At present, users hope to select the available network access quickly and meet their requirements of application, at any time, any place, without human intervention. Firstly, in this paper a novel approach to adjust the weights of fuzzy comprehensive evaluation using quantum genetic algorithm is adopted. Then, under the guidance of cognitive network, cross-layer aware framework oriented at access network selection was proposed. On this basis, access network was selected intelligently. Lastly, simulation experiments showed that this method could choose reasonable access networks without user intervention, and throughput, delay, packet loss and other performance indicators have significantly improved. So the method is superior to the traditional method of link capacity ways and network capacity means.