To further enhance the system performance of the heterogeneous network selection algorithm, a Markov process based utility optimization method for multi-access selection in heterogeneous wireless networks is presented in this paper. The method adopts the birth and death process to analyze the system performance of heterogeneous networks, fully considers users’ mobility and then obtains network joint utility optimization model. To maximize the joint utility, the proportions of different multi-access selections in heterogeneous wireless network overlapping region should be well organized. The subgradient method was adopted to obtain the best proportions for solving this optimization problem. The simulation results show that the method can maximize system utility, effectively improve throughput and make the radio resource allocation more efficient.