You can write down their tones using a variety of methods. An easy way is using numbers: 1 is low tone, 2 middle tone, and 3 high tone. You can then write bababa231, which means that the word bababa would start with middle tone, go up to high tone and then down again to low tone. This can then be easily analysed with any corpus method you like.
Sadly, you will need to codify the captured raw data before you try to put it into a corpus, otherwise you will end up with ambiguous entries in your corpus. You might be able to use context to speed up the codification, but some will still have to be done by hand. If you capture your raw data as voice recordings, you may be able to use voice recognition software to capture and codify the data into text.