A note on residual-based empirical likelihood kernel density estimation