摘要
对水下目标辐射噪声的混合高斯模型进行参数估计时,针对极大似然函数很难求解的问题,研究了一种使用海洋环境噪声统计信息的期望值最大算法(EM算法)。在对Bouvet和Schwartz水下目标辐射噪声信号和海洋环境噪声模型研究的基础上,修正了传统的EM统计算法,以降低计算的复杂度,提高迭代收敛速度。仿真结果和实船信号实验数据的一致性,表明基于EM算法的混合模型参数估计方法,使得参数估计复杂度降低,运算量明显减少,估计性能较好,具有很好的应用价值。
An iterative method was proposed to estimate the parameters of Gaussian-Gaussian mixture model which is very useful for passive sonars in shallow waters. Refs. 1 and 3 modeled the ship-radiated noise as a Gaussian-Ganssian mixture model, a further study of the model was given by analysis of underwater noise samples, and the aim of the proposed method focused on the parameters estimation of the model. Then a modified EM algorithm was proposed to speed up the convergence of the EM sequence. Experimental results on real-world data and simulation data show that the proposed EM algorithm is an efficient method of parameter estimation of the Gaussian-Gaussian mixture model, which does not affect EM algorithm＇s stability, flexibility and simplicity.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第10期2836-2839,共4页
Journal of System Simulation
基金
国防预研基金(51405020304HK0354)
关键词
混合高斯分布
期望值最大算法
参数估计
海洋环境噪声
Gaussian- Gaussian mixture distributions
Expectation-Maximization algorithm
parameter estimation
ambient ocean noise
