Segment data into overlapping windows

Hi!

I have a vector of time series data (1x300), that I would like to segment into 20 segments that are 100 ms in duration and have 90% overlap. I tried the buffer function, but the indices in the first segments start with zeros, which is not desirable. I would like the first epoch to start from the first time point and go from there. Here is a screenshot from the figure in the paper where the authors applied the method (Foster, Neuron, 2015). thank you in advance!

Answer 1

[William Rose]

@Panos Kerezoudis, you did not specify the sampling rate. You said you have time series data (1x300).

I assume 300 is the number of points.

You said you want 20 segments that are 100 ms long with 90% overlap. This means the segments overlap by 90 ms. This means you can fit 21 segments onto 300 ms of data.

Let us assume the sampling rate is 1 kHz.

%% Generate simulated data
fs=1000;                    % sampling rate (Hz
N=300;                      % signal length (points)
t=(0:N-1)/fs;               % time vector (s)
y=cos(2*pi*20*t);           % 20 Hz sinusoidal signal, for demonstration purposes

%% Prepare to segment the data
tseg=0.100;                 % segment duration (s) 
nseg=tseg*fs;               % segment length (points)
noverlap=0.9*nseg;          %overlap (point)
noffset=nseg-noverlap;      % offset (points)
K=floor(1+(N-nseg)/noffset);  % number of segments
fprintf('N=%d, nseg=%d, noffset=%d, number of segments=%d.\n',N,nseg,noffset,K)

N=300, nseg=100, noffset=10, number of segments=21.

yseg=zeros(K,nseg);  % allocate arrays for time and y for each segment
tseg=zeros(K,nseg);
%% Segment the data
for i=1:K
    tseg(i,:)=t(1+(i-1)*noffset:(i-1)*noffset+nseg);
    yseg(i,:)=y(1+(i-1)*noffset:(i-1)*noffset+nseg);
end
%% Plot all segments, with vertical offsets
figure;
for i=1:K
    plot(tseg(i,:),yseg(i,:)+(i-1)/5,'-r')
    hold on    
end
grid on; xlabel('Time (s)')

Looks decent.

Good luck with your research.

[William Rose]

@Panos Kerezoudis, if you require assistance in implementing Foster's method for determining response onset latency (B.L. Foster, Neuron 2015, Figure S4), email me by clicking the envelope icon which appears at the upper right corner of the window that pops up when you click on the "WR" circle, next to my answers.

The intracranial, subdural electrode recordings from lateral and medial (!) parietal cortex in humans, obtained by BL Foster et al., are really unique data. If you collect data anything like that, it is very impressive.

[Panos Kerezoudis]

awesome thank you so much for all your help! Will let you know if I run into any issues.

Panos

[William Rose]

@Panos Kerezoudis, you're welcome, good luck.