The learning device learns the target detection model used in the radar device. The learning device includes an acquisition unit, a learning data generation unit, and a learning processing unit. The acquisition unit acquires a reception signal generated based on the received wave and a tracking signal generated based on the reception signal from the radar device. The learning data generation unit generates learning data using the reception signal and the tracking signal. The learning processing unit learns a target detection model that detects a target from the reception signal, using the learning data.

False alarms in RADAR processing are reduced. One or more transforms may be performed to generate an array of spectrum values for a first domain spanning at least one of a range axis, a direction of arrival (DoA) axis, or a velocity axis. One or more spectrum values may be obtained from the array of spectrum values, wherein for each of the one or more spectrum values, (1) the spectrum value is associated with a range estimate, and (2) the spectrum value exceeds a range-dependent maximum threshold established based on a quartic function of the range estimate. The one or more spectrum values identified as exceeding the range-dependent maximum threshold may be excluded, or one or more reduced-magnitude values obtained, to generate an array of modified spectrum values for the first domain, used to generate a range estimate, a DoA estimate, or a velocity estimate, or any combination thereof.

Aspects of the present disclosure are directed to radar apparatuses and related methods. As may be implemented in connection with one or more embodiments, frequency-based representations of reflected radar signals received by different radar receivers are processed utilizing superposition of and combining of respective ones of the frequency-based representations. In response to said processing, phase noise in the frequency-based representations of reflected radar signals is corrected.

This document describes techniques and systems related to a radar system using a machine-learned model for stationary object detection. The radar system includes a processor that can receive radar data as time-series frames associated with electromagnetic (EM) energy. The processor uses the radar data to generate a range-time map of the EM energy that is input to a machine-learned model. The machine-learned model can receive as inputs extracted features corresponding to the stationary objects from the range-time map for multiple range bins at each of the time-series frames. In this way, the described radar system and techniques can accurately detect stationary objects of various sizes and extract critical features corresponding to the stationary objects.

A signal processing device, includes: an azimuth estimation unit configured to estimate an arrival azimuth of a radio wave based on a reception signal of plural antennas; an estimated reception signal calculation unit configured to calculate an estimated reception signal based on an estimation result of the arrival azimuth, for comparison with the reception signal; a residual signal calculation unit configured to calculate a residual signal which is a difference between the reception signal and the estimated reception signal; and a determination unit configured to determine whether the estimation result of the arrival azimuth is correct based on the residual signal.

Embodiments are disclosed that for synthetic aperture radar (SAR) systems and methods that process radar image data to generate radar images using vector processor engines, such as single-instruction-multiple-data (SIMD) processor engines. The vector processor engines can be further augmented with accelerators that vectorize element selection thereby expediting memory accesses required for interpolation operations performed by the vector processor engines.

Techniques and architectures for managing radar sensor processing chains. A first high-frequency radio signal is received with a first RF receiver in the plurality of RF sensor suites on a host platform. The received high-frequency radio signal is converted to a lower second frequency range. A chirplet transform is performed on the signal in the second frequency range. Stored relative location information for a second RF receiver in the plurality of RF sensor suites is retrieved. Radar waveform information corresponding to the second RF receiver in a processing stream corresponding to the first RF receiver is extracted by utilizing the retrieved information and results from the chirplet transform. A point cloud is generated based on the converted signal in the second frequency range and the extracted radar waveform information.

According to an aspect, a method of determining Doppler in a radar system comprising receiving a set of chirps, sampling in time the set of chirps to generate a set of non-uniform samples with one sample per chirp that is non-uniform in time in each chirp across the set of chirps, generating a first Doppler frequency from the set of non-uniform samples, generating a set of non-aliased Doppler frequencies for the first Doppler frequencies from a corresponding set of hypothesis, determining a first set of angles of arrival for every non-aliased Doppler frequency in the a set of non-aliased Doppler frequencies and selecting a first non-aliased Doppler frequency in the set of non-aliased Doppler frequencies that corresponds to the first angle of arrival with a minimum error in the a set of angles of arrival.

An imaging method using a doppler radar wherein the pointing direction in transmission (d.sub.ei) is modified from recurrence to recurrence; each detection block of duration T comprises a periodic repetition of a number C of pointing cycles, each of these cycles comprising a number P of recurrences, the set of these P recurrences covering the D.sub.e pointing directions (d.sub.ei) of the set; the order of the pointings is modified in a pseudo-random manner from pointing cycle to pointing cycle during a same detection block so as to create an irregular time interval between two pointings in a same direction; at least one beam is formed in reception on each recurrence in a direction included in the transmission-focused angular domain in the pointing direction corresponding to the recurrence.

This disclosure provides a signal processing method and apparatus. The method includes: obtaining Nr1×M1 signals, where the Nr1×M1 signals are echo signals of M1 signals that are sent by a radar to a target in a SIMO mode; obtaining Nt×Nr2×M2 signals, where the Nt×Nr2×M2 signals are echo signals of M2 signals that are sent by the radar to the target in a MIMO mode; performing first signal processing on the Nr1×M1 signals to obtain first processing data, where the first signal processing includes sequentially performing range FFT analysis, linear prediction, and Doppler FFT analysis; performing second signal processing on the Nt×Nr2×M2 signals to obtain second processing data, where the second signal processing includes range FFT analysis and Doppler FFT analysis; and performing velocity matching ambiguity resolution processing based on the first processing data and the second processing data.