Joint Direction of Arrival Estimation and Source Enumeration using Transformer for Sparse Linear Arrays with Robustness to Sensor Malfunctions

Date

2024-8-28

Author

Muslu, Burak Hayati

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

144
views

87
downloads

Cite This

Direction of arrival (DOA) estimation and source enumeration are crucial tasks in modern array processing with a wide range of applications. For applying these tasks, sparse linear arrays have been of great interest in the recent years due to their advantages over other array types. However, these arrays are more vulnerable to sensor malfunctions because of their coarray configuration. Sensor malfunctions in these arrays may cause performance degradation in the mentioned tasks if the failures are not handled. Handling the failures by repairing/replacing the sensors is challenging in practice. Therefore, there arises a need of an estimation method that is robust to sensor malfunctions. A transformer-based DOA estimation and source enumeration method is proposed to handle the sensor number/configuration variations occurring due to sensor malfunctions. Existing data-driven solutions cannot generalize to such variations due their input formulation and structural design. The proposed method exploits coarray-based input formulation and positional encoding and attention blocks for learning spatial arrangement of the sensors along with the nonlinear relationship between the measurements and source direction. Experimental results demonstrate that it achieves error reduction in DOA estimation performed for intact array especially for low SNR levels and snapshot numbers. It shows comparable performance to the state-of-theart methods for high SNR and snapshot regime. Introduction of sensor malfunctions causes less degradation in the performance of the proposed method compared to others which have reasonable processing time. The robustness to sensor malfunctions is complemented by performance improvement in source enumeration particularly for low SNR and snapshot regions.

Subject Keywords

Direction of Arrival Estimation, Source Enumeration, Sparse Array, Sensor Malfunction, Deep Learning

URI

https://hdl.handle.net/11511/111053

Collections

Graduate School of Natural and Applied Sciences, Thesis