Your lengthy question is not very clear. I would guess that you want to predict the outcome of upper limb prosthesis based on the EMG signals, and that outcome is categorical (nominal, discrete). This sounds like prediction of a categorical outcome with functional predictors - and I'm going to guess there are other predictor variables (e.g., age, pre-treatment measures of morbidity, etc.). Is that about right?

Dear Rabya,

As you mentioned in your question, the performance of the classifier depends on the quality of extracted features. Please find the attached article. This article compares and contrasts the performance of 4 different classifiers (LLC,GMM,MLP and SVM) for wrist motion classification.

Best wishes,

Fariba

You can use wavelet packet analysis (WPA) and obtain spectral features. For instance, you may compute subband energies within the range of 20 and 150Hz using WPA.

@Fariba i could not open the article you upload some error occurs while downloading could you pls. email it to meon [email protected]

@Serkan Gunal

I am actually looking forward to apply Wavelet neural network... what u suggest is a sub-part of it ... Thanks

@Dennis Cox

Yes Mr. Cox you are very correct it is indeed "prediction of a categorical outcome with functional predictors"

Hello Rabya,

I sent the article to your email. I hope you will get it.

Best wishes,

Have a look at ADC Chan's website at Carleton University for a emg processing toolbox