Maximum likelihood smoothing estimation in state-space models: An incomplete-information based approach