Automatic speech recognition using dynamic Bayesian networks