Can I design an adaptive filter using neural network pattern recognition tool?

I have data consisting normal and abnormal behavior (in terms of mean movement) and I want to filter out the abnormal behavior. I have 4 parameters for all the data like movement frequency, pulse duration, speed and acceleration. Is it possible to set the limit and filter out all the noise (abnormal behavior) from the raw data? I was trying the neural network pattern recognition tool but I am not sure whether that will work or not. Which method should I use, any suggestion is welcome.

[kedar Paul]

Can this can be done using matlab simulation ? I am less experience in coding.

[Greg Heath]

WOULD HELP IF YOU POSTED EXAMPLES OF NORMAL, ABNORMAL AND REVISED ABNORMAL

GREG