Recognition of speaker-dependent continuous speech with KEAL