I am currently working on the EEG Grasp Dataset from Kaggle. I am aware that I need to filter the EEG signal as it is visually noisy (is there any way to mathematically show it's noisy?). How do I choose what filter I should use from the vast number of types available. Also how do I make sure that the filter improves the SNR while not removing/damaging the information in the EEG signal?
From The kernels in kaggle, I have seen people use filters like discrete wavelet transform to get the frequency components and filter out unnecessary frequencies or they have used butterworth and notch filters, though one said that the notch filter "distorts" the signal and instead a high pass at 40Hz should be used. People have used different ones so, is there any procedure to determine which filter to use or should I start testing from the most basic filters then get into better frequency domain filters, then spatial filters and finally adaptive filters?