Metaheuristics algorithms such as Estimation of Distribution Algorithms use probabilistic modeling to generate candidate solutions in optimization problems. The probabilistic presentation and modeling allows the algorithms to climb the hills in the search space. Similarly in this paper, Continuous Gaussian Estimation of Distribution Algorithm (CGEDA) which is kind of multivariate EDAs is proposed for real coded problems. The proposed CGEDA needs no initialization of parameters; mean and standard deviation of solution is extracted from population information during optimization processing adaptively. Gaussian Data distribution and dependent Individuals are two assumptions that are considered in CGEDA. The fitting task model in CGEDA is based on maximum likelihood procedure to estimate parameters of assumed Gaussian distribution for data distribution. The proposed algorithm is evaluated and compared experimentally with Univariate Marginal Distribution Algorithm (UMDA), Particle Swarm Optimization (PSO) and Cellular Probabilistic Optimization Algorithm (CPOA). Experimental results show superior performance of CGEDA V.S. the other algorithms.