Pipelined cognitive signal processor
Techniques for denoising an electromagnetic signal are disclosed. The techniques utilize an antenna, a weight adaptation component, a reservoir computer including a computer interpretable neural network, a delay embedding component, and an output layer computer. The techniques include passively acquiring an electromagnetic signal by the antenna, producing a plurality of reservoir state values by the reservoir computer based on the electromagnetic signal, collecting the plurality of reservoir state values by the delay embedding component into a historical record, determining a plurality of reservoir state value weights by the weight adaptation component based at least in part of the historical record, scaling, by the plurality of reservoir state value weights, to produce a plurality of output values, the plurality of reservoir state values by the output layer computer, and outputting the plurality of output values, where the scaling occurs over a plurality of clock cycles of a clock for the system.
1. A cognitive signal processor system for denoising an electromagnetic signal, the cognitive signal processing system comprising:
a reservoir computer that receives an electromagnetic signal and produces reservoir state values, wherein the reservoir computer comprises a neural network, and wherein the electromagnetic signal comprises a mixture of waveforms over an ultra-wide bandwidth;
a delay embedding component that creates a finite temporal record of the reservoir state values, the delay embedding component coupled to the reservoir computer; and
a weight adaptation component that determines state value weights for the reservoir state values based at least in part on the finite temporal record, wherein the weight adaptation component uses an error between a predicted signal and the electromagnetic signal to tune output weights of the reservoir, the weight adaptation component coupled to the delay embedding component,
wherein the cognitive signal processor system is configured to scale the reservoir state values by the state value weights for the reservoir state values to produce and output a plurality of outputs.
2. The cognitive signal processor system of claim 1 , wherein the cognitive signal processor system is configured to scale the reservoir state values by the state value weights for the reservoir state values over a plurality of clock cycles of a clock for the cognitive signal processor system.
3. The cognitive signal processor system of claim 1 , wherein the reservoir computer models feedback connections between nodes of the neural network.
4. The cognitive signal processor system of claim 1 , wherein the reservoir state values are based on the mixture of waveforms of the electromagnetic signal received by the reservoir computer, and the reservoir computer maps the electromagnetic signal to a dynamical system.
5. The cognitive signal processor system of claim 1 , wherein the reservoir computer comprises electronic memory for storing feed forward connections from inputs to a plurality of reservoir states.
6. The cognitive signal processor system of claim 1 , wherein the reservoir computer, the delay embedding component, and the weight adaptation component are implemented one of: a Complementary Metal Oxide Semiconductor (CMOS), or a Field Programmable gate Array (FPGA).
7. The cognitive signal processor system of claim 1 , wherein the weight adaptation component is configured to determine the reservoir state value weights over a plurality of clock cycles of a clock for the cognitive signal processor system.
8. The cognitive signal processor system of claim 1 , wherein the cognitive signal processor system receives the electromagnetic signal from a single antenna.
9. A method of denoising an electromagnetic signal using a cognitive signal processor, the method comprising:
producing, by a reservoir computer, reservoir state values based on an electromagnetic signal, wherein the reservoir computer comprises a neural network, and wherein the electromagnetic signal comprises a mixture of waveforms over an ultra-wide bandwidth;
creating, by a delay embedding component coupled to the reservoir computer, a finite temporal record of the reservoir state values;
determining, by a weight adaptation component coupled to the delay embedding component, state value weights for the reservoir state values based at least in part on the finite temporal record, wherein the weight adaptation component uses an error between a predicted signal and the electromagnetic signal to tune output weights of the reservoir;
scaling the reservoir state values by the state value weights for the reservoir state values to produce a plurality of outputs; and
outputting the plurality of output values.
10. The method of claim 9 , wherein the scaling occurs over a plurality of clock cycles of a clock for the cognitive signal processor.
11. The method of claim 9 , further comprising modeling, by the reservoir computer, feedback connections between nodes of the neural network.
12. The method of claim 9 , wherein the reservoir state values are based on the mixture of waveforms of the electromagnetic signal received by the reservoir computer, and the reservoir computer maps the electromagnetic signal to a dynamical system.
13. The method of claim 9 , further comprising storing, by the reservoir computer, feed forward connections from inputs to a plurality of reservoir states in electronic memory.
14. The method of claim 9 , wherein the reservoir computer, the delay embedding component, and the weight adaptation component are implemented one of: a Complementary Metal Oxide Semiconductor (CMOS), or a Field Programmable gate Array (FPGA).
15. The method of claim 9 , wherein the determining occurs over a plurality of clock cycles of a clock for the cognitive signal processor.
16. The method of claim 9 , wherein the cognitive signal processor receives the electromagnetic signal from a single antenna.
17. A cognitive signal processor system for denoising an electromagnetic signal, the cognitive signal processing system comprising:
a reservoir computer that receives an electromagnetic signal and produces reservoir state values, wherein the reservoir computer comprises a neural network, and wherein the electromagnetic signal comprises a mixture of waveforms over an ultra-wide bandwidth;
a delay embedding component that creates a finite temporal record of the reservoir state values, the delay embedding component coupled to the reservoir computer; and
a weight adaptation component that determines state value weights for the reservoir state values based at least in part on the finite temporal record, the weight adaptation component coupled to the delay embedding component, wherein the weight adaptation component adapts an output of the reservoir computer via gradient descent to produce a prediction of the electromagnetic signal,
wherein the cognitive signal processor system is configured to scale the reservoir state values by the state value weights for the reservoir state values to produce and output a plurality of outputs.
18. The cognitive signal processor system of claim 17 , wherein the reservoir computer models feedback connections between nodes of the neural network.
19. A method of denoising an electromagnetic signal using a cognitive signal processor, the method comprising:
producing, by a reservoir computer, reservoir state values based on an electromagnetic signal, wherein the reservoir computer comprises a neural network, and wherein the electromagnetic signal comprises a mixture of waveforms over an ultra-wide bandwidth;
creating, by a delay embedding component coupled to the reservoir computer, a finite temporal record of the reservoir state values;
determining, by a weight adaptation component coupled to the delay embedding component, state value weights for the reservoir state values based at least in part on the finite temporal record, wherein the weight adaptation component adapts an output of the reservoir computer via gradient descent to produce a prediction of the electromagnetic signal;
scaling the reservoir state values by the state value weights for the reservoir state values to produce a plurality of outputs; and
outputting the plurality of output values.
20. The method of claim 19 , further comprising modeling, by the reservoir computer, feedback connections between nodes of the neural network.