My first attempt at performing inertial dead reckoning was with an adafruit flora, a small OLED, and an arduino friendly IMU (see the pictures below). I put these all into a glove and the tested it while I was playing racquetball. I was able to measure impact well enough and to filter out the force of gravity from the readings, but it was not very advanced and I had to stop working on it because I had to focus on my classes.
Since then, I've had some more experience with IMUs and trying to obtain better results from cheap hardware. I've used the BNO055 IMU from bosch. Which I really love because it gives you orientation data in quaternions, and because it gives you acceleration readings minus the effect of gravity. So no more wasting processing power computing those.
And I've learned a bit more about filtering signals and creating self correcting, closed loop systems. One of the biggest problems with inertial dead reckoning is that trying to obtain position from acceleration readings involves integrating twice. Which translates any original errors in the readings exponentially. So a lot of heavy processing is involved in filtering that error out and finding times to "re-set" the system so that the accumulated errors can be purged.
I hope that a hexiwear might be able to help me learn more about signal processing and maybe create a useful device for analyzing things like the swing of a racquet or the movement of your hands during a rowing stroke.
Comments