Heatmap-Based Multiple Co-Channel Transmitter Localization with Fully Convolutional Network

In this paper, a novel method based on deep learning for localizing multiple co-channel transmitters in complex electromagnetic (EM) environment is proposed, which can be applied in scenario with uncertainty about transmitters and propagation model parameters. The proposed method called HMTLNet, formulate the localization of multiple transmitters as a heatmap regression problem based on fully convolutional network. A representative numerical example considering dense urban scenario is discussed to demonstrate that compared to DeepTxFinder, which also leverages deep learning, the proposed method is more simplified, requiring only one network to localize multiple co-channel transmitters, while DeepTxFinder requires at least as many networks as the number of transmitters. And the localization performance is better when the sensor density is high than 0.1%.