FMCW radar with interference signal suppression using artificial neural network

Patent metadata

Field	Value
Publication number	US-11885903-B2
Application number	US-202016817385-A
Country	US
Kind code	B2
Filing date	Mar 12, 2020
Priority date	Mar 14, 2019
Publication date	Jan 30, 2024
Grant date	Jan 30, 2024

Abstract

A method for a radar device is described below. According to an example implementation, the method comprises transmitting an RF transmission signal that comprises a plurality of frequency-modulated chirps, and receiving an RF radar signal and generating a dataset containing in each case a particular number of digital values based on the received RF radar signal. A dataset may in this case be associated with a chirp or a sequence of successive chirps. The method furthermore comprises filtering the dataset by way of a neural network to which the dataset is fed in order to reduce an interfering signal contained therein. A convolutional neural network is used as the neural network.

First claim

What is claimed is:

1. A radar device, comprising:
a radar receiver that is designed to receive a radio-frequency (RF) radar signal associated with a plurality of frequency modulated chirps and, based thereon, to generate a digital signal that comprises a plurality of signal segments; and
a neural network having a plurality of layers each having one or more neurons,
wherein the plurality of signal segments are fed to an input layer of the plurality of layers and wherein the plurality of layers are designed to process the plurality of signal segments of the digital signal,
wherein an output layer of the plurality of layers, has at least one neuron that delivers an output value that indicates whether a respective signal segment or a sample of the respective signal segment is overlaid with an interfering signal originating from at least one radar transmitter of at least one other radar device,
wherein a convolutional kernel is associated with one or more neural network (NN) channels of one or more other layers of the plurality of layers, and
wherein the one or more other layers receive as input, output values of the one or more NN channels of a previous layer.

2. The radar device as claimed in claim 1 , further comprising:
a radar transmitter that is designed to generate an RF transmission signal that comprises a plurality of frequency-modulated pulses,
wherein each of the plurality of signal segments of the digital signal is associated with a respective frequency-modulated pulse of the plurality of frequency-modulated pulses, and
wherein the plurality of signal segments are in a time domain or in a frequency domain.

3. The radar device as claimed in claim 1 ,
wherein the output layer of the neural network is designed to deliver an output value that indicates whether the respective signal segment is overlaid with a frequency-modulated interfering signal from an external frequency-modulated radar transmitter.

4. The radar device as claimed in claim 1 ,
wherein the output layer has a neuron that indicates whether one or more of the plurality of signal segments fed to the input layer contains an interfering signal.

5. The radar device as claimed in claim 1 , further comprising:
a computing unit that is designed to detect a radar target based on the digital signal, wherein one or more of the plurality of signal segments for which the at least one neuron of the output layer of the neural network indicates overlaying of an interfering signal remain unconsidered.

6. The radar device as claimed in claim 1 ,
wherein the output layer has a plurality of neurons that are each associated with a sample of the respective signal segment and wherein each of the plurality of neurons indicates whether the sample associated therewith contains an interfering signal.

7. The radar device of claim 1 , wherein the output values of the one or more NN channels of the one or more other layers are based on a weighted sum of input that is provided to the respective layer, and
wherein the input that is provided to the respective layer is based on a respective convolution kernel.

8. A radar device, comprising:
a radar transmitter that is designed to output a radio-frequency (RF) transmission signal that comprises a plurality of frequency-modulated chirps;
a radar receiver that is designed to receive an RF radar signal associated with a plurality of frequency modulated chirps and, based thereon, to generate a dataset containing a particular number of digital values,
wherein a dataset is associated with a chirp or a sequence of successive chirps; and

a neural network to which the dataset is fed and that is designed to filter the dataset in order to reduce an interfering signal originating from at least one radar transmitter of at least one other radar device contained therein,
wherein the neural network is a convolutional neural network with a plurality of layers,
wherein a convolutional kernel is associated with one or more neural network (NN) channels of a plurality of further layers of the plurality of layers, and
wherein the plurality of further layers receive as input, output values of the one or more NN channels of a previous layer.

9. The radar device as claimed in claim 8 ,
wherein the radar receiver is designed, based on the RF radar signal, to generate a digital radar signal in a time domain that comprises a plurality of signal segments that are associated with a sequence of frequency-modulated chirps, and
wherein the plurality of signal segments represent a radar data array and the dataset is formed through two-dimensional Fourier transformation of the radar data array.

10. The radar device as claimed in claim 8 , wherein the radar device is designed,
based on the RF radar signal, to generate a digital radar signal in a time domain that comprises a plurality of signal segments that are associated with a sequence of frequency-modulated chirps, and
based on the digital radar signal, to calculate a range Doppler map that forms the dataset of digital values.

11. The radar device as claimed in claim 10 ,
wherein the dataset of digital values represents a dataset of complex values;
wherein the convolutional neural network has an input layer with two NN channels,
wherein a first channel of the two NN channels is to deliver real parts of the complex values of the dataset as output values, and

wherein a second channel of the two NN channels is to deliver imaginary parts of the complex values of the dataset as output values.

12. The radar device as claimed in claim 11 ,
wherein each of the plurality of further layers is associated with two or more NN channels.

13. The radar device as claimed in claim 12 ,
wherein a last layer of the plurality of further layers forms an output layer that has two NN channels, and
wherein a first channel of the two NN channels of the output layer is to deliver real parts as output values and a second channel of the two NN channels of the output layer is to deliver imaginary parts of complex values of a filtered dataset in which interfering signal components are reduced as output values.

14. The radar device as claimed in claim 13 ,
wherein the output layer has precisely two NN channels and each of the rest of the plurality of further layers has more than two NN channels.

15. The radar device as claimed in claim 12 ,
wherein each NN channel of each of the plurality of further layers of the neural network is to deliver a number of real output values that corresponds to a number of complex values of the dataset.

16. The radar device as claimed in claim 8 ,
wherein output values of an NN channel of each respective layer of the plurality of further layers depends on a weighted sum of input values that are fed to the respective layer.

17. A method for a radar device, comprising:
transmitting a radio-frequency (RF) transmission signal that comprises a plurality of frequency-modulated chirps;
receiving an RF radar signal associated with a plurality of frequency modulated chirps, and generating a dataset containing a particular number of digital values based on the RF radar signal,
wherein a dataset is associated with a chirp or a sequence of successive chirps; and

filtering the dataset using a neural network to which the dataset is fed in order to reduce an interfering signal originating from at least one radar transmitter of at least one other radar device contained therein,
wherein the neural network is a convolutional neural network with a plurality of layers,
wherein a convolutional kernel is associated with one or

Assignees

• Infineon Technologies Ag

Inventors

• Meissner Paul
• Messner Elmar
• Pernkopf Franz
• Rock Johanna
• Toth Mate Andras

Classifications

• G01S7/023 Primary

  Interference mitigation, e.g. reducing or avoiding non-intentional interference with other HF-transmitters, base station transmitters for mobile communication or other radar systems, e.g. using electro-magnetic interference [EMI] reduction techniques (auxiliary means for detecting or identifying radar signals or the like G01S7/021; means for anti-jamming G01S7/36) · CPC title

• G01S7/354

  Extracting wanted echo-signals (Doppler systems G01S13/50) · CPC title

• G01S7/417 Primary

  involving the use of neural networks · CPC title

• G01S7/356

  involving particularities of FFT processing · CPC title

• G01S7/36

  Means for anti-jamming {, e.g. ECCM, i.e. electronic counter-counter measures} · CPC title