The estimation of DOA is an important problem in sensor array signal processing and its industrial applications. Among all the solving techniques for DOA, the Stochastic Maximum Likelihood (SML) algorithm is well-known for its high accuracy of DOA estimation. However, its computational complexity is very high because a multi-dimensional nonlinear optimization problem is involved. The Particle Swarm Optimization (PSO) algorithm is considered as a rather efficient way for multi-dimensional non-linear optimization problems in DOA estimation. However Conventional PSO algorithm usually needs a large number of particles and the iteration number is also a litter high when all the particles converge. As a result, the computational complexity is still a litter high. This paper proposes a low complexity Joint-PSO (JPSO) algorithm for SML estimation. It uses the solution of Estimation of Signal Parameters via Rotational Invariance Techniques (ESPRIT) and stochastic Cramer-Rao bound (CRB) to determine a novel initialization space. In this case, smaller number of particles and less iteration number are required. Therefore, the computational complexity can be greatly reduced. Simulation results are also shown to demonstrate the validity of proposed JPSO algorithm.