基于高斯混合模型的EM算法改进与优化

Improvement and Optimization of EM Algorithm Based on Gaussian Mixture Model

作为聚类分析的一种重要模型,高斯混合模型在模式识别领域得到了广泛的应用。高斯混合模型的参数通常使用EM算法迭代训练获得。然而,传统的EM算法具有稳定性低,容易陷入局部极小值等缺点。针对传统EM算法的不足,改进了相关算法,在迭代过程中引入了自适应模型合并和模型分裂的策略。通过计算各高斯模型的熵,合并权值过低的模型,分裂熵过大的模型。此外,还优化了算法计算过程的相关步骤。相应实验结果表明,与传统EM算法相比,改进后的算法具有更强的适应性和更好的性能。

As an important model in the field of clustering analysis,Gaussian Mixture Model （GMM） has been widely used in pattern recognition.Usually,the iterative EM algorithm is applied to estimate GMM parameters.The traditional EM algorithm has a disadvantage of low stability,and it is easy to fall into the local minimum.In this paper,an improved EM algorithm is proposed to improve the stability of the algorithm by cutting off the low-weight component,and splitting the component with largest information entropy to reinitialize the empty mode.In addition,several steps of the calculation process are optimized.