机构地区: 华中理工大学电子与信息工程系
出 处: 《华中理工大学学报》 1994年第11期1-4,共4页
摘 要: 对窗口神经网络(BWNN)与鲁棒HMM’s之间的关系作了新的探讨,指出了鲁棒HMM’s是循环的BWNN(RBWNN)的特殊情况。讨论了由RBWNN构造用于噪音环境下语音识别的组合HMM’s和二阶HMM’s的算法,给出了它们相应的网络结构。 A new approach to the,relationdhip between the blocked-window neural network(BWNN)and the robust HMM is made.The robust HMM,especially the second-orderHMM(SHMM) and the composed HMM(CHMM),can be viewed as two special cases ofthe recurrent BWNN(RBWNN)。The retrieving and learning phases of SHMM and CHMMare derived,and the structures of the two models are given.