How to create a contour plot with frequencies on the y axis and time on the x axis but with the time in hh:mm:ss format?

Hello, I am trying to create a contour plot with frequencies on the y axis and time on the x axis but I want the time in hh:mm:ss format. The graph is generated well with the code below (time is in seconds, simple vector from 0 to 500) but I want the x ticks in hh:mm:ss with a specified start time (and that it adjusts when I zoom in). Thank you in advance. Marc

figure ('OuterPosition',[200 300 1000 600]); contour(time,freq_vector,log(freq_matrix),20); xlabel('time [s]'); ylabel('Frequency [Hz]');

[jonas]

So a correction to my now deleted answer. You need to use the old date format and datetick

[Marc]

Hi Jonas, I am not exactly new to Matlab, but am not very familiar with dates formats. It would be very much appreciated if you could write a bit more explicitly what you mean (with an example maybe?). Thanks in advance... again. Marc

Answer 1

[jonas]

Try this,

t is a double (seconds)

t=t/(60*60*24);
contour(t,freq,z)
datetick('x','hh:mm:ss','keepticks')

This is a bit hacky and will not work if your time-vector exceeds 24h, whereafter the xticks will return to 00:00. In other words, the xaxis shows the time of day, not the time elapsed.

Note that surf works with datetime, if that works for your application. I suppose contourf does not work because the xy-values of the contours are stored in the same matrix where datetime and doubles cannot be mixed.

I've also written an ugly but easy-to-use script for plotting contours with datetime because, quite honestly, it should be possible. Of course, some functionality of normal contour plots (such as clabel) are unavailable.

%%Some data
t=1:100;
y=1:100;
z=peaks(100);

%%Get contours
C=contourc(t,y,z);
cnt=[0];nc=[];idc=[];
while (sum(nc)+cnt)<size(C,2)
  cnt=cnt+1;
  idc=[idc sum(nc)+cnt];
  nc=[nc C(2,idc(end))];
end

%%Save levels and remove the corresponding cols
lvls=C(1,idc);
C(:,idc)=NaN;

So now we have a list of levels and a matrix with all contours saved, each contour enclosed by NaNs. We can easily plot the contours as a single line.

%%Plot
plot(seconds(C(2,:)),C(1,:))
xtickformat('hh:mm:ss')

However, we do not utilize the z-data, so all contours share the same colors. There are some solutions, but the most practical one is probably to split the matrix up in cell arrays. In hindsight, I should have stored them like this from the get go. Here's a piece of code written by Stephen Cobeldick:

%%Split matrix into cell array
C=C(:,2:end)';
idx = all(isnan(C),2);
idy = 1+cumsum(idx);
idz = 1:size(C,1);
C = accumarray(idy(~idx),idz(~idx),[],@(r){C(r,:)});

Now we can plot separate lines and access their individual colors in the handle. Let's define the linecolors by scaling their corresponding z-data to a colormap.

%%Plot
hold on
h=cellfun(@(x)plot(seconds(x(:,1)),x(:,2)),C)
xtickformat('hh:mm:ss')
cb=colorbar(gca)
set(gca,'clim',[min(lvls),max(lvls)])

%%Define colors
cmap=parula(300);
colormap(cmap)

%%Normalize and round for indexing
LevelColors = round(normalize(lvls,'range',[1 size(cmap,1)]))

%%Change color of handles
for j=1:numel(lvls)
    h(j).Color = cmap(LevelColors(j),:);
end

The result is a quite nice contour plot with dateformat on the x-axis. The colorbar is fake, in the sense that it is not scaled automatically.