GDNN: a gender-dependent neural network for continuous speech recognition

Most parametric representations of speech are highly speaker dependent, and probability distributions suitable for a certain speaker may not perform as well for other speakers. It is desirable to incorporate constraints on analysis that rely on the same speaker producing all the frames in an utterance. Experiments for speaker consistency modeling by using a classification network to help generate gender-dependent phonetic probabilities for a statistical recognition system are reported. Results show a good classification rate for the gender classification net. Simple use of such a model to augment an existing larger network that estimates phonetic probabilities does not help speech recognition performance. When the net is properly integrated in a hidden Markov model (HMM) recognizer, it significantly improves word accuracy