Digital speech analysis using sequential estimation techniques