摘要
针对隐马尔可夫模型(HMM)的传统参数估计方法容易陷入局部最优,对初始参数值要求较高且会出现过耦合现象,为了提高模型的鲁棒性和识别性能,提出一种基于改进蚁群算法的HMM参数训练估算法(HMM-ACO)。该算法根据信息素的变化实现全局搜索,较好地解决了迭代算法易发生的局部陷阱问题。与其他全局优化算法相比,该算法识别精度有较大提高。实验表明,利用HMM-ACO算法训练的隐马尔可夫模型具有较好的分类识别性能。
Hidden Markov models (HMMs) have been widely used in the area of speech and handwriting recognition owing to its excellent modeling power. The conventional method for parameter estimation of HMMs uses the Baum-Walch (BW) algorithm. However, the BW algorithm is highly sensitive to initial values of the model parameters. We propose a new model selection criterion using ACO algorithm for estimating the parameters of HMMs. The improved ACO algorithm provides a new model of artificial ants which are characterized by a relatively simple but efficient strategy of pray search. The experimental results show that ACO-BW obtains better values for the higher recognition accuracy than that of the HMMs trained by other existing methods.
出处
《江南大学学报(自然科学版)》
CAS
2009年第6期707-710,共4页
Joural of Jiangnan University (Natural Science Edition)
关键词
隐马尔可夫模型
参数估计
改进蚁群算法
连续优化
信息素
hidden markov model, parameter estimation,improved ant colony optimization, continuous optimization, pheromone