Deep neural network-based speech recognition with combination of speaker-class models | IEEE Conference Publication | IEEE Xplore