中文摘要：

运用带延迟拒绝的可逆跳马尔科夫链蒙特卡洛方法（DRJMCMC）来研究多元混合模型的参数估计和模型选择问题.在混合元的分裂和合并过程中,依然遵照转移前后模型的一阶和二阶矩不变的原则,同时引进随机产生的正交阵解决协方差矩阵不同的问题.还给出DRJMCMC算法在多元正态混合模型中的接受概率的具体表达式.最后给出了一些模拟数据的结果来验证这个算法的可行性及优良性.

英文摘要：

The Bayesian analysis of the multivariate Gaussian mixture model using the delaying rejection reversible jump MCMC algorithm was presented,which could deal with parameter estimation and model selection jointly in a signal sweep.Adhering to the principle of preserving the first two moments before and after the split and combination,special orthogonal matrix was generated to solve the problem of constructing the covariance matrix.The acceptance probability of the DRJMCMC algorithm was given.Experimental results on several data sets demonstrate the efficacy of our algorithm.