For a class of non Gaussian stationary processes, we develop the empirical likelihood approach. For this it is known that Whittle likelihood is the most fundamental tool to get a good estimator of unknown parameter, and that the score functions are asymptotically chi-square distributed. Motivated by the Whittle likelihood, we apply the empirical likelihood approach to its derivative. This paper provides a rigorous proof on convergence of our empirical likelihood to a chi-square distribution. Also, some numerical studies on confidence region will be given.