摘要
EM算法用于高斯混合模型参数估计时,具有对初始值敏感、易于陷入局部极小等缺点。将差分进化算法引入高斯混合模型参数估计问题,提出一种基于差分进化算法的高斯混合模型参数估计方法。该方法直接对模型参数进行编码,待优化目标函数简单且物理意义明显,具有算法实现容易、运行效率高及收敛速度快等优点。实验结果表明,新方法具有很强的全局搜索能力,参数估计精度更高、更稳定。
EM Algorithm is sensitive to the initial value of parameters and apt to fall into local minima, while applied to Gaussian Mixture Model for parameter estimation. Proposes a Differential Evolution Algorithm based Gaussian Mixture Model to solve the parameter estimation problems in Gaussian Mixture Model. In the method, the parameters are directly encoded, thus the target function is simpler. And the algorithm is easy to perform and quick to converge. The tests show that, the new method is powerful in performing a global search and it gives more precise and steady parameter estimation.
出处
《现代计算机》
2009年第5期29-31,共3页
Modern Computer
关键词
高斯混合模型
差分进化算法
EM算法
参数估计
Gaussian Mixture Model
Differential Evolution Algorithm
EM Algorithm
Parameter Estimation