Speaker-independent speech emotion recognition by fusion of functional and accompanying paralanguage features