I don't know if I understand the problem correctly, but maybe using probabilistic graphical models might help here. For example, you can treat every joint in the body as a node, and 'the bones' would be the links between nodes. This will give you a graph representing the human body. You can then learn the parameters of this graph using a dataset that contains the XYZ info of your sensor, and the corresponding body movement. Finally, after learning the parameters you make inference. Given a series of XYZ coordinates given by your sensor, what is the most likely body movement? Coursera has an excellent course in probabilistic graphical models, and one of the assignments is the identification of movements using data from an xbox kinect.
Another alternative, depending on the size of your dataset, is the use of Recurrent Neural Networks. Again, you will have a sequence of XYZ coordinates (or some features extracted from it) as your input, and the output will be the body movement. Since the RNN have many parameters, then you might need a lot of data to make this approach work.
I suspect that your question is too general. My wrist movement is somewhat independent of my body movement. For example, I may be playing a guitar and so my wrist will be moving in a certain manner. However, while playing the guitar I may be either, sitting down, standing up, walking or dancing to the music that I am playing. From this you can see that my wrist movements do not define my body movements and so one can not be inferred from the other.
Another situation where this "confusion" occurs can be manufactured if you have a chair that rotates. You can then sit in the chair with your wrist still while a friend quickly rotates the chair and (if you don't get dizzy) then if you look at the reading from your wrist sensor you will see that it seems to indicate that the wrist is moving (and it is but so is your whole body) and so you can not assume that a movement registered at the wrist only comes from the wrist.
There are a number of open source algorithms that can be used to filter and interpret the raw 3D accelerometer signals into outcome measures. The GGIR package in R may be useful:
Depending on the body movements you are interested in estimating, you will need to be aware of the limitations of sensor placement location and number of sensors used.