Skip to content

Accelerometer Research: October 11, 2009 Meeting

October 13, 2009

I was surprised to discover that the accelerometer does not simply measure acceleration, but orientation relative to gravity.  I programmed it initially in a “0” state facing up and level, with x and y at 0, and z at 1000.  Extreme values were mapped from -1000 (down) to +1000 (up).

Assuming that the accelerometer is facing up and perpendicular to the spine with:

  • xpos forward
  • ypos left
  • zpos up

we modeled simple head-down feeding:


This is logging data every half-second.  We modeled it again logging reads every 3 seconds, but given that the behavior is highly stereotyped, the difference in data stream visualization is trivial:


Data Storage Limitations

As noted in earlier posts, we were concerned that we would encounter data overflow before we were able to reasonably harvest it, so we then remapped the accelerometer to 255 values, with any extremes being max’d and min’d at 255 and 0 respectively.  For this test, we included the actual values to see what might be lost:


A small amount of the X value was lost, but could be corrected with remapping.  Regardless, depending on the acceleration, some information will be lost unless the entire range is mapped, but then patterns may be overly dampened.  Nonetheless, as this case shows stereotypical behavior is easily identified.

While there is no sense throwing it away, we now believe that our initial concerns regarding storage space were overblown:

1 gigabyte = 10^9 bytes.  If we map the accelerometer range to 255, each coordinate can be mapped with 1 byte, so 3 bytes total.  Including a date and time stamp would consume an additional 4, so 10^9/7=143 million sampling events/gigabyte minimum (actual storage space likely varies): that’s 1,650 days at one sample/second.

Bipedal Walking

We tried a couple of experiments walking with the accelerometer, but were unable to pick out anything that looked like a pattern.  In the first test, Christina held the accelerometer out in front of her at mid-torso height:


Given that the human frame is designed to keep the head steady, we weren’t surprised that we saw no rhythmic motion.  We tried it again with the accelerometer held firmly against her left hip:


These tests were inherently limited as the accelerometer was tethered to the Arduino, and the Arduino to my laptop.  It sat at the center of the PComp lab table, and she walked around two sides of it and back, two times.  3 peaks appear to be indicated where she turned to make her return trip.  With more data, we hypothesize we could pick out bipedal walking, and an untethered data logger would help here.

For quadrupedal behavior, however, we need a quadruped and a collar.

Next Tests

Quadrupedal Walking vs Rest

Brachiation (human on monkey bars)

Use Processing to visualize output instead of Excel

Leave a Comment

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: