In this paper, we build a central limit theorem for triangular arrays of sequences which satisfy a mild mixing condition. This result allows us to study asymptotic normality of density kernel estimators for some classes of continuous and discrete time processes.