System and method for synthetic aperture radar target recognition using multi-layer, recurrent spiking neuromorphic networks

1. A method for identifying a target in a synthetic aperture radar signal, the method comprising:

extracting, by a feature extractor, a plurality of features directly from phase history data of the synthetic aperture radar signal;

encoding, by an input spiking neural network, the features as a first plurality of spiking signals;

supplying the spiking signals to a multi-layer recurrent neural network to compute a second plurality of spiking signals;

computing, by a readout neural layer, a signal identifier based on the second plurality of spiking signals; and

outputting the signal identifier from the readout neural layer, the signal identifier identifying the target.

2. The method of claim 1 , wherein the plurality of features comprise an amplitude of the synthetic aperture radar signal.

3. The method of claim 1 , wherein the plurality of features comprise a phase of the synthetic aperture radar signal.

4. The method of claim 1 , wherein the readout neural layer comprises a linear classifier.

5. The method of claim 1 , wherein the input spiking neural network, the multi-layer recurrent neural network, and the readout neural layer are implemented by a neuromorphic chip.

6. The method of claim 1 , further comprising computing average spiking rates from the second plurality of spiking signals,

wherein the signal identifier is computed based on the average spiking rates.

7. The method of claim 1 , wherein the multi-layer recurrent neural network comprises:

a first excitatory neuron layer configured to receive the first plurality of spiking signals from the input spiking neural network;

a first inhibitory neuron layer;

a second excitatory neuron layer; and

a second inhibitory neuron layer connected to the second excitatory neuron layer,

the first excitatory neuron layer being configured to supply spiking signals to:

the first excitatory neuron layer;

the first inhibitory neuron layer; and

the second excitatory neuron layer,

the first inhibitory neuron layer being configured to supply spiking signals to the first excitatory neuron layer,

the second excitatory neuron layer being configured to supply spiking signals to:

the second excitatory neuron layer;

the second inhibitory neuron layer; and

the readout neural layer, and

the second inhibitory neuron layer being configured to supply spiking signals to the second excitatory neuron layer.

8. The method of claim 7 , wherein the input spiking neural network comprises a plurality of input neurons arranged in a grid,

wherein the first excitatory neuron layer comprises a plurality of first excitatory neurons arranged in a grid, the plurality of first excitatory neurons comprising a plurality of critical neurons uniformly distributed in the grid, and

wherein the plurality of input neurons are connected to the critical neurons to maintain spatial relationships between the input neurons in corresponding ones of the critical neurons.

9. The method of claim 7 , wherein the first excitatory neuron layer comprises a plurality of excitatory neurons arranged in a grid, and

wherein a neuron of the first excitatory neuron layer is configured to supply spiking signals to neurons in a local neighborhood of the grid around the neuron.

10. The method of claim 7 , wherein the first excitatory neuron layer comprises a plurality of first excitatory neurons arranged in a grid,

wherein the second excitatory neuron layer comprises a plurality of second excitatory neurons arranged in a grid, and

wherein the plurality of first excitatory neurons are connected to the second excitatory neurons to maintain spatial relationships between the first excitatory neurons in corresponding ones of the second excitatory neurons.

11. A system configured to identify a target in a synthetic aperture radar signal, the system comprising:

a feature extractor configured to extract a plurality of features directly from phase history data of the synthetic aperture radar signal;

an input spiking neural network configured to encode the features as a first plurality of spiking signals;

a multi-layer recurrent neural network configured to compute a second plurality of spiking signals based on the first plurality of spiking signals;

a readout neural layer configured to compute a signal identifier based on the second plurality of spiking signals; and

an output configured to output the signal identifier, the signal identifier identifying the target.

12. The system of claim 11 , wherein the plurality of features comprise an amplitude of the synthetic aperture radar signal.

13. The system of claim 11 , wherein the plurality of features comprise a phase of the synthetic aperture radar signal.

14. The system of claim 11 , wherein the readout neural layer comprises a linear classifier.

15. The system of claim 11 , wherein the input spiking neural network, the multi-layer recurrent neural network, and the readout neural layer are implemented by a neuromorphic chip.

16. The system of claim 11 , wherein the readout neural layer is configured to compute average spiking rates from the second plurality of spiking signals,

wherein the signal identifier is computed based on the average spiking rates.

17. The system of claim 11 , wherein the multi-layer recurrent neural network comprises:

a first excitatory neuron layer configured to receive the first plurality of spiking signals from the input spiking neural network;

a first inhibitory neuron layer;

a second excitatory neuron layer; and

a second inhibitory neuron layer connected to the second excitatory neuron layer,

the first excitatory neuron layer being configured to supply spiking signals to:

the first excitatory neuron layer;

the first inhibitory neuron layer; and

the second excitatory neuron layer,

the first inhibitory neuron layer being configured to supply spiking signals to the first excitatory neuron layer,

the second excitatory neuron layer being configured to supply spiking signals to:

the second excitatory neuron layer;

the second inhibitory neuron layer; and

the readout neural layer, and

the second inhibitory neuron layer being configured to supply spiking signals to the second excitatory neuron layer.

18. The system of claim 17 , wherein the input spiking neural network comprises a plurality of input neurons arranged in a grid,

wherein the first excitatory neuron layer comprises a plurality of first excitatory neurons arranged in a grid, the plurality of first excitatory neurons comprising a plurality of critical neurons uniformly distributed in the grid, and

wherein the plurality of input neurons are connected to the critical neurons to maintain spatial relationships between the input neurons in corresponding ones of the critical neurons.

19. The system of claim 17 , wherein the first excitatory neuron layer comprises a plurality of excitatory neurons arranged in a grid, and

wherein a neuron of the first excitatory neuron layer is configured to supply spiking signals to neurons in a local neighborhood of the grid around the neuron.

20. The system of claim 17 , wherein the first excitatory neuron layer comprises a plurality of first excitatory neurons arranged in a grid,

wherein the second excitatory neuron layer comprises a plurality of second excitatory neurons arranged in a grid, and

wherein the plurality of first excitatory neurons are connected to the second excitatory neurons to maintain the spatial relationships between the first excitatory neurons in corresponding ones of the second excitatory neurons.

21. A system for identifying a target in a synthetic aperture radar signal, the system comprising:

means for extracting a plurality of features directly from phase history data of the synthetic aperture radar signal;

means for encoding the features as a first plurality of spiking signals;

means for supplying the spiking signals to a multi-layer recurrent neural network to compute a second plurality of spiking signals;

means for computing a signal identifier based on the second plurality of spiking signals; and

means for outputting the signal identifier from the means for computing the signal identifier, the signal identifier identifying the target.