HDL Coder does not support dynamic matrices at any point. The matrix in question may not be your final output, but some intermediate temporary value inside your complicated expression.
I would try converting your code to a much simpler form with explicitly typed and sized variables for each operation. This may expose or even eliminate the dynamic matrix.
You can convert:
y=sum(sin(t_rad).*exp(-s*t_rad).*0.1)
to this style:
t1 = -s * t_rad;
t2 = exp(t1);
t3 = t2 .* 0.1;
and so on. If you are in the fixed point domain already, you can apply explicit types and sizing at each step.
