Hi all,

I am trying to do a project involved with signal classification.

On the input I have say four signals that can occur
So say, signal = ABCD {these signals are the basic signals that I can observe from my test system}. And I have four outputs such that

Whenever, where-ever I see A on the input, I see A in my output window 1 and zero elsewhere on the other output windows.

So for this I trained my network like this

SigA = A zero zero zero
Target A = A zero zero zero
zero zero zero zero
zero zero zero zero
zero zero zero zero

SigB = zero B zero zero
Target B = zero zero zero zero
zero B zero zero
zero zero zero zero
zero zero zero zero

SigC = zero zero C zero
Target C = zero zero zero zero
zero zero zero zero
zero zero C zero
zero zero zero zero

SigD = zero zero zero D
Target D = zero zero zero zero
zero zero zero zero
zero zero zero zero
zero zero zero D

so these constitue my training set {SigA, TarA, SigB, TarB and so on}
I am using following
Trainlm for training it, with initnw function for my layer weights and bias initialization.

The problem that I am facing is that, On the output of my network during training, if my two signals are similar, then I see some spikes along the way, which I dont want, as that can give me inaccurate results for classification.

Can anyone help me ?