I've recently wrote a segmented least squares algorithm for a Java project I'm working on that involves the accelerometers in our smart phones. While nice an neat test data (like a sine wave) does well when I apply my algorithm, when I use data collected from my smart phone, the noise that gets generated will occasionally get me some unexpected/unwanted results depending on what my penalty value is. While I have some ideas as to how to go about fixing this, I was wondering if anyone has any advice that they could give.
Asked
Active
Viewed 564 times
0
-
Do you have any code for us to review to see what is going on? – panoptical Sep 26 '13 at 17:50
-
I do, however it is is a large segment of code. Also my question is about handling noise, not issues with my code. – Magister Ludi Sep 26 '13 at 17:58
-
If you think you have chosen a good model, try to apply some form of noise reduction first [moving average]. Otherwise, choose another model. Can you post some plot illustrating your measurements? – user877329 Aug 08 '16 at 09:02