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