Nonlinear filter
View Patent ↗A system and method are disclosed for a nonlinear filter having a nonlinear transfer function. The nonlinear filter comprises a plurality of linear filters each having a filter output, a plurality of nonlinear elements each connected to one of the plurality of linear filters, and a combination network connected to the plurality of nonlinear elements. The nonlinear elements are used to produce nonlinear effects and generate a plurality of nonlinear outputs, and the combination network combines the nonlinear outputs.
1. A nonlinear filter having a nonlinear transfer function comprising:
a plurality of linear filters each having a filter output and a filter input, wherein the filter input of each of the plurality of linear filters receive a common input signal;
a plurality of nonlinear elements each connected to one of the plurality of linear filters; and
a combination network connected to the plurality of nonlinear elements; wherein:
the nonlinear elements are configured to perform absolute value operations to generate a plurality of nonlinear outputs;
a plurality of scalers scale each of the plurality of nonlinear outputs; and
the combination network combines the scaled nonlinear outputs.
2. A nonlinear filter as recited in claim 1 wherein the combination network comprises a summation network.
3. A nonlinear filter as recited in claim 1 wherein the nonlinear transfer function is a first nonlinear transfer function that includes a first order polynomial; and the first nonlinear transfer function approximates a second nonlinear transfer function.
4. A nonlinear filter as recited in claim 1 wherein the nonlinear transfer function is a first nonlinear transfer function that includes a first order polynomial; and the first nonlinear transfer function approximates a second nonlinear transfer function that includes a higher order polynomial.
5. A nonlinear filter as recited in claim 1 , wherein one of the plurality of linear filters is an averaging filter.
6. A nonlinear filter as recited in claim 1 wherein each of the linear filters has linear filter parameters that are derived adaptively.
7. A nonlinear filter as recited in claim 1 wherein each of the linear filters has linear filter parameters that are derived using least mean square fit.
8. A nonlinear filter as recited in claim 1 wherein each of the linear filters has linear filter parameters that are derived using back propagation.
9. A nonlinear filter as recited in claim 1 wherein the nonlinear transfer function is multidimensional.
10. A nonlinear filter having a nonlinear transfer function comprising:
a plurality of linear filters each having a filter output;
a plurality of nonlinear elements each connected to one of the plurality of linear filters; and
a combination network connected to the plurality of nonlinear elements; wherein:
the combination network comprises a min-max selection network;
the nonlinear elements are configured to perform absolute value operations to generate a plurality of nonlinear outputs; and
the combination network combines the nonlinear outputs.
11. A nonlinear filter as recited in claim 10 , wherein the min-max selection network comprises a minimum processor and a maximum processor in a nested configuration.
12. A method for providing a nonlinear output comprising:
receiving an input signal;
sending the input signal to a plurality of linear filters to generate a plurality of filter outputs;
performing an absolute value operation on the filter outputs to generate a plurality of nonlinear outputs;
performing a scaling operation on the nonlinear outputs to generate a plurality of scaled nonlinear outputs; and
combining the scaled nonlinear outputs.
13. A method as recited in claim 12 wherein combining the scaled nonlinear outputs comprises summing the nonlinear outputs.
14. A method as recited in claim 12 wherein the nonlinear transfer function is a first nonlinear transfer function that includes a first order polynomial; and the first nonlinear transfer function approximates a second nonlinear transfer function.
15. A method as recited in claim 12 wherein the nonlinear transfer function is a first nonlinear transfer function that includes a first order polynomial; and the first nonlinear transfer function approximates a second nonlinear transfer function that includes a higher order polynomial.
16. A method as recited in claim 12 , further comprising providing an average value for the nonlinear transfer function.
17. A method as recited in claim 12 wherein the nonlinear transfer function is multidimensional.
18. A method for providing a nonlinear output comprising:
receiving an input signal;
sending the input signal to a plurality of linear filters to generate a plurality of filter outputs;
performing an absolute value operation on the filter outputs to generate a plurality of nonlinear outputs; and
combining the nonlinear outputs, wherein combining the nonlinear outputs comprises performing a min-max selection.
19. A computer program product for producing a nonlinear output, the computer program product being embodied in a computer readable medium and comprising computer instructions for:
receiving an input signal;
sending the input signal to a plurality of linear filters to generate a plurality of filter outputs;
performing an absolute value operation on the filter outputs to generate a plurality of nonlinear outputs;
performing a scaling operation on the nonlinear outputs to generate a plurality of scaled nonlinear outputs; and
combining the scaled nonlinear outputs.
20. A computer program product as recited in claim 19 wherein combining the scaled nonlinear outputs comprises summing the scaled nonlinear outputs.
21. A computer program product as recited in claim 19 wherein the nonlinear transfer function is multidimensional.
22. A computer program product for producing a nonlinear output, the computer program product being embodied in a computer readable medium and comprising computer instructions for:
receiving an input signal;
sending the input signal to a plurality of linear filters to generate a plurality of filter outputs;
performing an absolute value operation on the filter outputs to generate a plurality of nonlinear outputs; and
combining the nonlinear outputs, wherein combining the nonlinear outputs comprises performing a min-max selection.