📄 icalab.html
字号:
MCA, NTC, NMF for both overdetermined and underdetermined (overcomplete) cases.</p>
<p><code><span style='font-size:10.0pt'>MATLAB<sup>®</sup> is a registered
trademark of The MathWorks, Inc.</span></code> </p>
<p>A similar package has been developed for <b>ICALAB for <a
href="http://www.bsp.brain.riken.go.jp/ICALAB/">Image Processing</a></b>. </p>
<p>See also NMFLAB and NTFLAB developed by A. Cichocki and R. Zdunek</p>
<p>The comprehensive reference for these toolboxes is in the following book: </p>
<p class=MsoNormal><a href="http://www.bsp.brain.riken.go.jp/ICAbookPAGE/"><b><u><span
style='font-size:13.5pt'>Adaptive Blind Signal and Image Processing</span></u></b>
<br>
<br>
by <b>Andrzej Cichocki</b> and <b>Shun-ichi Amari</b> <br>
<br>
John Wiley, Chichester, UK, 2003 (corrected and revised edition). </a></p>
<p class=MsoNormal style='margin-bottom:12.0pt;text-align:justify'> </p>
<p class=MsoNormal style='margin-bottom:12.0pt;text-align:justify'>The reference
for <span class=GramE>this</span> toolboxes is as follows: </p>
<p class=MsoNormal><b><span lang=PL style='mso-ansi-language:PL'>A. Cichocki,
S, Amari, K, Siwek, T. Tanaka, Anh Huy Phan, R. Zdunek, </span></b></p>
<p class=MsoNormal>ICALAB – MATLAB Toolbox Ver. 3 for signal
processing</p>
<p class=MsoNormal> </p>
<h2><span style='font-size:36.0pt'>The general concept of ICALAB</span></h2>
<p>The important and unique features of our ICALAB toolboxes are <b>preprocessing</b>,
<b>post-processing</b> tools (see Fig. 1 a) and <st1:place w:st="on"><b>Monte
Carlo</b></st1:place> <b>analysis</b></p>
<p class=MsoNormal align=center style='text-align:center'><span
style='color:blue'><img border=0 width=572 height=157 id="_x0000_i1025"
src="icalab_files/image001.gif" alt="Conceptual model of ICALAB Toolbox"></span></p>
<h6><span class=GramE><b>Fig. 1 a</b> Conceptual model of ICALAB Toolbox.</span></h6>
<p>Actual optional <b>PREPROCESSING</b> tools include: Principal Component
Analysi<b>s </b>(<b>PCA</b>), prewhitening, filtering: High Pass Filtering (<b>HPF</b>),
Low Pass Filtering (<b>LPF</b>), Sub-band filters (Butterworth, Chebyshev, Elliptic)
with adjustable order of filters, frequency sub-bands and the number of
subbands) or <b>user-defined</b> preprocessing functions.</p>
<p><b>POSTPROCESSING</b> tools actually includes: Ranking or ordering the
components according to various criteria, Deflation and Reconstruction
("cleaning") of original raw data by removing undesirable components,
noise or artifacts.</p>
<p><span class=GramE><b>SATISTICAL ANALYSIS</b> and evaluating of performance
and consistency, especially sophisticated Monte-Carlo (MC) Analysis.</span></p>
<p>Moreover, the ICALAB Toolboxes have flexible and extendable structure with
the possibility to extend the toolbox by the users by adding their own
algorithms. </p>
<p>The algorithms can perform not only ICA but also Second Order Statistics
Blind Source Separation (<b>BSS</b>), and in the near future Sparse Component
Analysis (<b>SCA</b>), Nonnegative Matrix Factorization (<b>NMF</b>), Smooth
Component Analysis (<b>SmoCA</b>), Factor Analysis (<b>FA</b>) and any other
possible matrix factorization of the form <b>X=HS+N</b> or <b>Y=WX</b>, where <b>H=W</b><b><sup><span
style='font-family:Arial'>+</span></sup></b> is a mixing matrix or a matrix of
basis vectors. <strong>X</strong> is a matrix of the observed data, <strong>S</strong>
is a matrix of original sources and <strong>N</strong> represents the matrix of
additional noise.</p>
<p>The ICA/BSS algorithms are pure mathematical formulas, powerful, but rather
mechanical procedures: There is not very much left for the user to do after the
machinery has been optimally implemented. The successful and efficient use of
the ICALAB strongly depends on <i>a priori</i> knowledge, common sense and
appropriate use of the preprocessing and postprocessing tools. In other words,
it is preprocessing of data and postprocessing of models where expertise is
truly needed (see the <a href="http://www.bsp.brain.riken.go.jp/ICAbookPAGE/">book</a>).</p>
<p>The assumed linear mixing models must be valid at least approximately and
original sources signals should have specified statistical properties.</p>
<p>The package contains a collection of algorithms for whitening, robust
orthogonalization, <st1:City w:st="on"><st1:place w:st="on">ICA</st1:place></st1:City>,
BSS and BSE. The user can easily compare various algorithms for Blind Source
Separation (BSS) employing the second order statistics (SOS), and <st1:City
w:st="on"><st1:place w:st="on">ICA</st1:place></st1:City> using the higher
order statistics (HOS). This package is hence quite versatile and extendable
for user algorithms. </p>
<p><b>ICALAB</b> can be useful in the following tasks:</p>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>1.<span style='font-size:7.0pt'>
</span>Blind Source Separation (BSS), Sequential Blind Sources Extraction
(BSE),</p>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>2.<span style='font-size:7.0pt'>
</span>Reduction of redundancy (<a href="#book">Book</a>, Chapter 3), </p>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>3.<span style='font-size:7.0pt'>
</span>Decomposition of multi-variable signals into independent components
(Chapters 6-8), </p>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>4.<span style='font-size:7.0pt'>
</span>Spatio-temporal decorrelation of correlated signals (Chapter 4), </p>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>5.<span style='font-size:7.0pt'>
</span>Extraction and removal of undesirable artifacts and interference by
applying deflation (see Chapters 1 and 4), </p>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>6.<span style='font-size:7.0pt'>
</span>Removal of noise or "cleaning" the raw sensor data, </p>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>7.<span style='font-size:7.0pt'>
</span>Extraction of features and patterns, </p>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>8.<span style='font-size:7.0pt'>
</span>Comparison of the performance of various algorithms for Independent
Component Analysis (ICA) and<span class=GramE> Blind</span> Source
Separation (BSS), </p>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>9.<span style='font-size:7.0pt'>
</span><st1:place w:st="on">Monte Carlo</st1:place> analysis</p>
<p>Several benchmarks are included to illustrate the performance of the various
algorithms for a selection of synthetic and real world signals (see <a
href="http://www.bsp.brain.riken.go.jp/ICALAB/ICALABSignalProc/benchmarks/">Benchmarks</a>).
</p>
<p><b>Limitations of version 3</b></p>
<p>The Version 3 of the package is limited to <strong>160</strong> channels.
Higher number of channels may be useful for high density array processing in
biomedical signal processing and in multivariate time series analysis
(available on request). Also the number of integrated algorithms is limited to
relatively fast and efficient algorithms.</p>
<div class=MsoNormal align=center style='text-align:center'>
<hr size=2 width="100%" align=center>
</div>
<p><strong>DISCLAIMER </strong></p>
<p class=style1>NEITHER THE AUTHORS NOR THEIR EMPLOYERS ACCEPT ANY
RESPONSIBILITY OR LIABILITY FOR LOSS OR DAMAGE OCCASIONED TO ANY PERSON OR
PROPERTY THROUGH USING SOFTWARE, MATERIALS, INSTRUCTIONS, METHODS OR IDEAS
CONTAINED HEREIN, OR ACTING OR REFRAINING FROM ACTING AS A RESULT OF SUCH USE.
THE AUTHORS EXPRESSLY DISCLAIM ALL IMPLIED WARRANTIES, INCLUDING
MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE. THERE WILL BE NO DUTY ON
THE AUTHORS TO CORRECT ANY ERRORS OR DEFECTS IN THE SOFTWARE. THIS SOFTWARE AND
THE DOCUMENTATIONS ARE THE PROPERTY OF THE AUTHORS AND SHOULD BE ONLY USED FOR
SCIENTIFIC AND EDUCATIONAL PURPOSES. ALL SOFTWARE IS PROVIDED FREE AND IT IS
NOT SUPPORTED. THE AUTHORS ARE, HOWEVER, HAPPY TO RECEIVE COMMENTS, CRITICISM
AND SUGGESTIONS ADDRESSED TO <a href="mailto:icalab@bsp.brain.riken.go.jp">icalab@bsp.brain.riken.go.jp
</a></p>
<div class=MsoNormal align=center style='text-align:center'>
<hr size=2 width="100%" align=center>
</div>
<h1><a name="Getting_started"></a>User guide</h1>
<h2>Starting ICALAB</h2>
<p>To start <b>ICALAB for Signal Processing</b> type: </p>
<p><span class=GramE><tt><span style='font-size:10.0pt'>icalab</span></tt></span>
</p>
<p><span class=GramE>in</span> the <b>MATLAB</b> command window <b>(Note: this package
runs on MATLAB 5.3 or higher)</b>. </p>
<p><b>ICALAB for Signal Processing</b> was developed under MATLAB version 7.1
and tested under MATLAB versions: 7.0, 7.1 and 7.2. <b>(Note: Previous versions
(i.e. 6.5) may not work properly due to some unsupported graphics functions.)</b>
</p>
<h2>Loading the processing data</h2>
<p>To load new signals or data for further processing: </p>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>1.<span style='font-size:7.0pt'>
</span>Click on the <tt><span style='font-size:10.0pt'>File</span></tt> in the
menu bar. It contains <tt><span style='font-size:10.0pt'>Open</span></tt> item.
Both <tt><span style='font-size:10.0pt'>HELP</span></tt> and <tt><span
style='font-size:10.0pt'>EXIT</span></tt> buttons in the main window will
become active after loading of the data file. </p>
<p class=MsoNormal align=center style='text-align:center'><span
style='font-size:10.0pt;font-family:Tahoma'><img border=0 width=518 height=543
id="_x0000_i1028" src="icalab_files/image002.jpg"></span></p>
<h6 style='margin-left:36.0pt;line-height:18.0pt'><span class=GramE><b>Fig. 1</b>
Initial window after starting the program ICALAB.</span> Please click on the <tt>File
| Open</tt> menu to load your data which can be in MATLAB format (*.mat), ASCI
format (*.txt and *.dat) or Excel formats (*.xls and *.csv).<br>
(Press on image to enlarge).</h6>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>2.<span style='font-size:7.0pt'>
</span>You can load the data (our benchmarks or your own data) stored in the
MATLAB (*.mat files), ASCI (*.txt and *.dat files) or Excel (*.xls, *.csv)
format. </p>
<p style='margin-left:36.0pt'>The signals must be stored in a 2-dimensional
matrix inside the corresponding files: </p>
<p class=MsoNormal style='margin-left:72.0pt;text-align:justify;text-indent:
-72.0pt;line-height:18.0pt'><span class=GramE><span style='font-size:10.0pt;
font-family:"Courier New"'>o</span></span><span style='font-size:7.0pt'>
</span>the number of rows corresponds to the number of sensors (the number of
observations) </p>
<p class=MsoNormal style='margin-left:72.0pt;text-align:justify;text-indent:
-72.0pt;line-height:18.0pt'><span class=GramE><span style='font-size:10.0pt;
font-family:"Courier New"'>o</span></span><span style='font-size:7.0pt'>
</span>the number of columns corresponds to the length of the signals, i.e.,
number of samples.</p>
<p style='margin-left:36.0pt'>The loaded Matlab files can contain one or more
data matrices. The user can choose any one of the matrix data. However, data loaded
in ASCII or Excel format should contain only one data file. There is no limit
to the length of the signals (number of samples), but in the version 3
available on the web, only a maximum of <strong>160 signals</strong> (time
series) can be loaded and processed. </p>
<p style='margin-left:36.0pt'>Usually, the convergence speed of the algorithms
in this package strongly depends on the dimension of signals (i.e., the number
of observations and the number of samples) and the computer resources, e.g.,
available memory or processor speed. In the case of processing of large files,
we recommend that you increase the swap space on the hard disk. </p>
<p class=MsoNormal align=center style='text-align:center'><span
style='font-size:10.0pt;font-family:Tahoma'><img border=0 width=360 height=222
id="_x0000_i1029" src="icalab_files/image003.gif" alt="File load window"></span></p>
<h6 style='margin-left:36.0pt;line-height:18.0pt'><span class=GramE><b>Fig. 2</b>
Window illustrating how to load a benchmark or your data.</span> After
selecting the desired data file, click on the <tt>Open</tt> button.</h6>
<p class=MsoNormal style='margin-left:36.0pt;text-align:justify;text-indent:
-36.0pt;line-height:18.0pt'>3.<span style='font-size:7.0pt'>
</span>You can optionally discard some signals and select an arbitrary time
window using the <i>Select channels</i> window. Click on the <tt><span
style='font-size:10.0pt'>SELECT CHANNELS</span></tt> button. </p>
<p class=MsoNormal align=center style='margin-left:36.0pt;text-align:center;
line-height:18.0pt'><img border=0 width=518 height=543 id="_x0000_i1030"
src="icalab_files/image004.jpg"></p>
<h6 style='margin-left:36.0pt;line-height:18.0pt'><span class=GramE><b>Fig. 3</b>
Window after loading of the data.</span> It is possible to input the sampling
frequency of the data to obtain the axis of the plots in time units or leave it
empty to use normalized units. If <span class=GramE>your</span> MATLAB data
consist of several matrices click on <tt>Select variable</tt> popup menu to
choose desired data. Click optionally on <tt>SELECT CHANNELS</tt> in order to
choose the time window (the number of samples) and desired signals or directly
on <tt>Algorithm</tt> to select one of the algorithms (beginners could start
with <code>AMUSE</code> and <code>SANG</code> algorithms). In order to achieve
a fast convergence it is recommended to process less than 10.000 samples at a
time. </h6>
<p style='margin-left:36.0pt'><a name="select_channels"></a>The <i>Select
channels</i> window will appear. This window allows you to mark the signals (or
data) that you want to use for further processing. You can also choose a specific
time window for the input signals (number of samples) by entering the first and
last sample numbers into the fields marked as <tt><span style='font-size:10.0pt'>start</span></tt>
and <tt><span style='font-size:10.0pt'>end</span></tt> at the bottom of the
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -