Feature Extraction for classification

Started by Michael Bezenchuk July 8, 2010
I'm doing a project dealing with classification
I have database of 4 classes each class have number of samles
you can see the database here (matlab files - smps parameter):
I've been asked to identify the features visually and then to estimate the
distribution for each class. also it says that i can assume that the
features vector is Gaussian.

i need help to identify the features and how to find the distribution
function for each class.

I've guess about the features, but i don't know how letter to measure them
if i pass the data through the filter
*Code:* BPfilter = fdesign.bandpass(.40,.47,.53,.60,80,1,80);
BandPassFilt = design(BPfilter, 'equiripple');

i get something similar to speech.

Michael Bezenchuk
M.Sc Student
Electrical and Computer Engineering
Ben Gurion University /