SBL for Multiple Parameterized Dictionaries

This repository contains the python-code used in [1].

Included are (i) the code for the proposed sparse Bayesian learning (SBL)-based algorithm, (ii) the code for the newtonized orthogonal matching pursuit (NOMP) algorithm [2] used for comparison, (iii) additional files such as for the generalized subpattern assignment (OSPA) metric [3].

The repositor contains three demo examples (example_<name>.py). The script example_crossing.py reproduces Figure 2 from [1].

Repository structure:

|- pySBL/pySBL.py     Code for the proposed algorithm
|- pyOMP/pyOMP.py     Implementation of the NOMP method for comparison
|- dictionary_functions.py    Implementation of the (parameterized) dictionary functions
|- gopsa.py           Implementation of the OSPA used to evaluate the results.
|- example_radar.py   Demo example with a single radar comparing SBL and OMP
|- example_multiradar.py      Demo Example applying SBL to multiple radars
\  example_crossing.py        Demo example of two crossing targets (Fig. 2)

The code was tested using python 3.13 and numpy 2.2.3. The full spec of the miniconda environment is found in python_env.txt

References

[1] Moederl J., Westerkam, A. M., Venus, A. and Leitinger, E., "A Block-Sparse Bayesian Learning Algorithm with Dictionary Parameter Estimation for Multi-Sensor Data Fusion", submitted to the IEEE 28th International Conference on Information Fusion, Rio de Janeiro, Brazil, Jul 7-11, 2025.

[2] B. Mamandipoor, D. Ramasamy, and U. Madhow, "Newtonized orthogonal matching pursuit: Frequency estimation over the continuum," IEEE Trans. Signal Process., vol. 64, no. 19, pp. 5066-5081, Oct. 2016.

[3] A. S. Rahmathullah, A. F. Garcia-Fernandez, and L. Svensson,"Generalized optimal sub-pattern assignment metric," in 20th Int. Conf. Inf. Fusion, Xi'an, China, Jul. 10-13, 2017.