Using SVM for cllasification

Hello Matlab Geeks, I am new to this area and my focus is on classification algorithms using Matlab. The application that I am looking for is classifying the propulsion vs push activities of a wheelchair using an accelerometer-based dataset. I had few questions on how to work about it. Any help appreciated :) 1) My dataset is for different days with approximately 2 hrs of data sampled at 30 Hz. So how do I use the whole dataset for classification? 2) And there are few features based on which I want to classify it into the above-mentioned categories. Any leads on how to extract features for the whole dataset? Features I am looking for are: mean, SD, RMS, mean Absolute deviation, zero crossings and mean crossings. 3) How do I compare different algorithms to find the accuracy of each one for eg: SVM, Naive C4.5 etc., 4) Given I have only data for testing what probably can be used as training dataset?

Thanks!