We are recording LFP signals (Using 2 probes ) from freely moving rats. This induces a large volume of motion-related noise as the rats move the cable, grooming etc.
I am looking for a way to filter out these noises.
As Niels mentioned above this is a very tricky one. Something I had used with neuronexus 16 chn probes and it actually works (partially) is the following:
If we assume that the motion artifact is more or less common for the nearby recording sites but the neural activity is not, then you can subtract the signal from two nearby probes and the common component ie noise will go away.
You can also try with processing by defining the frequency range of the noise, usually low frequencies dominate, and filter them away or even try to apply pca or ica to decontaminate your signal...
This question is quite interesting to me, as I have the same problem (with one 32-channel probe recording in mice). The motion artifacts interfere with my detection of sharp wave ripples-spectrally the artifacts have more power in the ripple band than real ripples. I can therefore detect sleep ripples, but not wake ripples. In newer recordings, this is not a problem, probably because my surgery technique has improved, but hoping to recover the data from the first two mice. The PCA idea is interesting, I will give that a try. I have attached small screenshots of the motion noise vs. the 'clean' sleep data.