摘要
传统的最大互信息训练中一般采用梯度方法 ,这就使得所得模型往往只是一个局部最优模型 .文中将最大互信息 (MMI)和演化计算 (EC)相结合 ,引入到隐马尔柯夫模型 (HMM)的训练中去 .各模型集用个体表示 ,个体的适应值采用模型的最大互信息 .这样借助于进化计算的全局搜索及种群的特点 ,得到了基于最大互信息估计的 HMM模型的更优解 .实验结果表明 。
Maximum mutual information based training produces models with better discriminant power. With the complexity of the objective function in this training, gradient based methods were adopted in traditional training. And this has led to the local optimality of the models. This paper proposed a hybrid maximum mutual information/evolutionary computation (MMI/EC) architecture to be embedded in the training of HMMs. Each individual in evolutionary computation represents a set of HMMs, while the fitness value of each individual represents the maximum mutual information. Since evolutionary computation is noted for its global search and population based operations, the globally optimal solutions, or the sub optimal solutions can be obtained. The experimental results indicate that the system trained with the proposed method is superior to the one trained with the traditional gradient based training method.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2002年第3期348-350,354,共4页
Journal of Shanghai Jiaotong University
关键词
最大互信息
进化计算
混合结构
语音识别
speech recognition
maximum mutual information(MMI)
evolutionary computation(EC)
