Thread Subject:

EEG SIGNALS CLASSIFICATION

I want to apply LDA for only dimensionality reduction and classify signals after that with SVMTRAIN and SVMCLASSIFY.
I've used PCA + SVMTRAIN + SVMCLASSIFY to classify EEG SIGNALS.
I want to compare between tow methods (PCA + SVM) AND (LDA + SVM).
So I need "Linear Discriminant Analysis (LDA)" code for dimensionality reduction.

On Apr 29, 8:40 am, "MESBAH Amine" <kerami...@gmail.com> wrote:
> I want to apply LDA for only dimensionality reduction and classify signals after that with SVMTRAIN and SVMCLASSIFY.
> I've used PCA + SVMTRAIN + SVMCLASSIFY to classify EEG SIGNALS.
> I want to compare between tow methods (PCA + SVM) AND (LDA + SVM).
> So I need "Linear Discriminant Analysis (LDA)" code for dimensionality reduction.

help classify
doc classify

help ClassificationDiscriminant
doc ClassificationDiscriminant

Hope this helps.

Greg

Thank you Greg,
but, I think that I can't use CLASSIFY for dimensionality reduction :(

On Apr 30, 11:56 am, "MESBAH Amine" <kerami...@gmail.com> wrote:
> Thank you Greg,
> but, I think that I can't use CLASSIFY for dimensionality reduction :(

Correct. Sorry for the Bum Steer.

Re using ClassifierDiscriminant:

I no longer use the linear and quadratic discriminant functions.They
are designed to minimize the ratio of sum(within-class squared
Mahalanobis distance) to sum(between-class squared Mahalanobis
distance), not
classification error rate. As a result, time is spent estimating
covariance matrices, even when there is not enough data to obtain
precise estimates. In addition, dimensionality reduction results in
eliminating variables in a transformed space, not original variables.

My alternative is to use STEPWISE or STEPWISEFIT to directly
eliminate variables, squared variables and/or crossproducts. The
classifier is designed to minimise MSE between target and predicted
class posterior probabilities. Class assignments are made to the class
corresponding to
of the maximum predicted posterior.

Hope this helps.

Greg

assalamualaikum,
sir,
    At first sorry for posting this in this......My name is MOHAMMED JASEEM.K.M. I am doing my final year B.E. BIOMEDICAL ENGINEERING at COLLEGE OF ENGINEERING GUINDY, ANNA UNIVERSITY, CHENNAI, INDIA........
I am doing my final year project on decoding viewed alphabets from subjects EEG signals and i need your help to specify the methode of classification and if possible please mail any MATLAB coding to read the obtained EEG data in MS excel file and analysing it for classification of the two alphabets namely 'Y' AND 'N'......... kindly waiting for your response !!!!!
emailid: jaseemlulu@gmail.com