Why is the filtered signal mirroring the raw one ?

I am trying to filter EMG signal.. I have applied high-pass, then low-pass, then notch filter for powerline interferences.

My filtered signal looks like it mirrors the raw signal.. there are random peaks in the filtered signal that are a mirror of the raw one

How can I solve this?

Here is my filter function:

cutoff_freq_highpass = 30; % Cutoff frequency for high-pass filter in Hz

highpass_order = 5; % High-pass filter order

% Design the high-pass filter

[b_highpass, a_highpass] = butter(highpass_order, cutoff_freq_highpass/(fs/2), 'high');

% Apply the high-pass filter

HP = filter(b_highpass, a_highpass, y);

% Define the low-pass filter parameters for noise removal

cutoff_freq_lowpass = 450; % Cutoff frequency for low-pass filter in Hz

lowpass_order = 4; % Low-pass filter order

% Design the low-pass filter

[b_lowpass, a_lowpass] = butter(lowpass_order, cutoff_freq_lowpass/(fs/2) , 'low');

% Apply the low-pass filter

LP = filter(b_lowpass, a_lowpass, HP);

% Define the notch filter parameters

powerline_freq = 50; % Powerline frequency in Hz

notch_order = 4; % Notch filter order

% Design the notch filter

[b_notch, a_notch] = butter(notch_order, [(powerline_freq-1)/(fs/2), (powerline_freq+1)/(fs/2)], 'stop');

% Apply the notch filter to remove powerline interference

NCH = filter(b_notch, a_notch, LP);

% Apply the low-pass filter to remove noise

e = filter(b_notch, a_notch, NCH);

Answer 1

[Mathieu NOE]

hello

I think what you interpret as "mirrored" peaks is just pure luck, for the good reason that there are no scientific reasons why filtering a signal (except doing a simple polarity inversion) could creat the exact mirrored version of the raw signal

but still depending of what filters are used ,you could have a phase shift of + or - 180° in a given frequency range, so if that matches your signal frequency, yes , the output would look mirrored ,

but in that case also your filter has no amplitude impact (gain = 1) so it's not really a filter

beside all that bla bla , I would recommend you use filtfilt instead of filter to have zero phase distorsion , but notice that filtfilt will apply your filtering twice (forward and backward in time) , so either reduce the filter order and change the cut off frequency to keep things equal