A DOA and TOA joint estimation algorithm based on deep transfer learning

This letter proposes a direction of arrival (DOA) and time of delay (TOA) joint estimation algorithm with deep transfer learning. Recently deep learning technique has been applied to solve the joint estimation problem by using the pretrained network and perform well. But in real applications, different scenarios require to cost much time to obtain different pretrained network. In order to overcome these problems, a transfer scheme for DOA and TOA joint estimation is proposed based on a multi-task network, which uses a shared-private structure to enhance the transferability of the pretrained network in different signal-to-noise ratio (SNR) scenarios. Thus, for different target scenarios, the proposed transferring scheme just uses a few of data from new scenario to fine-tune pretrained network, which can effectively reduce the computation complexity with satisfied estimation accuracy. Simulation results show that the proposed algorithm is superior to other traditional methods in estimation accuracy and efficiency under different SNR testing scenarios.