Effects of spatial correlation between signal subspaces on indoor localization using subspace matching

We present experimental results about a spatial correlation between signal subspaces on indoor localization using a subspace matching. An eigenvector spanning signal subspace is used in the subspace matching. The eigenvector is based on not received signal strengths (RSS) but direction of arrival (DOA) of incident signals on an antenna array, and is unique to the location of the transmitter terminal. To estimate the location of the transmitter terminal, we use the magnitude of inner product between the eigenvector obtained by observation and that obtained in advance at each reference location in the database. The spatial correlation between signal subspaces influences the localization accuracy. Subspace matching localization requires high spatial correlation between neighborhood signal subspaces. In this paper, we evaluate the spatial correlation between signal subspaces by obtaining the eigenvectors at the fine-grid locations in an actual indoor environment. Experimental results show that, because the high spatial correlation between signal subspaces is quite local, the localization accuracy of subspace matching is high only when the database is constructed such that the distance between the adjacent reference locations is short.