Monkey Accelerometer Project

October 11, 2009

Morgen has wired the accelerometer up to the Arduino with ribbon cable, and we did some preliminary testing of the system this week. Results seem promising. We're working under the assumption that the accelerometer will be mounted on a collar and fixed on the back of the animal's neck. The accelerometer can then measure monkey movement on three axes:

X data is movement towards the head or tail. In anatomical terms: cranial (+) and caudal (-).
Y data is movement to the monkey's left (+) or right (-).
Z data is movement up or down if quadrupedal. In anatomical terms: dorsal (+) and ventral (-).

I thought that an easy position to recognize would be head-down suspension in atelines. I would guess that prolonged periods in this position would suggest that a monkey is foraging, as in the following videos.

In our tests, the graphs clearly showed when the accelerometer moved into this position. With more data, we think we could have picked out bipedal walking, and an untethered data logger would help here. For recognizing quadrupedal behavior, however, we will need a quadruped and a collar.

We also started to talk a bit about sampling rates and data-logging constraints. The program now maps the output from the accelerometer to a value from 1 to 254. Extreme values outside of our range are set to 0 or 255. This method thus requires three bytes for each sample: one each for X, Y, and Z coordinates. The SD datalogger has a maximum capacity of 2 gigabytes. That means:

(2 gigabytes) / (3 bytes) = 715,827,883 logging events

If we are starting and stopping the data-logging to conserve battery power, then we'd have to also record the date and time. Assuming 32 bits to encode the date (Unix time):

(2 gigabytes) / (3 bytes for coordinates + 4 bytes for time) = 306,783,378 logging events

It would take almost ten years to fill up our SD card if we sample once a second. Needless to say, we have plenty of room.

Here are the next steps for this project, as we currently foresee them:

Quadrupedal Walking vs. Rest requires a collar, an untethered data-logger, and a quadruped (Kenny's dog?)
Brachiation requires a collar, an untethered data-logger, a human (Kenny?), and monkey bars
Processing Output instead of Excel requires a bit of coding know-how
Tilt Switch to Conserve Battery requires tilt switch, electronics magic
Fine Tuning the Mapping and Sampling requires better estimation of battery constraints