Deep neural network techniques for monaural speech enhancement and separation: state of the art analysis