US 11,852,750 B2
Method and apparatus for radar signal processing using recurrent neural network
Jong Il Kim, Yongin-si (KR)
Assigned to SMART RADAR SYSTEM, INC., Seongnam-si (KR)
Filed by SMART RADAR SYSTEM, INC., Seongnam-si (KR)
Filed on Jun. 28, 2019, as Appl. No. 16/457,887.
Prior Publication US 2020/0408881 A1, Dec. 31, 2020
This patent is subject to a terminal disclaimer.
Int. Cl. G01S 7/41 (2006.01); G01S 7/35 (2006.01)
CPC G01S 7/417 (2013.01) [G01S 7/352 (2013.01); G01S 7/356 (2021.05)] 10 Claims
OG exemplary drawing
 
1. A method for radar signal processing performed by an apparatus for tracking a radar target, the method comprising:
a signal reception operation to receive a frequency modulated continuous wave (FMCW) digital radar signal by a radar antenna;
a first Fourier transform operation to perform a first Fourier transform on the FMCW digital radar signal at every pulse repetition interval (PRI) to generate range data, which indicates range-index-specific coefficient values;
a recurrent neural network (RNN) processing operation to determine the probability of the existence of a target in each range index based on the range data output at every PRI, wherein the RNN is configured as a supervised learning type network trained using information regarding whether the target exists in each of the range indices;
a target selection operation to select first range indices each representing that a target exists based on the probabilities of the existence of the target in the range indices, wherein the range indices further comprise second range indices each representing that a target does not exist based on the probabilities of the existence of the target;
a storage operation to store range data of the selected first range indices in which the target exists over a plurality of PRIs;
a second Fourier transform operation to perform a second Fourier transform on range data of the first range indices in the direction of the PRI index, wherein the first range indices are selected by the target selection operation, each representing the target exists to calculate a speed-index-specific coefficient value, wherein the second Fourier transform operation is not performed for the second range indices each representing that a target does not exist based on the probabilities of the existence of the target; and
an output operation to output ranges and speeds of the selected targets.