Channel assignment is an efficient method to increase the network capacity of multi-radio mesh networks by using orthogonal channels for otherwise interfering transmissions in the network. With the massive increase of commercial and personal WLAN deployments in urban areas, co-located networks may cause a performance degradation if operated on the same wireless channels. Thus, in order to fully exploit the benefits of spectral diversity, channel assignment needs to become aware of external interference. In this paper, we propose EICA, a distributed algorithm for external interference-aware channel assignment. We show how EICA is capable to detect co-located sources of interference with spectrum sensing using commodity IEEE 802.11 radios of the network nodes. Thus, no additional hardware is required. Based on the channel conditions, the algorithm favors the lesser congested channels for the channel assignment. We evaluate EICA in a large-scale testbed with 126 multi-radio network nodes in different scenarios with increasing levels of interference. The testbed study shows that EICA can significantly increase the network capacity in scenarios with high levels of external interference.