Fingerprint-Based Localization for Massive MIMO-OFDM System With Deep Convolutional Neural Networks

Abstract

Fingerprint technique is a promising enabler for mobile terminals (MTs) localization in rich scattering environments, such as urban areas and indoor corridors. In this paper, we investigate fingerprint based location for massive multiple-input multiple-output (MIMO) orthogonal frequency-division multiplexing (OFDM) systems with deep convolutional neural networks(DCNNs). By taking full advantage of the high resolution in the angle domain and the delay domain in massive MIMO-OFDM systems, we first propose an efficient angle-delay channel amplitude matrix (ADCAM) fingerprint extraction method. Then a DCNN enabled localization method is proposed, in which the modeling error for fingerprint similarity calculation can be overcome. Both DCNN classification and DCNN regression are considered. For practical implementation, a hierarchical DCNN architecture is proposed. The performance of the proposed DCNN localization method is evaluated via simulation performed in a geometry-based ray tracing signal propagation scene. Numerical results demonstrate that DCNN performs well in achieving high localization accuracy as well as reducing storage overhead and computational complexity.