These are the code and saved models of the paper "A DOA Estimation Neural Network with Hierarchical Codebook for Hybrid Millimeter Wave Systems".
The HC-EST algorithm has been proposed to channel estimation for Hybrid Millimeter Wave Systems. However, it does not perform very well in the multipath scenario. In this letter, we aim to design a neural network with hierarchical codebook (HCNet) for DOA estimation in hybrid mmWave systems. The HCNet adopts the HC-EST algorithm as the feature extractor to pre-process the raw signal at the antenna array.
We provide trained models for all networks. You can directly load the trained model and test it by running ".py" files. You can also train the network by yourself. Due to the space limitation of github, we provide training, validation and test data sets in Baidu SkyDrive (Extraction code: 52zg) or Google Cloud Disk.