I want to filter a certain data set based on accelerometer data. As the baseline of the data set is perturbed by movement. I don't want to apply a high pass filter, as I do want to measure physiological perturbation in the baseline. Do you know any methods to do this?
Hi Denise Mannée,
I would suggest to use moving average filter to keep physiological information.
https://www.gaussianwaves.com/2010/11/moving-average-filter-ma-filter-2/
Regards
Hilal
Hi,
Have you tried sensor fusion for calibration?
https://pdfs.semanticscholar.org/73d8/dfae89b42e01046d1a12fa412ff8062d17c4.pdf
Regards
Counter question: What kind of physiologic movement, orientation or acceleration you want to measure? If you can give a bit more information about the data.
Try EMD. This will decompose a signal in elemental components. You can reject the unwanted. The others can be joined again.
The moving average filter should work, if it will not work then apply adaptive filter based on EMD.
It is necessary to know as many parameters as possible of the useful and disturbing signals. When both are spectrally different, a suitable frequency filter can be designed. When their spectra overlap, it is necessary to look for another distinct feature and a corresponding mathematical procedure to suppress interference.
A possible solution is use both a gyroscope and a accelerometer, then apply Kalman filter, or adaptative filter.
Thanks to all for responding. I don't want to filter the acceleration data. I want to use the acceleration to filter artefacts from another dataset, which is measure at the same time. I'll will try to do spectra analysis. If this doesn't work EMD/Kalman filter will be the next step.
As far as I can understand the problem, source separation may be an option.
If I got it right, you have a measurement of some physical quantities, and you are interested in the low frequencies perturbations which may appear for some reason. On another side, you have low frequencies perturbations due to an influence quantity, namely a movement known from an accelerometric measurement. If both perturbations occupy the same frequency band, no standard filter will help you separate them.
Someone else just asked a very similar question. You may find my answer helpful (I hope) ...
https://www.researchgate.net/post/Has_anyone_any_experiences_in_designing_a_simple_filter_for_an_IMU_acceleration_and_gyroscope_sensor
Yes, a Kalman filter should work OK. But in less than a second it will converge and all the extra computations required to updated the covariance matrix and compute the gain vector are unnecessary. So you may as well just go straight for a fixed-gain filter (e.g. alpha-beta or alpha-beta-gamma, for pos & vel or pos, vel & acc, respectively) - much simpler, for the same results (at steady-state). If that doesn't give you what you want, you could increase the order of the filter for wider bandwidth and/or improved high-frequency attenuation. For some ideas, see my paper ...
Conference Paper Fixed-Gain Augmented-State Tracking-Filters
Here is the link of my article ,hope is helpful:
Article SVM Multi-Classification of Induction Machine's bearings def...
What dataset you are trying to filter? Generally speaking, if you want to use one signal as reference signal to filter another signal, then there are a number of methods that can be used such as: cross correlation, adaptive filtering, adaptive noise cancellation and principal component analysis.
see for example adaptive filtering using matlab:
https://it.mathworks.com/help/dsp/ug/enhance-a-signal-using-lms-and-normalized-lms-algorithms.html
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling