You can directly choose 64bits variables (double) if you want highest precision

http://fr.mathworks.com/help/matlab/matlab_prog/complex-numbers.html

http://fr.mathworks.com/help/matlab/numeric-types.html

Thank you for your kind answer! By default, MATLAB stores all numeric values as double-precision floating point. (We cannot change the default type and precision.) Indeed, I want to improve the precision of the intermediate process, not the final precision of the results. 

In MATLAB, the command "format long" displays the numbers with 14 decimal places. It performs the intermediate calculations also with this much precision.

Another important function is "vpa" (Variable Precision Arithmetic). For example, if you need the value of pi correct up to 10000 digits, then vpa(pi,10000) will display the result.

I hope, this is what you need.