Data Segmentation by proportion

Hi. I want to segmentate an ECG signal recorded for 10 h from a database. Every signal have about 9 milion sample with a sample rate of 250 Hz, so about 3000 seconds. I want to segmentate this signal in time domain so I can have 5 beats in every signal. Every beat is 1.2 s so every signal segmentated will be 6 s and about 1500 samples. I need this for feature extraction and a cnn clasification. How can i extract every signal. I first throught i need to find every QRS complex, but in some cases the P wave will be higher in amplitude than R and findpeaks will get that as a peak. I tried getting every signal in an matrice but it exceeds maximum array size and it can be difficult for further processing.

[signal,Fs,tm]=rdsamp('08455'); %recording name have 2 leads
sig = signal(:,1); %we will wok on the first one
s= 9205760;   %samples
fs = 250;     %sample rate
t = floor(s/ts);     %time for the recording
t_i = 0.8;      %initial time for the inteval 2/3 of first 1.2 s
i = floor(t/6);    %nr of signals
for j =1:i        
    t_f = t_i +6;   %final time for the interval
    m(s,j) = sig((t_i*fs):(t_f*fs));   %I wanted to get every signal in a matrice in which every column will be a signal 
    t_i = (t_f*ts +1)/ts;         %initial time for further intervals
end

Answer 1

[Mathieu NOE]

hello

you can use the attached function to detect positive and negative slope threshold crossing points

this will give you the time index of the signal individual blocks

I could try it inside your code if you could share some signal data

% demo code
n=100;
x=(1:n)/n;
y= 0.5 + sin(20*x)+ 0.02*randn(1,n);

threshold = 1; % 

[ind_pos,t0_pos,s0_pos,ind_neg,t0_neg,s0_neg]= crossing_V7(y,x,threshold,'linear'); % positive slope zero crossing points 

plot(x,y,'b',t0_pos,s0_pos,'+r',t0_neg,s0_neg,'+g','linewidth',2,'markersize',12);grid
legend('signal','positive slope crossing points','negative slope crossing points');

[Theodora Chivu]

Thank youfor your answer, but i didn't wanted to plot the line that divide the graph but store the graph divided. I ended up doing a matrix with 30000 signals x 300 samples.

y = zeros();

samp =9205760;

for i=1:a-1

poz=(1.2/c);

for j=1:poz

if i==1

poz1=1;

else poz1=poz*(i-1);

end

for l=1:d

if adn(l,1)>=poz1 && adn(l,1)<=(poz*i)

y(i,1)=1;

end

end

y(i,j+1) = s((poz*(i-1)+j),:);

end

end