Gaussian Mixture Modeling GUI (GMM DEMO)

GUI for an Expectation-Maximization algorithm (EM) variant (Split-EM-Discriminant)

The Expectation-Maximization algorithm (EM) is widely used to find the parameters of a mixture of Gaussian probability density functions (pdfs) or briefly Gaussian components that fits the sample measurement vectors in maximum likelihood sense [1]. In our work, the expectation-maximization (EM) algorithm for Gaussian mixture modeling is improved via three statistical tests:
a) A multivariate normality test,
b) a central tendency (kurtosis) criterion, and
c) a test based on marginal cdf to find a discriminant to split a non-Gaussian component.

-Input Buttons
Button 1: Open Data file (.mat) or (.tif)
Button 2: Draw Gaussian Data with Mouse
Left mbutton = Draw
Right mbutton = Jump a point
Return key = Finish

-Operational Buttons
Button 3: Start GMM modeling
Button 4: Stop GMM modeling

-Output Button
Button 5: Save GMM parameters as a .mat file

Requirements:
The DEMO was writen in Matlab 7.5 and Windows XP.

References:
Dimitrios Ververidis and Constantine Kotropoulos, "Gaussian mixture modeling by exploiting the Mahalanobis distance," IEEE Trans. Signal Processing, vol. 56, issue 7B, pp. 2797-2811, 2008.

T. Anderson. An Introduction to Multivariate Statistical Analysis. J. Wiley & Sons: N.Y., 1984.