Speech fusion based on rough neural network for humanoid intelligent robots
Chen Xing-lin · Kongzhi yu juece · 2003
Integrating rough set theory with neural network theory, a novel information fusion method based on rough neural network is proposed. It is used in speech fusion of humanoid intelligent robots. The input could be qualitative, i.e. a range or variational during the observation, as well as quantitative. Because the error transfer function of rough neural network is not differentiable, genetic algorithms are applied for training the network. The simulation results indicated that the novel information fusion method based on rough neural network does improve the speech recognition probability.