Spatial filtering from FFT results

I have a spatial signal that has frequencies that I'd like to filter out, the signal is seen below. I need some advice in removing those valleys and leaving just the top portion. I still need the large valley seen around 700.

My code to filter it has been:

dx=0.2; %mm
Fs=1/dx; %1/mm

L = length(metrology_s1);   % Length of signal
t = (0:L-1)*Fs;       
Y = fft(metrology_s1);

P2 = abs(Y/L);
P1 = P2(1:L/2+1);
P1(2:end-1) = 2*P1(2:end-1);

f = Fs*(0:(L/2))/L;
plot(f,P1) 
title('Single-Sided Amplitude Spectrum of X(t)')
xlabel('f (Hz)')
ylabel('|P1(f)|')

From here I am not applying filters correctly:

% Second attempt
wo = 0.012207/(.2);   % 
bw = wo/.5;
[b,a] = iirnotch(wo,bw);
fil = filtfilt(b,a,metrology_s1);       

Any advice?

Answer 1

[Mathieu NOE]

hello

my suggestion below

look at the yellow trace

y = readmatrix('metrology_s1.csv');

samples = numel(y);

% create x vector
dx = 0.2;
Fs = 1/dx;

x = (0:samples-1)*dx;

% smooth data
ys = smoothdata(y,'rloess',300);

% add min (max negative) point
[y_min,ind] = min(y);
x_min = x(ind);
%create an exponential window centered on the negative peak and use that to
%"blend" the smoothed data with the raw data
blend_ratio = exp(-(x-x_min).^4/1000);
% plot(x,blend_ratio); % debug / test only
y_blended = ys.*(1-blend_ratio)+ y.*(blend_ratio);

plot(x,y,x,ys,x,y_blended)
legend('raw','smoothed','blended');