论文部分内容阅读
提出了将最小差别信息(MDI)和进化计算(EC)相结合引入到HMM的训练中去的方法.各个模型用个体来表示,个体的适应值采用模型的最小差别信息.这样借助于进化计算全局搜索的特点,能克服传统的MDI局部搜索的不足,从而得到系统的全局最优解.实验结果表明,该方法训练所得的系统识别率高于传统的MDI方法训练所得的系统.
A method of introducing MDI and EC into the training of HMM is proposed, in which each model is represented by an individual, and the fitness of the individual is modeled by the minimum difference information, so that by means of evolutionary computation The global search can overcome the shortcomings of traditional MDI local search and get the global optimal solution.The experimental results show that the system recognition rate obtained by this method training is higher than the traditional MDI training system.