Links to earlier accelerometer studies can be found here:
The early research focus was on human gait studies, but there are papers referenced here that deal with accelerometer use in wildlife/animal tracking, and specifically primate.
Main Issues to Date
1. Data Precision. Studies typically achieve precision by adding sensors and harnesses. This seems like the wrong direction to go in our case, so a few papers have focused on limiting sensors to one (Gait Analyzer based on a Cell Phone[...] by Iso & Yamazaki in the case of human gait studies, and the animal studies referenced above). There is more noise in the single sensor approach, but this can be dealt with mathematically with transforms (wavelet in Iso's case, not sure which transforms in the others yet).
2. Data Quantity. Obviously data sampling rates affect quantity, and it can add up fast, so sampling rates need to be held to a minimum. Typically studies are working with 100Hz, which Tom has pointed out is going to be way too much given our constraints. The transforms act as compression, which could be helpful, but it is unlikely that this could be done on the monkey. One idea would be for the accelerometer to only record data at moments of change, i.e., from swinging to resting, etc., so it would be helpful to know exactly what monkey activities we are trying to map. With the GPS we know their path, so what additional information would be helpfully added with the accelerometer?