Speech analysis for health: Current state-of-the-art and the increasing impact of deep learning