Real time communication (RTC) over wireless multi-hop networks has two major challenges: strong QoS provision and severe interference in wireless networks. To route the coming RTC flow in a high quality path, the system needs to accurately predict the path quality in advance before the traffic really being accepted. In wireless multi-hop networks, interference is the key factor impacting the network performance. Two types of traffics will interfere with the coming RTC flow, one is neighboring traffic including the traffic cross the same node and adjacent nodes, the other is this RTC traffic itself (we call it self-traffic) along the path. However, traditional measurement based routing schemes, such as ETX [1], can not get an accurate estimation of path quality since they can not reflect the self-traffic interference. Let us think about the scenario which is shown in Fig.1, where each node has an 802.11b radio. There is an existing flow (flow-1) from 6 to 0 with 256kbps, now considering a new RTC flow (flow-2) from 3 to 4. There are two possible paths for this coming RTC flow, path-1 and path-2, path-1 is interfered with the existing flow. In this case, the measurement shows that due to interference of the existing flow, the packet loss rate of the three links of path-1 is 13%, 23%, 14% respectively, and the packet loss rate of all links of path-2 is 0. Thus the ETX of path-1 is 3.61, while ETX of path-2 is 4. Therefore, ETX will select path-1 for new coming RTC flow. However contrary to ETX?s selection, the performance of path-1 is worse than that of path-2: the goodput along path-1 is 318kbps, while the goodput along path-2 is 384kbps. This is just because ETX can not reflect self-traffic interference. Since at the route-selection stage the coming RTC flow is not injected into network, ETX can not measure the possible interference of self-traffic. After flow-2 is injected into path-1, the joint impact of both neighboring traffic and self-traffic cause that performance of path-1 is worse than path-2. This weak-