Deep Transfer Learning for Site-Specific Channel Estimation in Low-Resolution mmWave MIMO
Wesin Alves, Ilan Correa, Nuria González‐Prelcic, Aldebaro Klautau
Abstract
We consider the problem of channel estimation in low-resolution multiple-input multiple-output (MIMO) systems operating at millimeter wave (mmWave) and present a deep transfer learning (DTL) approach that exploits previously trained models to speed up site adaptation. The proposed model is composed of a feature extractor and a regressor, with only the regressor requiring training for the new environment. The DTL approach is evaluated using two 3D scenarios where ray-tracing is performed to generate the mmWave MIMO channels used in the simulations. Under the defined testing setup, the proposed DTL approach can reduce the computational cost of the training stage without decreasing the estimation accuracy.