Indeed, SPARSE is not supported, and if you declare it extrinsic, the result will be a MATLAB type (mxArray), which you cannot do much with in an Embedded MATLAB block. More to the point, unless this is expected and one knows what one is doing, it will probably create more problems. If the sparse matrix computations are self-contained and the final output is not sparse, one is probably better off running ones own MATLAB function as an extrinsic function. So, for example if the final output is a double of the same size as the first input, I might write
function y = EmbeddedMatlabFun(x,a,b,c,d)
eml.extrinsic('myfun');
y = zeros(size(x)); % declares the output type for myfun.
y = myfun(x,a,b,c,d);
Where myfun is a .m file somewhere on the path (probably in the same directory)
function y = myfun(x,a,b,c,d)
... code ...
t = melbankm(p,n,fs,fl,fh);
... more code ...
% Here foo takes a sparse input t and returns a
% non-sparse output with the same size as x.
y = foo(x,t);
-- Mike
