Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Overview
Today we learn about
The cocktail party problem -- called also blind
source separation (BSS)
Independent Component Analysis (ICA) for solving
BSS
Other applications of ICA / BSS
At an intuitive & introductory & practical level
A bit like
S 1(t)
S 2(t)
Amplitude Amplitude
Signals
time
marginal
densities
a11
x1
s3
s2
a12
x2
a13
s4
a14
x3
xi(t) = ai1*s1(t) +
ai2*s2(t) +
ai3*s3(t) +
ai4*s4(t)
Here, i=1:4.
x4
In vector-matrix
notation, and dropping
index t, this is
x=A*s
Recovered signals
x2
a1
a2
a1
s1
Original s
x1
Mixed signals
x1 (t ) a11 s1 (t ) a12 s2 (t )
x2 (t ) a21 s1 (t ) a22 s2 (t )
t 1 : T
a1
a2
a1
x1
x1 (t ) a11 s1 (t ) a12 s2 (t )
x2 (t ) a21 s1 (t ) a22 s2 (t )
t 1 : T
Mixed signals
Step1:
Sphering
Step2:
Rotatation
x2
a1
a2
a1
s1
Original s
x1
x1 (t ) a11 s1 (t ) a12 s2 (t )
x2 (t ) a21 s1 (t ) a22 s2 (t )
t 1 : T
Mixed signals
Step1:
Sphering
Step2:
Rotatation
Excluded case
There is one case when rotation
doesnt matter. This case cannot be
solved by basic ICA.
Obj ( W ) G ( W T x t ) ( W T W I )
t 1
Obj
Xg ( W T X)T W 0
W
where g(.) is derivative of G(.),
W is the rotation transform sought
is Lagrange multiplier to enforce that
W is an orthogonal transform i.e. a rotation
Solve by fixed point iterations
The effect of is an orthogonal de-correlation
Output: W, S
The overall transform
then to take X back to S is
(WTV)
There are several g(.)
options, each will work best
in special cases. See
FastICA sw / tut for details.
Image denoising
Original
image
Wiener
filtering
Noisy
image
ICA
filtering
Clustering
In multi-variate data
search for the direction
along of which the
projection of the data is
maximally non-Gaussian =
has the most structure
50
100
150
200
250
300
350
Summing Up
Assumption that the data consists of unknown
components
Individual signals in a mix
topics in a text corpus
basis-galaxies
Related resources
http://www.cis.hut.fi/projects/ica/cocktail/cocktail_en.cgi
Demo and links to further info on ICA.
http://www.cis.hut.fi/projects/ica/fastica/code/dlcode.shtml
ICA software in MatLab.
http://www.cs.helsinki.fi/u/ahyvarin/papers/NN00new.pdf
Comprehensive tutorial paper, slightly more technical.