A speech recognition method for noise environments using dual inputs