There are many ways to do this. The most direct will be to define the data type of the MATLAB Function block output using the Model Explorer. It will explicitly allow you to set any desired data type on any port of any block.
A less direct method is to use Simulink's type propagation. The answer above suggests a Data Type Conversion block, placed on the output. This won;t quite do it. You would need the receiving block's input to be specified to your desired data type.
This can also be accomplished inside your MATLAB code, by explicitly setting the type of the data assigned to your function output in the assignment. For example, for the final output assignment, instead of writing:
y = my_computed_data;
Write this instead:
y = fi(my_computed_data, 1, 16, 4);
This specifies the output to be a signed 16 bit value with 4 bits to the right of the binary point. If this is the first (or only) assignment of output y, it will define the data type of y. With no explicit or implicit definition the data type defaults to double.
