Thanks for giving it a try!

I'm using a low pass filter on the raw sensor data to minimize any jittery input.

The code I'm using is as follows:

  alpha = 0.1;
  output = previous_output + alpha * (new_input-previous_output);

If anyone knows a better way to get a stable and also current read-out, I would be more than interested!

I haven't implemented one myself, but Kalman filters are frequently used for this. Search HN's history to find several (presumably) good introductions.

Kalman filters can get really complicated (EKF, UKF, adaptive, etc etc) and needs to be tuned with the correct covariances.

Complementary filters works well enough and is very simple to implement even if you're a noob (e.g., high pass on accelerometer and low pass on gyro).

There's a lot of libraries that do the hard work for you though... Anyway, if an EKF is too much effort, you could also try the "Madgwick" filter:

Madgwick, S. O., Harrison, A. J., & Vaidyanathan, R. (2011). Estimation of imu and marg orientation using a gradient descent algorithm. In Proceedings of the IEEE International Conference on Rehabilitation Robotics

The Madgwick filter seems to be popular for gyroscope applications, this might be very helpful. Thanks!

Awesome! I'll check it out

I'm no expert, but when I had to make a cursor on the screen from Kinect hand location data I used what's called an exponential weighted average using output from an exponential moving average function on the raw data.

http://stackoverflow.com/questions/7947352/exponential-movin...

http://blogs.interknowlogy.com/2011/10/27/kinect-joint-smoot...

Worked pretty well!

That sounds very helpful! I'll give it a close read and see what I can come up with.

i think i small windowed median filter would do a good job in this case

I tried a small windowed mean filter, which gave really bad results. I'll try implementing the median filter.

Also be aware, that the accuracy is solely dependent on the accuracy of your device. If your compass is not correctly calibrated or there is any magnetic interference, the software can't do much.