'double' input argument not found using a function

Hello,

I have got a int16 array (called allAudio) and would like to run the function 'findpeaks' on it. When I write pks = findpeaks(allAudio(1,:)), the error message says: "Error using findpeaks (line 59) Input arguments must be 'double'."

 I don't understand it... Any thoughts?

Answer 1

[Iain]

findpeaks only operates on numbers that are 64 bit floating point numbers (they're called doubles).

Convert your 16bit integers to double with the function "double".

double_allAudio = double(allAudio);

[Joachim]

Great!

here's the last bit of the code I wrote to make it work (Had to transpose the matrix)

allAudio_transp=allAudio.';
double_allAudio = double(allAudio);
[pks,locs] = findpeaks(double_allAudio(1,:))

[kemo]

Hello all, But what if we need to specify a certain datatype for the arguments and give the inputs accordingly??

Answer 2

[Steven Lord]

In release R2017b we introduced functions islocalmax and islocalmin that support a data input A of double, single, or one of the integer types.

rng default
x = randi(10, 1, 20, 'int8');
ILM = islocalmax(x);
y = [x; ILM]

If you run those four commands, you'll see y is an int8 array and that most of the instances of 10 in x are local maxima, as is one of the instances of 9. To see that graphically:

% NaN values don't get plotted so preallocate then fill the local maxima
z = NaN(size(x));
z(ILM) = x(ILM);

ind = 1:numel(x);
plot(ind, x, '-', ind, z, 'ro');

The red circles correspond to the local maxima. The two plateaus each have one of their two points of equal height marked; see the documentation for islocalmax for more information about how you can customize how that behavior is handled.