There are different human falls and different changes in acceleration curve. It depends on his/her height, weight and posture. But how to model the falling based on these parameters? Has any one done the work or alike ?
Many studies recently involved this problem, in particular with the aim to detect elderly safety in the smart house approach. For this reason, I cannot answer you in a complete way, but I can suggest you some possible search starting points. For example, the FORITAAL 2013 and 2014 congresses deal with similar issue and many works about it were provided (e.g.: also posing attention on preventive detection of possible falls http://dee.poliba.it/DEIS/Projects4.html). Indoor positioning system, accelerators, and also EEG devices can be used in order to understand the human falling; image motion analysis such (e.g.: Kinetic systems) could be proposed in order to analyze the falling behaviours. I would also like to suggest you to pose the attention on some elderly care projects, such as http://fate.upc.edu/index.php.