Multi-target direction-of-arrival estimation of deep models with frame-level permutation invariant training in marine acoustic environment.

Direction-of-arrival (DoA) estimation is an important part in sonar signal processing, providing a reliable foundation for tasks, such as underwater object detection and tracking. Although the deep learning model has powerful data fitting capabilities, accurately estimating the orientation of multiple targets with a single model remains a challenging task. To address this challenge, we enhance the permutation invariant training (PIT) technique and propose two different types of methods: multi-group classification with PIT (MC-PIT) and multi-group regression with PIT (MR-PIT). These two frame-level PIT schemes utilize a single model for both training and testing in multi-target scenarios. Furthermore, we evaluate the performance of MR-PIT and MC-PIT with different network backbones and demonstrate that the frame-level PIT has excellent portability. Compared with the model trained with the general multi-label strategy, simulation experiments show that our proposed methods have better multi-target DoA estimation performance. Finally, when the array configuration of simulated and recorded data are consistent, the model with frame-level PIT can achieve good performance on recorded data even only trained on simulation data.

Reliable underwater multi-target direction of arrival estimation with optimal transport using deep models.

Multi-target direction of arrival (DoA) estimation is an important and challenging task for sonar signal processing. In this study, we propose a method called learning direction of arrival with optimal transport (LOT) to accurately estimate the DoAs of multiple sources with a single deep model. We model the DoA estimation problem as a multi-label classification task and introduce an optimal transport (OT) loss based on the OT theory to capture the intrinsic continuity within the angular categories. We design a cost matrix for the OT loss in LOT approach to characterize the order and periodicity of the angular grid. The LOT approach encourages reliable predictions closer to the ground truth and suppresses spurious targets. We also propose a lightweight channel mask data augmentation module for deep models that use items related to the covariance matrix as input. The proposed methods can be seamlessly integrated with different model architectures and we indicate the portability with experiments on several typical network backbones. Experiments across various scenarios using different measurements show the effectiveness and robustness of our approaches. Results on SwellEx-96 experimental data demonstrate the practicality in real applications.