
%
% Use help on any of these files for more information.
%
% BUG FIXES
% butter, cheby1 - Exact zeros and numerator polynomials for analog case.
% buttord, cheb1ord, cheb2ord, ellipord - The minimum filter order was
% incorrectly overestimated for some bandstop filters. This has been
% corrected.
% decimate - Uses a lower order Chebyshev anti-aliasing filter in case
% the default 8th order filter is bogus. Prevents problems when using
% very high decimation factors. See the help for more information.
% impinvar - The filter is now scaled by 1/Fs. This causes the magnitude
% response of the discrete filter to match that of the analog filter.
% rc2poly - Modified to correctly deal with complex inputs.
% remez, firls - The coefficients in the differentiator case are now correct
% so that when applied to a signal the output is the correct sign.
% remez - The maximum number of iterations was increased from 25 to 250 to
% prevent the design of non-equiripple filters. Also seg faults are now
% avoided in the case of a large number of bands and a short filter.
%
% ENHANCEMENTS TO OLD FUNCTIONS
% cceps - New output parameter for keeping track of rotation applied before
% FFT, useful in inversion.
% fftfilt - Support for multiple filters.
% fir1 - Now works for multiple band filters (in addition to low, high,
% band-pass and band-stop filters). New 'noscale' option to prevent
% scaling of response after windowing.
% firls - No matrix inversion when full band is specified. This makes the
% design of these filters much more efficient.
% levinson - Support has been added for complex inputs, and multiple column
% input.
% lpc - Support has been added for complex and multiple column inputs. Also,
% the gain is now output for the AR estimates.
% remezord - Cell array output with 'cell' option for convenience.
% resample - Uses upfirdn and is MUCH faster when q (the decimation factor)
% is larger than one. Also, resample is now vectorized to work on the
% columns of a signal matrix.
% specgram - Works on a set of specified frequencies using either czt or
% upfirdn.
% strips - New scaling parameter allows control of the vertical height of
% the strips.
% psd, csd - Chi-squared confidence intervals have been added.
% xcorr, xcov - Option for computing the correlation at a specified number
% of lags.
%
% NEW FUNCTIONS
% SIGNAL GENERATION
% chirp - Swept-frequency cosine generator.
% gauspuls - Gaussian pulse generator.
% pulstran - Pulse train generator.
% rectpuls - Sampled aperiodic rectangle generator.
% tripuls - Sampled aperiodic triangle generator.
%
% FILTER DESIGN
% cremez - FIR filter design which minimizes the complex Chebyshev error
% to design arbitrary, including non-linear phase and complex, FIR
% filters.
% fircls, fircls1 - Constrained Least-Squares algorithm for minimizing
% LS error subject to maximum ripple constraints.
% firrcos - Raised cosine FIR filter design from frequency domain
% specifications for communications applications.
% kaiserord - Order estimation formula for finding the minimum
% order FIR Kaiser windowed filter to meet a set of frequency
% domain specifications.
% maxflat - Maximally flat IIR and symmetric FIR lowpass filter design.
% Also known as generalized Butterworth filters.
%
% MULTIRATE FILTER BANKS
% upfirdn - MEX-file implementing upsampling, FIR filtering, and
% downsampling using an efficient multirate implementation.
% Algorithm supports multiple signals and/or multiple filters.
%
% LATTICE FILTER SUPPORT
% latc2tf, tf2latc - Conversion of lattice (or lattice/ladder)
% coefficients to and from transfer function form.
% latcfilt - Fast MEX implementation of lattice and lattice/ladder filters.
%
% SPECTRAL ANALYSIS
% pmem - PSD estimate using Maximum Entropy method.
% pmusic - PSD estimate using MUSIC algorithm.
% pmtm - PSD and confidence intervals using Multiple-taper method.
% dpss - Discrete Prolate Spheroidal sequences (Slepian sequences).
% dpsssave, dpssload, dpssdir, dpssclear - DPSS data base for storing
% long sequences.
%
% OTHER
% icceps - inverse Complex Cepstrum.
%
% NEW - GRAPHICAL USER INTERFACE (GUI) TOOLS
%
% SPTOOL - graphical environment for analyzing and manipulating Signals,
% Filters, and Spectra. You manage and keep track of these objects in
% the SPTool figure, and bring up client tools for more detailed
% analysis. The client tools are:
%
% SIGNAL BROWSER - Interactive signal browsing allows display, measurement,
% and analysis of signals.
% FILTER VIEWER - Graphical tool for viewing the magnitude & phase
% response, group delay, zeros & poles, impulse response,
% and step response of a digital filter.
% FILTER DESIGNER - filter design tool for designing lowpass, highpass,
% bandpass and bandstop filters to meet a frequency
% domain attenuation criterion.
% SPECTRUM VIEWER - Graphical analysis of frequency domain data using
% different methods of spectral estimation.
%
% ************************************************************************
% *** WARNING !!! ***
% The following functions have been fixed or enhanced in a way that might
% affect your existing code.
%
% csd - default detrending mode changed to 'none'.
% - confidence intervals have changed.
% cohere - default detrending mode changed to 'none'.
% psd - default detrending mode changed to 'none'
% - confidence intervals have changed.
% tfe - default detrending mode changed to 'none'.
% resample - uses upfirdn for efficiency. The output of this function
% will differ from previous versions in two cases:
% i) Zero-order hold.
% Previously the output was purely causal, now returns the nearest
% sample.
% ii) Input filter with even filter length.
% Sometimes would error out. Now will always work accurately,
% but give slightly different output.
% impinvar - filter now scaled by 1/Fs for correct scaling of magnitude
% response.
% remez, firls - differentiators have changed sign - need to remove
% minus sign from your code where you use this filter.
%
% Use HELP on these files or TYPE them for more information.
% Copyright (c) 1988-98 by The MathWorks, Inc.
% $Revision: 1.18.1.2 $ $Date: 1999/01/22 03:42:30 $