how can i build a time series classifier

Hi,

In my problen, I want to train a machine learning algorithm to classify patterns of absolute acceleration. Lets assume that i have many examples of acceleration vectors, from t=1 to t=20. A vector can be A = [ 1 1 1 ... 2 3 4 2 1... 1 1 ] ; or B = [ 1 1 1 ... 2 2 3 2 1 ... 1 1 ] ;

There are 2 different patterns nested in A and in B. I assume that all of my examples have one of these patterns, with added noise, BUT I don't know in which t the pattern starts in each example. I also assume that the t_initial of the patterns can be different in each example.

Obviously, a simple neural network or svm model won't work.

Is there any other way to build a classifier for these examples?

Thanks! Yaniv.

[Greg Heath]

I deleted an innapropriate question by the author posted in an ANSWER box.

Greg