Systems and methods of nonlinear RF beamforming
According to some embodiments, a method of nonlinear RF beamforming includes receiving a reference RF signal and computing linear beamformed signals. The method further includes determining a frequency content for each of the plurality of linear beamformed signals, determining a reduced subset of linear beamformed signals, and computing a minimum element-wise magnitude across the reduced subset of linear beamformed signals. The method further includes computing a FFT of a reference channel signal, computing an amplitude mask ratio between the minimum element-wise magnitude and the FFT of the reference channel signal, and applying a learned polynomial response to the amplitude mask ratio to create a modified amplitude mask. The method further includes applying the modified amplitude mask to the FFT of the reference channel signal to create a masked frequency response of the reference channel signal and computing an inverse FFT in order to generate a final beamformed signal.
1. A system comprising:
one or more memory units; and
a processor communicatively coupled to the one or more memory units, the processor configured to:
receive a reference radio frequency (RF) signal;
compute a plurality of linear beamformed signals using the reference RF signal;
determine a frequency content for each of the plurality of linear beamformed signals;
determine a reduced subset of linear beamformed signals from the plurality of linear beamformed signals;
compute a minimum element-wise magnitude across the reduced subset of linear beamformed signals;
compute a fast Fourier transform (FFT) of a reference channel signal;
compute an amplitude mask ratio between the minimum element-wise magnitude and the FFT of the reference channel signal;
apply a learned polynomial response to the amplitude mask ratio to create a modified amplitude mask;
apply the modified amplitude mask to the FFT of the reference channel signal to create a masked frequency response of the reference channel signal; and
compute an inverse FFT of the masked frequency response of the reference channel signal to generate a final beamformed signal.
2. The system of claim 1 , wherein computing the plurality of linear beamformed signals using the reference RF signal comprises using:
delay and sum beamforming; and
a plurality of different phase shifts.
3. The system of claim 1 , wherein determining the frequency content for each of the plurality of linear beamformed signals comprises:
performing FFT; or
performing fast integral transforms.
4. The system of claim 1 , wherein determining the reduced subset of linear beamformed signals from the plurality of linear beamformed signals comprises using a neural network.
5. The system of claim 4 , wherein the neural network comprises:
a plurality of fully-connected layers; and
a plurality of nonlinear activation functions.
6. The system of claim 4 , wherein an amount of signals in the reduced subset of linear beamformed signals is a hyperparameter in the neural network.
7. The system of claim 1 , wherein applying the learned polynomial response to the amplitude mask ratio to create the modified amplitude mask comprises using a neural network.
8. A method by a computing system, the method comprising:
receiving a reference radio frequency (RF) signal;
computing a plurality of linear beamformed signals using the reference RF signal;
determining a frequency content for each of the plurality of linear beamformed signals;
determining a reduced subset of linear beamformed signals from the plurality of linear beamformed signals;
computing a minimum element-wise magnitude across the reduced subset of linear beamformed signals;
computing a fast Fourier transform (FFT) of a reference channel signal;
computing an amplitude mask ratio between the minimum element-wise magnitude and the FFT of the reference channel signal;
applying a learned polynomial response to the amplitude mask ratio to create a modified amplitude mask;
applying the modified amplitude mask to the FFT of the reference channel signal to create a masked frequency response of the reference channel signal; and
computing an inverse FFT of the masked frequency response of the reference channel signal to generate a final beamformed signal.
9. The method of claim 8 , wherein computing the plurality of linear beamformed signals using the reference RF signal comprises using:
delay and sum beamforming; and
a plurality of different phase shifts.
10. The method of claim 8 , wherein determining the frequency content for each of the plurality of linear beamformed signals comprises:
performing FFT; or
performing fast integral transforms.
11. The method of claim 8 , wherein determining the reduced subset of linear beamformed signals from the plurality of linear beamformed signals comprises using a neural network.
12. The method of claim 11 , wherein the neural network comprises:
a plurality of fully-connected layers; and
a plurality of nonlinear activation functions.
13. The method of claim 11 , wherein an amount of signals in the reduced subset of linear beamformed signals is a hyperparameter in the neural network.
14. The method of claim 8 , wherein applying the learned polynomial response to the amplitude mask ratio to create the modified amplitude mask comprises using a neural network.
15. One or more computer-readable non-transitory storage media embodying instructions that, when executed by a processor, cause the processor to perform operations comprising:
receiving a reference radio frequency (RF) signal;
computing a plurality of linear beamformed signals using the reference RF signal;
determining a frequency content for each of the plurality of linear beamformed signals;
determining a reduced subset of linear beamformed signals from the plurality of linear beamformed signals;
computing a minimum element-wise magnitude across the reduced subset of linear beamformed signals;
computing a fast Fourier transform (FFT) of a reference channel signal;
computing an amplitude mask ratio between the minimum element-wise magnitude and the FFT of the reference channel signal;
applying a learned polynomial response to the amplitude mask ratio to create a modified amplitude mask;
applying the modified amplitude mask to the FFT of the reference channel signal to create a masked frequency response of the reference channel signal; and
computing an inverse FFT of the masked frequency response of the reference channel signal to generate a final beamformed signal.
16. The one or more computer-readable non-transitory storage media of claim 15 , wherein computing the plurality of linear beamformed signals using the reference RF signal comprises using:
delay and sum beamforming; and
a plurality of different phase shifts.
17. The one or more computer-readable non-transitory storage media of claim 15 , wherein determining the frequency content for each of the plurality of linear beamformed signals comprises:
performing FFT; or performing fast integral transforms.
18. The one or more computer-readable non-transitory storage media of claim 15 , wherein determining the reduced subset of linear beamformed signals from the plurality of linear beamformed signals comprises using a neural network.
19. The one or more computer-readable non-transitory storage media of claim 18 , wherein the neural network comprises:
a plurality of fully-connected layers; and
a plurality of nonlinear activation functions.
20. The one or more computer-readable non-transitory storage media of claim 18 , wherein an amount of signals in the reduced subset of linear beamformed signals is a hyperparameter in the neural network.