We focus on channel estimation and indoor positioning of wideband multiuser millimeter wave systems. Leveraging the sparse channel characteristics, we transform the estimation problem to recovery problems of multipath parameters, e.g., angle of arrival, time delay and path fading. We model the training signal as a matricization form of a third-order canonical polyadic tensor, which consists of three factor matrices containing channel information. Inspired by array signal processing methods, we exploit the structural feature of our tensor and develop an uniqueness condition of tensor factorization, leading to a tensor-based channel estimation algorithm. In order to distinguish the multiuser signal components, we further develop a clustering-aided indoor positioning scheme. Simulation results show that the proposed methods apply to different system configurations and can achieve satisfactory estimation and positioning performance.

channel estimation; hybrid beamforming; indoor positioning; mmWave systems; tensor signal processing