Removing bias drift from integration of noisy signal

Hello,

I am working on a Simulink simulation where I need to perform a double integration of a noisy signal. The input signal is generated by exciting the system with a chirp signal that varies in frequency from 0.01 Hz to 10 Hz. To reduce noise before integration, I apply a moving mean filter.

However, despite the filtering, the double-integrated signal exhibits an unwanted upward trend (bias drift), which should not be present. I suspect this may be due to numerical drift and low-frequency components.

What would be the best approach to mitigate this bias? Should I use a different filtering technique, remove the DC component, or apply a different integration method?

I have attached plots showing:

• The signal pre filtering
• The signal after filtering
• The output after single and double integration, highlighting the bias issue

Any advice would be greatly appreciated. Thank you!

Answer 1

[Nithin]

Hi @Antonio,

The double integral is anticipated to show an upward trend since the single integral is not converging to zero, which is expected given that the mean of the filtered signal is not zero.

If your goal is to reduce noise in the signal, it is advised to use a "High-Pass Filter" from the "DSP System Toolbox" to eliminate low-frequency components.

However, if your objective is different, it might be better to consider changing the integration method.

I hope this helps in resolving the issue.