Call Admission Control is a key function that guarantees the Quality of Service (QoS) for users. In radio networks, this function is usually based on traffic models and ensures that sessions are admitted only if the estimated available bandwidth is enough for the entire call duration. For video on IEEE 802.16, the CAC function must ensure that the bandwidth to be reserved is compatible with the resource availability. For the enhanced SVC (Scalable Video Coding) systems, the CAC function must take into account all the layers and their characteristics. In this paper, we propose an enhanced CAC function for SVC that adapts the admission according to the statistical behaviour of the video sessions. The main goal is to use measurements in the 802.16 base station (BS) to update the traffic model of SVC video flows, this for the different layers of SVC flows. We then use the variability of the traffics generated to adapt the CAC according to the characteristics of incoming flows. To perform that, we use a Markovian model that adapts for each flow instead of using a generic static one as used in most of the papers. Performance evaluation is given to illustrate the interest of our proposal.