Granted Patent B2

US 12,625,227 · App. 18/483,792 · Granted May 12, 2026

Inventors: Jeroen Overdevest (Eindhoven, NL); Marco Jan Gerrit Bekooij (Empel, NL); Arie Geert Cornelis Koppelaar (Giessen, NL)

Assignee: NXP B.V.

G01S7/023G01S7/352G06F17/11

Abstract

Claims (115)

1 . A method of processing radar data, the method comprising:

receiving a mask that identifies a set of samples in received radar signalling that are detected as including interference, wherein the mask comprises a matrix of data having a fast-time dimension and a slow-time dimension;

receiving radar data comprising a matrix of samples of received radar signalling having a fast-time dimension and a slow-time dimension wherein the set of samples identified by the mask have been set to a predetermined value to remove said samples including interference; and

determining a reconstruction of the radar data in which at least the set of samples of the radar data are replaced with estimated samples, wherein said determining a reconstruction of the radar data comprises formulating an optimization problem based on the radar data and the mask, and applying an iterative method to solve the optimization problem at least in part in the range-Doppler domain wherein an output of each iteration of the iterative method is converted to the time domain and wherein reconstruction of the radar data comprises said output after at least one iteration, wherein

a first iteration of said application of the iterative method to solve the optimization problem includes

determining a two-dimensional Fourier Transform of the radar data multiplied by a predetermined scalar, μ, wherein the two-dimensional Fourier Transform provides for conversion to the range-Doppler domain,

applying a soft thresholding function to the two-dimensional Fourier Transform of the radar data multiplied by the predetermined scalar, to determine a thresholded dataset, and

determining an output of the first iteration by determining an Inverse two-dimensional Fourier Transform of the thresholded dataset to provide for the conversion to the time domain, and

each subsequent iteration of said iterative method includes determining an output of the subsequent iteration by the steps of

determining a first function comprising the difference between an element-wise multiplication of the mask and an output of an iteration comprising an immediately prior iteration, and the radar data,

determining a second function comprising a scalar multiplied by the first function, wherein the scalar is termed a step-size scalar,

determining a third function comprising the output of the iteration that comprises the immediately prior iteration minus the second function,

determining a fourth function comprising the application of a complex soft thresholding function to a two-dimensional Fourier Transform of the third function, and

determining an inverse two-dimensional Fourier Transform of said fourth function.

2 . The method of claim 1 , wherein the iterative method includes application of a thresholding function in the range-Doppler domain.

3 . The method of claim 1 , wherein the first iteration of said iterative method is configured to apply the soft thresholding function to a function of the range-Doppler processed radar data.

4 . The method of claim 3 , wherein the determination of the reconstruction of the radar data comprises a plurality of iterations of the iterative method; and

wherein the subsequent iteration of said iterative method, after the first iteration, is configured to apply the soft thresholding function to a function of the output of a previous iteration, the mask and the radar data.

5 . The method of claim 1 , wherein said complex soft thresholding function comprises T λ (x)=e j∠ x (|x|−λ) + wherein x represents the data to which the complex soft thresholding function is applied and λ represents the threshold of the thresholding function, wherein values of x that have |x|<λ will be set to zero and the other values will be scaled to |x|−λ.

6 . The method of claim 1 , wherein the step-size scalar comprises one.

7 . The method of claim 1 , wherein said iterative method is performed based on the step-size scalar μ k which defines a step-size for each iteration of the iterative method and a shrinkage-threshold λ k which defines a threshold of the complex soft thresholding function applied in each iteration and wherein said method includes using an updated step-size scalar μ k and updated shrinkage-threshold λ k in the subsequent iteration.

8 . The method of claim 7 , wherein the updated step-size scalar μ k and the updated shrinkage-threshold λ k for use in the subsequent iteration or iterations is determined using a deep learning process involving back-propagation.

9 . A method of processing radar data, the method comprising:

receiving a mask that identifies a set of samples in received radar signalling that are detected as including interference, wherein the mask comprises a matrix of data having a fast-time dimension and a slow-time dimension;

receiving radar data comprising a matrix of samples of received radar signalling having a fast-time dimension and a slow-time dimension, wherein the set of samples identified by the mask have been set to a predetermined value to remove said samples including interference; and

determining a reconstruction of the radar data in which at least the set of samples of the radar data are replaced with estimated samples, wherein

said determining a reconstruction of the radar data comprises formulating an optimization problem based on the radar data and the mask, and applying an iterative method to solve the optimization problem at least in part in the range-Doppler domain,

an output of each iteration of the iterative method is converted to the time domain, reconstruction of the radar data comprises said output after at least one iteration,

a first iteration of said application of the iterative method to solve the optimization problem includes

determining a two-dimensional Fourier Transform of the radar data multiplied by a predetermined scalar, μ, wherein the two-dimensional Fourier Transform provides for conversion to the range-Doppler domain,

applying a soft thresholding function to the two-dimensional Fourier Transform of the radar data multiplied by the predetermined scalar, to determine a thresholded dataset, and

determining an output of the first iteration by determining an Inverse two-dimensional Fourier Transform of the thresholded dataset to provide for the conversion to the time domain, and

each subsequent iteration of said iterative method comprises determining an output of the subsequent iteration by the steps of

determining a first function comprising the element-wise multiplication of a function of the mask and an output of an iteration comprising an immediately prior iteration, wherein the function of the mask comprises (1−μm) wherein μ comprises a predetermined scalar termed a step-size scalar,

determining a second function comprising the first function added to the radar data scaled by said step-size scalar,

determining a third function comprising the application of a complex soft thresholding function to a two-dimensional Fourier Transform of the second function, and

determining an inverse two-dimensional Fourier Transform of said third function.

10 . A method of processing radar data, the method comprising:

receiving a mask that identifies a set of samples in received radar signalling that are detected as including interference, wherein the mask comprises a matrix of data having a fast-time dimension and a slow-time dimension;

receiving radar data comprising a matrix of samples of received radar signalling having a fast-time dimension and a slow-time dimension wherein the set of samples identified by the mask have been set to a predetermined value to remove said samples including interference; and

determining a reconstruction of the radar data in which at least the set of samples of the radar data are replaced with estimated samples, wherein said determining a reconstruction of the radar data comprises formulating an optimization problem based on the radar data and the mask, and applying an iterative method to solve the optimization problem at least in part in the range-Doppler domain wherein an output of each iteration of the iterative method is converted to the time domain and wherein reconstruction of the radar data comprises said output after at least one iteration, and

wherein an output of an iteration of said iterative method is defined by x k wherein:

x k =F −1 {T λ k ( F{s k −μ k ( m⊙s k −y )})}

wherein F{ } and F −1 { } represent a two-dimensional Fourier transform and inverse two-dimensional Fourier transform respectively, T λ k represents a complex soft thresholding function with threshold λ k , m represents said mask; y represents said radar data and μ k represents an step-size scalar and ⊙ represents an element-wise multiplication; and wherein:

t

k

+

1

=

1

+

1

+

4

⁢

t

k

2

2

⁢

and

s

k

+

1

=

x

k

+

t

k

-

1

t

k

+

1

⁢

(

x

k

-

x

k

-

1

)

.

11 . A method of processing radar data, the method comprising:

receiving a mask that identifies a set of samples in received radar signalling that are detected as including interference, wherein the mask comprises a matrix of data having a fast-time dimension and a slow-time dimension;

receiving radar data comprising a matrix of samples of received radar signalling having a fast-time dimension and a slow-time dimension wherein the set of samples identified by the mask have been set to a predetermined value to remove said samples including interference; and

determining a reconstruction of the radar data in which at least the set of samples of the radar data are replaced with estimated samples, wherein said determining a reconstruction of the radar data comprises formulating an optimization problem based on the radar data and the mask, and applying an iterative method to solve the optimization problem at least in part in the range-Doppler domain wherein an output of each iteration of the iterative method is converted to the time domain and wherein reconstruction of the radar data comprises said output after at least one iteration, and

wherein only said set of samples are replaced with estimated samples such that said reconstruction of the radar data is designated {circumflex over (x)} wherein

{circumflex over (x)}=m⊙x +(1 −m )⊙ x x

wherein m designates the mask, x designated the radar data and x k designates the output of at least one iteration of said iterative method.

12 . A processor configured to:

receive a mask that identifies a set of samples in received radar signalling that are detected as including interference, wherein the mask comprises a matrix of data having a fast-time dimension and a slow-time dimension;

receive radar data comprising a matrix of samples of received radar signalling having a fast-time dimension and a slow-time dimension wherein the set of samples identified by the mask have been set to a predetermined value to remove said samples including interference; and

determine a reconstruction of the radar data in which at least the set of samples of the radar data are replaced with estimated samples, wherein said determination of a reconstruction of the radar data comprises formulating an optimization problem based on the radar data and the mask, and applying an iterative method to solve the optimization problem at least in part in the range-Doppler domain, wherein

an output of each iteration of the iterative method is converted to the time domain,

reconstruction of the radar data comprises said output based on at least one iteration,

a first iteration of said application of the iterative method to solve the optimization problem includes

determining a two-dimensional Fourier Transform of the radar data multiplied by a predetermined scalar, μ, wherein the two-dimensional Fourier Transform provides for conversion to the range-Doppler domain,

applying a soft thresholding function to the two-dimensional Fourier Transform of the radar data multiplied by the predetermined scalar, to determine a thresholded dataset, and

determining an output of the first iteration by determining an Inverse two-dimensional Fourier Transform of the thresholded dataset to provide for the conversion to the time domain, and

each subsequent iteration of said iterative method includes determining an output of the subsequent iteration by

determining a first function comprising a difference between an element-wise multiplication of the mask and an output of an iteration comprising an immediately prior iteration, and the radar data,

determining a second function comprising a scalar multiplied by the first function, wherein the scalar is termed a step-size scalar,

determining a third function comprising the output of the iteration that comprises the immediately prior iteration minus the second function,

determining a fourth function comprising the application of a complex soft thresholding function to a two-dimensional Fourier Transform of the third function, and

determining an inverse two-dimensional Fourier Transform of said fourth function.

13 . The processor of claim 12 , wherein the iterative method includes application of a thresholding function in the range-Doppler domain.

14 . The processor of claim 12 , wherein the first iteration of said iterative method is configured to apply the soft thresholding function to a function of the range-Doppler processed radar data.

15 . The processor of claim 12 , wherein the determination of the reconstruction of the radar data comprises a plurality of iterations of the iterative method; and

wherein the subsequent iteration of said iterative method, after the first iteration, is configured to apply the soft thresholding function to a function of the output of a previous iteration, the mask and the radar data.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (34)