Radar signal processing with stepwise peak detection

Abstract

An automotive radar system includes at least one transmitter and at least one receiver and a processor configured to: receive a received radar signal from the at least one receiver; generating a first sub-section of a radar cube using the received radar signal; detecting a first set of candidate peaks in the first sub-section of the radar cube; generating a second sub-section of the radar cube using the received radar signals; detecting a second set of candidate peaks in the second sub-section of the radar cube; determining a set of locations in a candidate peak dataset, wherein each location in the set of locations is associated with a candidate peak in both the first set of candidate peaks and the second set of candidate peaks; and estimating a direction of arrival of the object using the candidate peaks associated with the set of positions.

Description

Technical Field

[0001] This disclosure generally relates to radar systems and related operating methods. In one aspect, this disclosure relates to an automotive radar system configured to iteratively process received radar signals using a stepwise peak detection scheme to reduce the memory requirements of the radar system's signal processing subsystem. Background Technology

[0002] A radar system transmits electromagnetic signals and receives the reflected signals. The time delay between transmitting and receiving the signals can be determined and used to calculate the distance and / or velocity of the object causing the reflection. For example, in automotive applications, automotive radar systems can be used to determine the distance and / or velocity of oncoming vehicles and other obstacles.

[0003] Automotive radar systems enable the implementation of advanced driver assistance systems (ADAS) functions, which can achieve increasingly safer driving and ultimately realize fully autonomous driving platforms.

[0004] During operation, frequency modulated continuous wave (FMCW) radar systems typically store reflected, received, and sampled radar reflections in so-called radar cubes. These radar cubes are three-dimensional data structures containing received signal data for a specific transmitted signal sample number, chirp number, and transmitter antenna combination. During normal operation, these data cubes can be very large, occupying up to tens of megabytes (MB) of data.

[0005] In automotive radar system applications, such as corner radar systems, the size of the radar unit dictates that inexpensive external memory, such as dynamic random access memory (DRAM), cannot be used. The system must therefore rely on internal and relatively expensive static random access memory (SRAM). Considering the cost of such systems, reducing the memory footprint of radar processing operations in civilian automotive radar systems may be beneficial. Summary of the Invention

[0006] In some aspects, the technology described herein relates to an automotive radar system comprising: at least one transmitter and at least one receiver, wherein the at least one transmitter and the at least one receiver are configured to transmit and receive radar signals, wherein the at least one transmitter and the at least one receiver are coupled to a vehicle; and a processor configured to: receive received radar signals from the at least one receiver; process the received radar signals to generate a first sub-segment of a radar cube; process the first sub-segment of the radar cube to detect a first set of candidate peaks, wherein candidate peaks in the first set of candidate peaks are associated with positions in a range-Doppler matrix; process the received radar signals to generate a second sub-segment of the radar cube; process the second sub-segment of the radar cube to detect a second set of candidate peaks, wherein each candidate peak in the second set of candidate peaks is associated with a position in the range-Doppler matrix; combine the positions of the first set of candidate peaks and the positions of the second set of candidate peaks into a candidate peak dataset; use the candidate peak dataset to determine a set of positions in the candidate peak dataset, wherein each position in the set of positions is associated with a candidate peak in both the first set of candidate peaks and the second set of candidate peaks; and use the candidate peaks associated with the set of positions to estimate the direction of arrival of an object.

[0007] In some respects, the technology described herein relates to an automotive radar system, wherein the processor is configured to delete the first sub-segment from the memory of the automotive radar system after processing the first sub-segment.

[0008] In some respects, the technology described herein relates to an automotive radar system in which the processor is configured to process the range-Doppler matrix from the first sub-segment to detect a first candidate set of peaks using a constant false alarm rate (CFAR) peak detection algorithm.

[0009] In some respects, the technology described herein relates to an automotive radar system, wherein a first sub-segment of the radar cube and a second sub-segment of the radar cube together constitute a complete radar cube of the automotive radar system.

[0010] In some respects, the technology described herein relates to an automotive radar system in which the processor is configured to zero-padding the first sub-segment in the Doppler dimension before processing the first sub-segment to detect the first candidate peak set.

[0011] In some respects, the technology described herein relates to an automotive radar system, wherein the processor is configured to zero-padding the first sub-segment such that the first sub-segment includes the values ​​of all chirped signals encoded into the received radar signal.

[0012] In some respects, the technology described herein relates to an automotive radar system, wherein the automotive radar system is at least one of a frequency modulated continuous wave (FMCW) radar system and an orthogonal frequency division multiplexing (OFDM) radar system.

[0013] In some aspects, the technology described herein relates to an automotive radar system comprising: at least one transmitter and at least one receiver, wherein the at least one transmitter and the at least one receiver are configured to transmit and receive radar signals, wherein the at least one transmitter and the at least one receiver are coupled to a vehicle; and a processor configured to: receive received radar signals from the at least one receiver; generate a first sub-segment of a radar cube using the received radar signals; detect a first set of candidate peaks in the first sub-segment of the radar cube; generate a second sub-segment of the radar cube using the received radar signals; detect a second set of candidate peaks in the second sub-segment of the radar cube; determine a set of locations in the set of candidate peaks, wherein each location in the set of locations is associated with a candidate peak in either the first set of candidate peaks or the second set of candidate peaks; and estimate the direction of arrival of an object using the candidate peaks associated with the set of locations.

[0014] In some respects, the technology described herein relates to an automotive radar system, wherein the processor is configured to delete the first sub-segment from the memory of the automotive radar system after processing the first sub-segment.

[0015] In some respects, the technology described herein relates to an automotive radar system, wherein the processor is configured to detect the first candidate peak set using a constant false alarm rate (CFAR) peak detection algorithm.

[0016] In some respects, the technology described herein relates to an automotive radar system, wherein a first sub-segment of the radar cube and a second sub-segment of the radar cube together constitute a complete radar cube of the automotive radar system.

[0017] In some respects, the technology described herein relates to an automotive radar system in which the processor is configured to zero-padding the first sub-segment of the radar cube in the Doppler dimension before processing the first sub-segment to detect the first candidate peak set.

[0018] In some respects, the technology described herein relates to an automotive radar system, wherein the processor is configured to zero-padding the first sub-segment such that the first sub-segment includes the values ​​of all chirped signals encoded into the received radar signal.

[0019] In some respects, the technology described herein relates to an automotive radar system, wherein the automotive radar system is at least one of a frequency modulated continuous wave (FMCW) radar system and an orthogonal frequency division multiplexing (OFDM) radar system.

[0020] In some aspects, the technology described herein relates to a method comprising: receiving received radar signals from at least one receiver; processing the received radar signals to generate a first sub-segment of a radar cube; processing the first sub-segment of the radar cube to detect a first set of candidate peaks, wherein candidate peaks in the first set of candidate peaks are associated with positions in a range-Doppler matrix; processing the received radar signals to generate a second sub-segment of the radar cube; processing the second sub-segment of the radar cube to detect a second set of candidate peaks, wherein each candidate peak in the second set of candidate peaks is associated with a position in the range-Doppler matrix; combining the positions of the first set of candidate peaks and the positions of the second set of candidate peaks into a candidate peak dataset; using the candidate peak dataset to determine a set of positions in the candidate peak dataset, wherein each position in the set of positions is associated with a candidate peak in both the first set of candidate peaks and the second set of candidate peaks; and using the candidate peaks associated with the set of positions to estimate the direction of arrival of an object.

[0021] In some respects, the technology described herein relates to a method that also includes deleting the first sub-segment from the memory of the automotive radar system after processing the first sub-segment.

[0022] In some respects, the techniques described herein relate to a method that also includes processing the distance-Doppler matrix from the first sub-segment to detect a first candidate peak set using a constant false alarm rate (CFAR) peak detection algorithm.

[0023] In some respects, the technology described herein relates to a method in which the first sub-segment of the radar cube and the second sub-segment of the radar cube together constitute a complete radar cube of an automotive radar system.

[0024] In some respects, the techniques described herein relate to a method that further includes zero-padding the first sub-segment in the Doppler dimension prior to processing the first sub-segment to detect the first candidate peak set.

[0025] In some respects, the techniques described herein relate to a method that further includes zero-padding the first sub-segment such that the first sub-segment includes the values ​​of all chirped signals encoded into the received radar signal. Attached Figure Description

[0026] Considering the following figures, and with reference to the detailed description and claims to obtain a more complete understanding of the subject matter, similar reference numerals are used throughout the figures to refer to similar elements.

[0027] Figure 1A A simplified schematic block diagram depicts an automotive radar system, including a radar unit connected to a radar controller processor.

[0028] Figure 1B Graphically depicting what can be processed by a processor from Figure 1A The processing steps of digital signals received by the car radar system.

[0029] Figure 2 It is a diagram depicting a slice of radar cube data after range FFT and Doppler FFT processing.

[0030] Figure 3 This is a flowchart depicting a method for iteratively processing sub-segments of radar cube data, according to the present disclosure.

[0031] Figure 4 This is a diagram that provides a visual illustration of a method for generating candidate peak datasets.

[0032] Figure 5 It is a three-dimensional curve depicting the peak detected by conventional signal processing methods used for simulation.

[0033] Figure 6 It is a three-dimensional curve depicting the peak detected by a signal processing method used for simulation, according to this disclosure. Detailed Implementation

[0034] The following detailed description is illustrative in nature only and is not intended to limit the embodiments of the subject matter of this application or the use of such embodiments. The terms “exemplary” and “example” as used herein mean “serving as an example, instance, or illustration.” Any embodiment or example described herein as exemplary or illustrative should not be construed as preferred or advantageous over other embodiments. Furthermore, one is not to be bound by any express or implied theory presented in the prior art, background art, or the following detailed description.

[0035] In the context of this disclosure, it should be understood that radar systems can be used as sensors in automotive radar sensors for road safety and vehicle control systems such as advanced driver assistance systems (ADAS) and automated driving (AD) systems.

[0036] For example, an automotive radar system can be implemented as a frequency-modulated continuous wave (FMCW) radar system. An FMCW radar system transmits frequency-modulated signals (chirps) and receives their echoes as reflections from nearby objects. After down-mixing the received signals to baseband, the resulting signal consists of several sine waves, each with a beat frequency proportional to the distance to a specific object. Within each sine wave, an additional phase term carries Doppler phase information for each object near the radar system. This Doppler phase typically changes slowly and encodes information about the relative velocity of the corresponding object.

[0037] FMCW radar signal processing attempts to identify the range and Doppler components of the received reflected signals from each nearby object (e.g., other vehicles, road signs) that generated the reflected signals. This processing involves arranging the sampled data values ​​of the received chirped signals into a range-Doppler matrix as several horizontal vectors. The row length of the matrix is ​​equal to the number of samples for each chirped signal (typically a power of two, e.g., 1024 values), and the column length of the range-Doppler matrix is ​​the number of measured chirps (typically also a power of two, e.g., 256 values). These two-dimensional matrices are generated for each receiving antenna in the radar system. Several two-dimensional matrices for each receiving antenna are arranged together in a three-dimensional matrix, the size of which is the number of samples x the number of chirps x the number of receiving antennas, called a "radar cube".

[0038] In traditional signal processing methods, received signal data is buffered until a complete radar cube is received. Once received, the entire radar cube is arranged in memory as a matrix with dimensions equal to the number of Doppler elements (N_Dbins) x the number of range elements (N_Rbins) x the number of receive antennas (receive_antenna). Once the complete radar cube is available, further processing steps are performed to process it to identify potential objects and their properties (e.g., direction of arrival and velocity). This processing initially involves a Fast Fourier Transform (FFT) performed on the range dimension of the radar cube (called "fast-time" FFT or "R-FFT"), and an FFT performed on the Doppler dimension of the radar cube (called "slow-time" FFT or "D-FFT"). Peak detection methods are performed on the entire three-dimensional dataset containing the data processed by the fast and slow-time FFTs. In modern automotive radar systems, this is a massive dataset compared to the typical memory resources of such systems.

[0039] Therefore, difficulties may arise in automotive radar applications, such as corner radar systems, where the radar system relies on expensive internal SRAM memory instead of cheaper DRAM due to constraints on the device's size. Given these memory limitations, providing sufficient memory resources to load the entire radar cube into memory to perform the aforementioned operations can be prohibitively expensive. In these cases, reducing the memory footprint of the radar signal processing subsystem can help reduce system cost and / or size.

[0040] This disclosure provides a radar signal processing method that can be implemented with reduced memory footprint. Specifically, in this radar signal processing method, radar signal processing (R-FFT, D-FFT, and peak detection) is decomposed into several (N) partial processing steps, progressively detecting actual peaks present in the radar cube spectrum. Each processing step processes subsequent and different portions of the incoming data that constitute a sub-segment of the complete radar cube dataset.

[0041] When a new portion (1 / N) of the total chirps becomes available, new processing steps (covering R-FFT, D-FFT, and peak detection) are performed on the data sub-segment (even before the entire radar cube is received). At this point, previously received and processed portions of the same radar cube data can be discarded, thereby reducing memory storage requirements.

[0042] Based on the available data in each processing step, each processing step in each sub-segment of the entire radar cube identifies a distinct set of candidate peaks within each sub-segment. Therefore, when all N processing steps have been executed, indicating that all sub-segments of a single radar cube dataset have been processed, the information regarding candidate peak detection in each sub-segment is combined. In this process, the coordinates of the identified candidate peaks are projected onto the same axis of the range-Doppler dimension, and peak information associated with the set of coordinate pairs is collected.

[0043] After combining the candidate peaks identified in each sub-segment into a single dataset for the entire radar cube, further analysis is performed using the final pass criteria described herein to determine whether each candidate peak associated with a specific coordinate within the radar cube is associated with a true spectral peak. In one embodiment, a counter records the number of times a candidate peak at a specific coordinate is detected in each sub-segment of the radar cube. If the counter indicates that the same candidate peak is detected at the same coordinate in more than a threshold number of sub-segments of the radar cube, this may indicate that the candidate peak is a true spectral peak within the entire radar cube. After identifying the true spectral peak, the peak location can be identified as the output of this processing system. The peak can then be further analyzed to perform arrival direction analysis of objects near the vehicle.

[0044] As described herein, because this disclosure envisions iteratively processing sub-segments of the entire radar cube, the D-FFT operation performed on each sub-segment can be extended with zero-padding to allow D-FFT processing on the same number of samples, equal to the total number of chirped signals in the radar system. This zero-padding enables uniform coordinate axis processing across all sub-segments of the radar cube. This zero-padding results in broadening of the spectral energy and reducing the processing gain in the Doppler dimension of the radar cube, which may weaken any existing spectral peaks, potentially leading to peak detection loss. To mitigate this, threshold-based peak detection in each processing step of each sub-segment of the radar cube can be relaxed to detect weaker candidate peaks. In this case, a final pass criterion can be selected to filter out identified candidate peaks that are actually noise during the processing of the radar cube sub-segments but have passed the relaxed peak detection criterion.

[0045] To illustrate the design and operation of a vehicle radar system, refer to the following: Figure 1A , Figure 1A A simplified schematic block diagram of an automotive radar system 100, including a radar device 10 connected to a radar controller processor 20, is depicted. In selected embodiments, the radar device 10 may be embodied as a field-replaceable unit (LRU) or a modular component designed for quick replacement at the operating location. Similarly, the radar controller processor 20 may be embodied as a field-replaceable unit (LRU) or a modular component. Although a single or monostatic radar device 10 is shown, it should be understood that additional distributed radar devices can be used to form a distributed or multistatic radar system. Furthermore, the depicted radar system 100 may be implemented as an integrated circuit, wherein, depending on the application, the device 10 and the radar controller processor 20 may be formed using separate integrated circuits (chips) or a single chip.

[0046] Within the radar system 100, each radar device 10 includes one or more transmitting antenna elements 102 and receiving antenna elements 104, respectively connected to one or more radio frequency (RF) transmitter (TX) units 11 and receiver (RX) units 12. For example, each radar device (e.g., 10) is shown as including separate antenna elements 102, 104 (e.g., TX1,i, RX1,j) connected to three transmitter modules (e.g., 11) and four receiver modules (e.g., 12), but these numbers are not limiting, and other numbers are possible, such as four transmitter modules 11 and six receiver modules 12, or a single transmitter module 11 and / or a single receiver module 12.

[0047] Each radar unit 10 also includes a chirp generator 112, which is configured and connected to provide a chirp input signal to the transmitter module 11. For this purpose, the chirp generator 112 is connected to receive individual and independent local oscillator (LO) signals and chirp start trigger signals. The operation of the transmitter module 11 can be controlled by a controller 110, which can be implemented wholly or partially by the processor 20. The chirp signal 113 is generated and transmitted to the transmitter module 11, typically following a predefined transmission schedule, wherein the chirp signal 113 is filtered at the RF conditioning module 114 and amplified at the power amplifier 115, and then fed to the corresponding transmit antenna 102 (TX1,i) and radiated.

[0048] Radar signals transmitted by transmitter antenna elements 102 (TX1,i, TX2,i) can be reflected by objects, and a portion of the reflected radar signal reaches receiver antenna elements 104 (RX1,i) at radar device 10. At each receiver module 12, the received (RF) antenna signal is amplified by a low-noise amplifier (LNA) 120 and then fed to a mixer 121, where the received signal is mixed with a transmit chirp signal generated by an RF conditioning module 114. The resulting intermediate frequency (IF) signal is fed to a first high-pass filter (HPF) 122. The resulting filtered signal is fed to a first variable gain amplifier 123, which amplifies the signal and then feeds it to a first low-pass filter (LPF) 124. This re-filtered signal is fed to an analog-to-digital converter (ADC) 125 and output as a digital signal 126 (D1) by each receiver module 12. The receiver modules compress object echoes with various delays into multiple sinusoidal subcarriers whose frequencies correspond to the round-trip delay of the echoes.

[0049] The radar system 100 also includes a radar controller processing unit 20, which is connected to supply input control signals (e.g., via controller 110) to the radar device 10 and receive digital output signals (e.g., digital signal 126) generated by the receiver module 12.

[0050] In the selected embodiment, the radar controller processing unit 20 may be embodied as a microcontroller unit (MCU) or other processing units configured and arranged for signal processing tasks, such as, but not limited to, object recognition, object distance, object velocity and object orientation calculation, and control signal generation. For example, the radar controller processing unit 20 may be configured to generate calibration signals, receive data signals, receive sensor signals, generate spectrum shaping signals (such as ramp generation in the case of FMCW radar), and / or register programming or state machine signals for RF (radio frequency) circuit enable sequences. Additionally, the radar controller processor 20 may be configured to program the transmitter module 11 to operate in a time-division manner by sequentially transmitting chirps, thereby coordinating communication between the transmitting antenna elements 102TX1,i and RX1,j.

[0051] The radar controller processor 20 is configured to process digital signals 126 to ultimately identify the distance to objects and the angular positions of those objects relative to the radar system 100. The digital signals 126 comprise a sequence of digital values ​​representing the amplitude of radar signals captured by the receiving antenna elements 104 over time. Typically, each digital value is associated with a specific chirp number and sample number.

[0052] Figure 1A This diagram illustrates a series of signal processing steps implemented by processor 20 to properly process the digital signal 126 received from radar device 10 in order to identify potential nearby objects. To supplement... Figure 1A , Figure 1B The processing steps that can be implemented by the processor 20 to process the digital signal 126 are depicted graphically at a high level.

[0053] The content of digital signal 126 consists of a series of data frames including several digital sample values ​​(e.g., captured by ADC 125 of receiver unit 12), where the sample values ​​are arranged in a two-dimensional matrix generated based on a pulse signal sequence. The data structure that makes up a single capture frame is... Figure 1B This is depicted as matrix 150. As depicted, the single-frame data in matrix 150 comprises a two-dimensional matrix, the first dimension of which is called the "fast time" dimension, representing the data values ​​captured from different pulse signals. The second dimension of matrix 150 is called the "slow time" dimension, representing the data values ​​captured in response to different chirp signals that may be included in a specific pulse signal emitted by transmitter module 11. Figure 1BAs shown, signal processing may involve processing multiple data frames represented as several matrices 150. In this disclosure, signal processing may involve processing individual data frames, as described herein, to identify a set of peaks within the frame. Alternatively, as described herein, this may involve processing different sub-segments of the ADC data stream to identify candidate sets of peaks, which are ultimately combined across the entire radar cube to perform final peak detection. Typically, during such signal processing, data frames represented as matrices 150 are captured for each received channel. Therefore, Figure 1B Multiple matrices 150 are depicted, each associated with a different receiving channel and capable of serving as input data for signal processing chains.

[0054] For a sub-segment of radar cube data that may include all or part of one or more data represented as matrix 150, radar controller processor 20 first performs a fast time-range fast Fourier transform (FFT) 21. Figure 1A This generates a new frame of data represented as matrix 152. FFT 21 is performed on the array of 1-D data (i.e., signals) associated with each different chirp in the original input matrix 150 to generate 1-D transformed signals of the same length. The FFTs of each chirp in the original input frame, represented as matrix 150, are combined to generate the transformed frame, indicated by matrix 152. This process is repeated for each frame associated with each received channel. The resulting data frame representing the range map is... Figure 1B This is represented as matrix 152 and can be used to determine the distance to a specific object, as shown in the distance map.

[0055] In the next step, the radar controller processor 20 performs an additional Fast Fourier Transform (FFT) 22 on the range map. Figure 1A (This is referred to as slow-time or Doppler-FFT) to generate new range-Doppler frame data, represented as matrix 154. However, in this step, FFT 22 is applied along the dimension opposite to FFT 21. Therefore, FFT 22 is performed on the array of 1-D data (i.e., signals) associated with each range cell in matrix 152 to generate transformed 1-D signals of the same length. The FFTs of each signal in the frames of matrix 152 are combined to generate range-Doppler data frames, indicated by matrix 154. This process is repeated for each frame associated with each receive channel. The range-Doppler data frames associated with matrix 154 provide information about the movement of a potential object between different sample numbers over time. After the data frames associated with matrix 154 are generated, the data encoded therein can be processed to begin identifying potential objects, and, if an object is detected, its velocity and direction of arrival are determined.

[0056] Therefore, the radar controller processor 20 performs constant false alarm rate (CFAR) object detection 23. Figure 1A ), that is, 156 ( Figure 1B ).

[0057] If a potential object is detected, then the radar controller processor 20 executes the MIMO array measurement configuration 24. Figure 1A ), that is, 158 ( Figure 1B ), to determine each object 25 ( Figure 1A The direction of arrival (DOA) of 160 ( Figure 1B Next, the radar controller processor 20 transmits the final object information, which may include the object identifier (DOA) and other relevant information (in...). Figure 1A Step 26 Figure 1B (In step 162) to ADAS or other systems configured to use object information to control one or more vehicle systems.

[0058] In traditional radar systems, it is typically assumed that the entire radar cube dataset is available before processing. As discussed above, this approach requires significant memory resources to process the entire radar cube dataset in memory at each step of the algorithm. To reduce these requirements, this disclosure proposes a radar system and method configured to implement steps in radar signal processing (e.g., R-FFT, D-FFT, and peak detection steps), where each step is decomposed into N partial processing steps, in which sub-segments of the entire radar cube are iteratively processed, and after processing a specific data sub-segment, the data sub-segment can be discarded, thereby reducing memory storage requirements. In this method, peaks in the radar cube data are progressively detected as various data sub-segments are processed.

[0059] Specifically, within the radar system, radar cube data is sequentially made available as radar device 10 in the signal processing chain of the radar system outputs radar cube data in response to the analog radar signal received by the receiving antenna element 104 of radar device 10. Essentially, the radar cube data is generated as a data stream, which flows out as a digital signal 126 from radar device 100 to radar controller processor 20. In conventional methods, radar controller processor 20 stores the contents of the data stream in memory until the entire radar cube dataset is received. Only then does radar controller processor 20 begin processing the radar cube data.

[0060] However, in this system, the radar controller processor 20 can be configured to process radar cube data simultaneously with and before receiving the complete radar cube dataset. In this configuration, the radar controller processor 20 processes the streaming radar cube data into a series of N "chunks" or sub-segments, such that when the radar controller processor 20 receives a new sub-segment of the radar cube, a processing step is performed on said sub-segment of the radar cube dataset, including R-FFT, D-FFT, and peak detection steps. At this time, previously processed sub-segments of the same radar cube dataset can be discarded to free up memory for previously processed data.

[0061] This process is performed on each new sub-segment of the radar cube dataset as it becomes available, such that each processing step for each sub-segment outputs a set of different detected candidate peaks associated with each radar cube dataset sub-segment. After processing all the sub-segments of the entire radar cube dataset (i.e., when all N sub-segments of the radar cube dataset have been processed), information about the candidate peaks identified in each sub-segment of the radar cube is combined into a candidate peak dataset.

[0062] When generating the candidate peak dataset, data associated with candidate peaks identified in each subset are combined into a single dataset. This data may include the coordinates of each candidate peak and information associated with peaks for the same coordinate pair. In an embodiment, the candidate peak dataset includes the locations of candidate peaks identified from different sub-segments, where the locations are represented with respect to a hypothetical distance-Doppler matrix (e.g., with dimensions equal to the dimension of the entire radar cube). Figure 1BThe origin (row 1, column 1) of matrix 154 defines the two-dimensional coordinates. In this coordinate system, the size of the range dimension is equal to the number of samples per chirp, and any sub-segment can be constructed to fit this number. The size of the Doppler dimension is equal to the number of chirs in the entire radar cube. As described in this paper, to ensure that the coordinates associated with the candidate peaks identified in each sub-segment are equivalent, during sub-segment processing (and specifically, during D-FFT processing), the number of available chirs in each sub-segment can be expanded in the Doppler dimension (e.g., by zero-padding) such that the dimension of the expanded sub-segment being processed (e.g., the number of chirs per sub-segment) matches the dimension of the entire radar cube. This allows the range-Doppler spectrum of each sub-segment to be described using the same coordinate system. Furthermore, the contribution of each sub-segment to the range-Doppler spectrum of the (hypothetical) entire radar cube (never stored or computed) is also clearly defined. The candidate peak dataset can be generated using a general candidate peak detection algorithm. However, in some embodiments, peak values ​​at the spectral level can be aggregated before analyzing the dataset using the DoA algorithm to improve the SNR and bring it closer to the SNR obtained through conventional processing. If such aggregation is performed, normalization may not be necessary, but depending on the sub-segments, some phase correction of the phase term may be required. Phase correction may be needed to coherently combine candidate peak values ​​obtained from different sub-segments and achieve an increase in SNR.

[0063] After combining candidate peak data detected within each sub-segment of the radar cube dataset, a final pass criterion is applied to the candidate peak dataset to determine whether each candidate peak represents the true peak that should be output to the DOA estimation subroutine of the radar system. The final pass criterion may involve determining the number of times a candidate peak is found at the same location on the range-Doppler spectrum of one or more sub-segments of the radar cube dataset. The location of a peak can be defined by the offset (in cells) of the peak relative to the origin (row 1, column 1) of the range-Doppler matrix of the sub-segment. As described above, the range-Doppler matrix of each sub-segment has the same size due to zero-padding. If the number of times a candidate peak is found at the same location on the range-Doppler spectrum of each radar cube sub-segment exceeds a predefined threshold, then the location can be determined to be associated with the true peak used for object DOA estimation.

[0064] During radar cube sub-segment processing, because processing is performed only on sub-segments of the radar cube, the "slow-time" Doppler FFT operation can utilize zero-padding to increase the size of the sub-segment dataset, ensuring that the obtained Doppler spectrum has an appropriate size. Zero-padding is used to ensure that the radar cube sub-segment being processed includes the values ​​of all radar system chirps (potentially including true data values ​​or zero-padding values). This allows candidate peaks identified at locations within each radar cube sub-segment to be combined into a single dataset with uniform coordinate axes consistent across all identified candidate peaks. As described herein, such zero-padding leads to spectral energy broadening and reduced processing gain in the Doppler dimension, which can weaken any existing spectral peaks in the data being processed, potentially resulting in detection loss. As a mitigation, it is conceivable that threshold-based peak detection performed while processing each radar cube sub-segment can be relaxed to detect and identify weaker candidate peaks during the sub-segment processing phase. In this case, a criterion is ultimately responsible for detecting which weaker candidate peaks are most likely noise and which are most likely to reflect true peaks.

[0065] In radar processing systems, both range FFT and Doppler FFT operations concentrate most of the signal energy in their respective outputs around the input signal beat frequency and Doppler coordinates in the range-Doppler two-dimensional dataset generated as the output of the two FFT operations. Therefore, the outputs of the range FFT and Doppler FFT processing steps include spectral peaks that can be detected by the corresponding peak detection operations. Once peak detection is performed, the dataset containing the peak detection information is typically much smaller than the original radar cube dataset (e.g., usually one-tenth or one-hundredth of its size). Essentially, the detected peak dataset is a representation of the original radar cube, which is extremely sparse across all dimensions.

[0066] To illustrate, Figure 2 This is a diagram depicting slices of radar cube data after range FFT and Doppler FFT processing. The horizontal axis represents the Doppler dimension, while the vertical axis represents the reference signal amplitude (power) at the corresponding Doppler cell (horizontal axis) of the data slice. As depicted, the concentration of signal energy caused by the range FFT and Doppler FFT processing results in several energy peaks 202 concentrated at specific locations in the Doppler dimension (called Doppler cells). Each peak 202 indicates a signal reflection generated by an object near the radar system. The amplitude of each peak 202 is determined by radar equations based on object distance. In the peak detection dataset, only information associated with peak 202 is retained, while other data can be discarded as noise. This significantly reduces the data size compared to the original signal data.

[0067] Figure 3This is a flowchart depicting a method 300 for iteratively processing sub-segments of radar cube data according to the present disclosure. Method 300 can be implemented by a signal processor of a radar system. In an embodiment, method 300 is implemented by a radar controller processor 20 of radar system 100. Method 300 assumes the availability of a predetermined input value N, which defines the number of sub-segments of the input radar cube dataset to be processed. In various embodiments, N can be a relatively small number (e.g., 2 or 3). With N defined, at block 302, the processor initiates receiving radar cube data in the form of a data stream from a radar signal processing system (e.g., radar device 10 of radar system 100).

[0068] Upon receiving the radar cube data stream, at box 304, the processor determines whether data for the current sub-segment "i" of the radar cube dataset has been received (and stored in memory accessible to the processor or controller). For a given radar system, the sub-segment may include a subset of J chirps x N antennas out of the total number of system chirps (i.e., n chirps x N antennas) within the system's "radar cube". If not, the method returns to box 302, where the processor continues receiving the radar cube data stream.

[0069] After all data for the current sub-segment "i" has been received at box 304, method 300 moves to box 306, where peak detection is performed on the radar cube data contained within the current sub-segment "i". This step involves performing range FFT and Doppler FFT on the data of sub-segment "i". After performing range FFT on all chirps available in sub-segment "i", a "slow-time" Doppler FFT is performed on all range cells produced by the range FFT processing in the Doppler dimension. If sub-segment "i" does not contain data values ​​for all available chirps of the radar system, then the columns of the matrix produced by the range FFT processing can be padded with zero values ​​so that the columns have full size and contain values ​​for all chirps of the radar system (some zeroed out). Zero padding of the i-th sub-segment out of the N sub-segments is accomplished by pre-appending (i-1) zero tuples and appending (Ni) zero tuples, each tuple consisting of (total number of chirps) / N zeros. Peak detection is typically performed iteratively on one or more adjacent columns of the Doppler spectrum matrix. Once peak detection is complete for this smaller subset of columns of the matrix, the processed column data can be discarded.

[0070] Peak detection is then performed on the resulting distance-Doppler matrix to identify a set of candidate peaks. In one embodiment, peak detection may begin with a maximum search to accurately obtain the peak locations in the distance-Doppler matrix, followed by a variant of the CFAR algorithm (e.g., OS-CFAR).

[0071] After box 306 is completed, the peak detected for the current sub-segment "i" is stored in memory as a candidate peak (requiring only a relatively small amount of data storage). The input and intermediate data for sub-segment "i" can be discarded, thereby reducing the memory storage requirements of method 300. The data associated with the candidate peak is typically the distance-Doppler spectrum value at and within a 3x3 range around the peak, as well as some metadata including the peak location.

[0072] At box 308, it is determined whether there are any additional sub-segments to process in the current radar cube dataset. If so, the process moves to box 310, where the value of "i" is incremented. The method then returns to box 304 to determine if all data for the sub-segment has been received. If not, method 300 returns to box 302 to continue receiving the radar cube data stream until all data for the new sub-segment has been received.

[0073] However, if the radar cube has no more sub-segments to process, indicating that the candidate peak set for all sub-segments of the current radar cube dataset has been determined, then the method moves to box 312, where the candidate peak data identified for each sub-segment is processed (e.g., by determining the number of times a candidate peak is identified at each location within the dataset). For example, box 312 may involve combining information about the candidate peaks with matching locations into a form that supports a final pass decision (at box 314 below). In one embodiment, the peak detection data associated with a location is an implicit binary decision (yes / no), and the aggregation process may be a simple counting of such decisions across all sub-segments at a given location. In another embodiment, the (complex) spectral values ​​of candidate peaks matching the same location are considered, and the aggregation may include phase correction and coherently combining these values ​​across all sub-segments.

[0074] Figure 4 This is a diagram providing a visual illustration of a method for generating candidate peak datasets. Figure 4 In this diagram, a specific radar cube is subdivided into four sub-segments (i.e., N=4). Element 402 describes the candidate peak processing for the second sub-segment 404 of the entire radar cube. Element 406 describes the candidate peak processing for the third sub-segment 408 of the entire radar cube. The processing for the first and fourth sub-segments is not described.

[0075] As shown, candidate peak detections in each sub-segment 404 and 408 (and the first and fourth sub-segments not shown) are combined into a candidate peak dataset 410, thereby enabling the formation of a final pass analysis on the combined dataset. In an embodiment, the candidate peak dataset 410 identifies the locations where several candidate peaks are detected in the respective sub-segments of the radar cube (e.g., sub-segments 404 and 408). Figure 4As depicted herein, sub-segments 404 and 408 both include candidate peaks located at the same position 412 within a single candidate peak dataset 410. Given a final pass criterion (as described herein), position 412 is determined to be associated with a true peak because it is associated with candidate peaks occurring two or more times (or another threshold number of times). Conversely, position 414 in the single candidate peak dataset 410 is associated with candidate peaks that exist only within a single sub-segment—sub-segment 404. Because position 414 is associated with candidate peaks only within a single radar cube sub-segment, it is determined to be associated with a false peak and can be discarded or ignored, as described herein.

[0076] Return to Figure 3 At box 314, the final pass criteria are applied to the data in the candidate peak dataset to identify so-called “true peaks.” In one embodiment, the final pass criteria involve determining whether the number of times a particular candidate peak appears at the same location in the candidate peak dataset within the range-Doppler 2D grid exceeds a predetermined threshold. Therefore, at box 314, each location in the candidate peak dataset is analyzed to determine the number of candidate peaks detected at that location. If the number exceeds the threshold, then that location can be reported as a true peak location. This set of true peak locations is reported at box 316 and, as described herein, can be used to perform DOA estimation for objects near the radar system.

[0077] Therefore, method 300 implements this iterative processing by dividing the iterative processing of the radar cube dataset into processing of sub-segments with reduced storage requirements. Since the candidate peak detection algorithm (i.e., the algorithm associated with box 306) is applied to each sub-segment of the radar cube, and specifically, the Doppler FFT is applied to only one sub-segment of the entire radar cube, specifically 1 / N of the total useful signal samples, the processing gain of the Doppler FFT may only be 1 / N of that of conventional methods that process all radar cube data.

[0078] By processing sub-segments of the radar cube in this manner, the spectral energy around candidate peaks in the range-Doppler spectrum of the sub-segment may be broadened due to the application of a shorter window. This broadening of useful spectral energy in the Doppler dimension can present several challenges, as it may reduce the signal-to-noise ratio (SNR) for further signal processing. Therefore, there is a risk that if conventional peak detection methods are used on the range-Doppler spectrum of the sub-segment, some true peaks may be discarded or undetected if their amplitudes are below or close to the signal thresholds of the peak detection algorithm, which are typically defined as signals above the noise level. Thus, this broadening of spectral energy weakens peak detection in individual radar cube sub-segments, potentially causing signal peaks that would be detected when processing the entire radar cube dataset to go undetected.

[0079] To mitigate these difficulties, this method provides, at block 306 of method 300, the use of a peak detection algorithm with a relaxed threshold to identify more peaks during the sub-segment processing stage, thereby ensuring that lower amplitude peaks with amplitudes closer to the noise level in the sub-segment distance-Doppler spectrum are included in the candidate peak set. In one embodiment, this involves, for example, using a peak detection algorithm at the sub-segment stage, such as configured with an ordered statistical constant false alarm rate (OS-CFAR) with relaxed tolerance, to detect peaks weaker than normal peaks (described in more detail below). By configuring the peak detection algorithm to detect weaker peaks, the risk of lower amplitude peaks going undetected due to reduced processing gain can be mitigated.

[0080] The result of the slack tolerance of the peak detection algorithm implemented in conjunction with block 306 of method 300 is that some noise may be identified as candidate peaks by the peak detection algorithm. However, these noisy (or “false”) peaks can be filtered out using a final pass criterion, which is the final processing step described in this paper prior to DOA estimation.

[0081] As the number of sub-segments N used when processing radar cube datasets increases, peak detection algorithms can become increasingly lenient in detecting more candidate peaks. Therefore, the degree of leniency can depend on the number of segments "N," because as N increases, the processing gain decreases, and the candidate peaks that may collectively form the true peaks become weaker.

[0082] To illustrate the potential memory footprint reduction advantage of this radar cubes dataset processing method, a comparison of current sub-segment-based radar cubes processing methods with conventional methods is now described. Both conventional and proposed methods enhance their performance primarily through the following industry-known measures: 1) Windowing can lead to wide spectral peaks, which can mislead conventional OS-CFAR algorithms because many points within and around the true peak will meet the OS-CFAR test conditions, resulting in multiple false detections of a single strong peak. Furthermore, and specifically, for the proposed method, peak detection is performed using 1 / N of the available chirp, which in turn results in spectral peaks that are N times wider in the Doppler dimension. To avoid the problem of multiple false detections, a maximum filter is needed that can filter out any non-maximum points within the slope drop or rise (existing around strong spectral peaks). This can be achieved by combining a 4x4 cell 2D maximum filter with OS-CFAR. The maximum filter is performed on the range-Doppler spectral power data. The OS-CFAR check is evaluated at the output points produced by the 2D maximum filter.

[0083] The windowing applied to sub-segment samples in both the "fast time" and "slow time" dimensions shapes the energy distribution around the range-Doppler spectral peaks. Windows with sharp, narrow spectral main lobes also have strong side lobes, which stand out strongly against background noise when the SNR is high. These side lobes can be filtered by a maximum value filter and verified after using the OS-CFAR algorithm, which may lead to spurious peak detection for each side lobe. Conversely, windows with relatively wide spectral widths typically have low side lobes, which may not cause this problem. When the SNR is low, the problem is exactly the opposite: windows with wide spectral widths propagate the energy of weak peaks that cannot stand out against noise. Therefore, a Chebyshev window (wide spectral width, main lobe to side lobe ratio of -100dB) can be chosen for short-range signal processing (distance < 50% of the maximum support distance), while a sinusoidal window (narrow spectral width, main lobe to side lobe ratio of -50dB) can be used for longer distances. In long-range scenarios, the window has high sidelobes, but since the window is applied to a low SNR region, these sidelobes may be below the background noise and may not interfere with the maximum filter or OS-CFAR algorithm. By taking these measures, the reference processing flow can even detect the weakest peaks in testing. When comparing this method with conventional methods, the same range-based windowing method and the same peak detection method (a combination of maximum filter and OS-CFAR) are used. However, according to this method, the OS-CFAR threshold is reduced to detect candidate peaks that may be ignored in individual radar cube sub-segments. When the OS-CFAR threshold is reduced, the power threshold of the algorithm may be reduced compared to the nominal value. In a specific implementation of OS-CFAR, the spectral component must exceed a specific threshold power difference over the background noise to become a detectable peak. The threshold power difference is configured as a function of the background noise distribution, the desired target false detection rate, and the structural parameters of the algorithm (e.g., the window size used). In the relaxed peak detection criterion, the threshold power difference is reduced relative to the same threshold in the non-relaxed OS-CFAR. The relaxation ratio (the ratio of the relaxation threshold to the nominal threshold) can be determined through simulation, with the aim of achieving a balance between detection sensitivity and excessive noise detection at the end of the entire process (after finally passing the standard).

[0084] As discussed above, relaxing peak detection algorithms can lead to the misidentification of noise peaks as candidate peaks during processing of individual sub-segments. Therefore, the final dataset applied to the combined candidate peak dataset is configured to detect these false peaks, thereby enabling the removal of false peaks from the final dataset.

[0085] As described herein, the information contained in the complete radar cube is also contained (distributed) across N sub-segments. In this method of processing radar cube data sub-segments, each sub-segment of the radar cube is discarded once processed. Therefore, there is a risk of information loss. For example, a valid but weak peak may only appear after the third sub-segment of the radar cube data has been processed. In this case, data associated with the same peak that should have appeared in the first and second sub-segments may be inaccessible because these sub-segments may have been discarded. To minimize the risk of peak data loss, this disclosure envisions that additional data can be stored during the processing of method 300 and retrieved later if necessary. Specifically, during sub-segment processing, regardless of whether a peak is detected at the current sub-segment location, information relating to: 1) a hard decision for each sub-segment given a peak: a "hit" counter, and 2) 3x3 neighborhood data surrounding the peak if the peak was detected at a specific location ("hit") in the previous sub-segment. It should be noted that, in order to minimize storage requirements, 3x3 snapshots of the same location can be accumulated (averaged or weighted).

[0086] In this disclosure, a final criterion is used as a counter to track the number of times candidate peaks appear at the same location throughout the candidate peak dataset. Candidate peaks caused by noise and unrelated to the true peaks are unlikely to repeat at the same coordinates in a radar cube sub-segment. Therefore, if this countering method detects two or more peaks at the same location throughout the candidate peak data, then it can be determined that the location contains the true peak data. Because some true peaks do not fall exactly in a single cell (i.e., integer coordinate pairs of the range-Doppler spectrum matrix), but rather the energy of the peak is split into two cells, setting the counter threshold to 3 or 4 may be too stringent. Therefore, in various embodiments, the location associated with two or more peaks is designated as the true peak according to the final criterion.

[0087] The simulation results of this radar signal processing method are now compared with those produced by conventional processing. The simulation results are 256 chimes x 512 samples per chime, with N=4. The final pass standard for this simulation sets the true peak detection threshold to 2. During radar cube sub-segment processing, a relaxed OS-CFAR peak detection algorithm is used, where the relaxation parameter (the ratio of the relaxed power threshold to the nominal power threshold) is set to 0.33.

[0088] Figure 5 It is a three-dimensional curve depicting the peak detected by conventional signal processing methods used in this simulation. Figure 5In this diagram, the horizontal axis represents range cells, the depth axis represents Doppler cells, and the vertical axis represents indicator signals. A value of "1" is assigned at each detected peak, and a value of "0" is assigned at all other locations; therefore, vertical line 502 represents the detected peak. In this method, all beat frequencies are detected in the radar cube data (represented by peaks 502 with rounded tips). This method does indeed introduce some false alarm peaks (represented by peaks 502 with forked tips).

[0089] Figure 6 This is a three-dimensional curve depicting the peak value detected by the signal processing method used in this simulation. Figure 6 In the diagram, the horizontal axis represents range cells, the depth axis represents Doppler cells, and the vertical axis represents the indicator signal. A value of "1" is assigned at each location where the final peak is detected, and a value of "0" is assigned at other locations; therefore, vertical line 602 represents the detected peak. As depicted, this radar signal processing method can detect most of the beat frequencies of the input radar cube data, but in this example, some beat frequencies are missed when the SNR drops below 5 dB (the missed peaks are represented by circles 604 at the expected locations). Therefore, in this simulation, this signal processing method introduces fewer false alarms than conventional processing. Thus, both methods achieve the same detection probability (i.e., 100%) for nearby objects and under high SNR conditions.

[0090] In embodiments of this system, a radar signal processor (e.g., radar controller processor 20 of radar system 100) is coupled to a circular buffer configured to store data of the radar cube sub-segment being processed (typically, approximately 1 / N the size of the entire radar cube). In a specific embodiment, the size of the circular buffer is set to store the current radar cube sub-segment being processed, plus an additional buffer (approximately 1 / 8 the size of the entire radar cube) for storing data transmitted via the data stream (e.g., during processing of the current sub-segment). Figure 3 Other radar cube input data received at frame 302. The memory system accessed by the radar signal processor includes a small buffer (typically about 1 / 16 the size of the entire radar cube) allocated for storing candidate and final peak data. In this case, each candidate or true peak is stored as a cell with the peak location and its 3x3 neighborhood in a 2D grid. Therefore, in various embodiments, the total memory requirement of this radar signal processing method is approximately half the size of the entire radar cube. Thus, this processing method provides sufficient object detection while reducing memory storage requirements, which is typically manifested as a reduction in the need for area-expensive SRAM.

[0091] Various examples of this radar signal processing method have been presented for FMCW radar systems. However, it should be understood that this radar cube sub-segment processing method can be used in conjunction with other types of radar systems, such as those implementing Orthogonal Frequency Division Multiplexing (OFDM). In this case, the radar cube may include a three-dimensional dataset comprising K (predetermined values) x radar channel transfer function x N antennas. OFDM-based radar signal processing can be implemented according to this disclosure as follows: First, for the transmitted radar signal, a set of (arbitrary) spectral components on an array of length N is defined. The array of length N is transformed into the time domain by applying an inverse FFT. This is repeated K times to construct a matrix with K x N values. Then, the values ​​defined by the K x N matrix are converted into analog values ​​by serialization, and the resulting analog signal is up-converted to radio frequency to transmit the values ​​as the transmitted radar signal to the radio channel.

[0092] Next, at the receiver side, the reflections of the transmitted signals are received (e.g., via a receiving antenna). The received signals are then down-converted. After down-conversion, the upcoming samples are packaged into a dataset of K arrays of length N to generate a radar cube dataset of size K x N times the number of receiving antennas. Then, each array of length N in the radar cube is restored to the spectral domain by applying an FFT operation, and the radio channel transfer function is estimated for each array of length N. That is, since the transmitted spectral components are known, the system can estimate the transfer function primarily by dividing the received spectrum by the expected spectrum. The obtained radio channel transfer function is then transformed to the time domain by an inverse FFT (IFFT), resulting in the impulse response of the communication channel, equivalent to the radar delay distribution (echo). This result is similar to (and very similar in quantity to) the range processing output of FMCW radar after the range FFT. By compiling the impulse responses of the K arrays of length N into a K x N matrix and performing another FFT in the remaining dimensions, a range-Doppler spectrum of size K x N, similar to that of FMCW radar, can be obtained. Such distance-Doppler spectra have the same peak values ​​in the distance-Doppler dimension and have the same or similar signal sparsity.

[0093] Therefore, referring to this disclosure, in such OFDM radar schemes, the value N will be equal to the FMCW parameter "chirp length" (range dimension), and the value K will be equal to the FMCW parameter "chirp number" (Doppler dimension). After all processing is complete, the resulting radar cube spectrum resembles the FMCW spectrum. Therefore, in OFDM schemes, the teachings of this disclosure can be applied to iteratively process the OFDM radar cube, and then the intermediate results can be combined to obtain the final peak value.

[0094] Although examples have been described with reference to automotive radar systems, the systems and methods described herein can be implemented in conjunction with other types of radar systems. Devices or components described as separate can be integrated into a single physical device. Furthermore, units and circuits can be suitably combined in one or more semiconductor devices. That is, the devices described herein can be implemented as a single integrated circuit or as multiple integrated circuits.

[0095] The foregoing detailed description is illustrative in nature only and is not intended to limit the subject matter or the use of embodiments of this application and such embodiments.

[0096] As used herein, the term “exemplary” means “serving as an example, instance, or illustration.” Any embodiment described herein as exemplary should not be construed as preferred or advantageous over other embodiments. Furthermore, one is not to be bound by any express or implied theory presented in prior art, background art, or detailed description.

[0097] The connecting lines shown in the figures contained herein are intended to represent exemplary functional relationships and / or physical couplings between various elements. It should be noted that many alternative or additional functional relationships or physical connections may exist in embodiments of the subject matter. Furthermore, certain terms may be used herein for reference only, and therefore are not intended to be limiting, and unless the context clearly indicates otherwise, the terms “first,” “second,” and other such numerical terms referring to structures do not imply order or sequence.

[0098] As used herein, "node" means any internal or external reference point, connection point, junction, signal line, conductive element, etc., where a given signal, logic level, voltage, data pattern, current, or quantity exists. Furthermore, two or more nodes can be implemented with a single physical element (and although receiving or outputting at a common node, two or more signals can still be multiplexed, modulated, or otherwise distinguished).

[0099] The above description refers to elements, nodes, or features being "connected" or "coupled" together. As used herein, unless otherwise explicitly stated, "connected" means that one element is directly engaged to (or directly communicates with) another element, and not necessarily mechanically. Similarly, unless otherwise explicitly stated, "coupled" means that one element is directly or indirectly engaged to (or directly or indirectly communicates with) another element electrically or otherwise, and not necessarily mechanically. Therefore, although the schematic diagrams shown depict an exemplary arrangement of elements, additional intervening elements, means, features, or components may be present in embodiments of the depicted subject matter.

[0100] While at least one exemplary embodiment has been presented in the foregoing detailed description, it should be understood that numerous variations exist. It should also be understood that the one or more exemplary embodiments described herein are not intended to limit the scope, applicability, or configuration of the claimed subject matter in any way. In fact, the foregoing detailed description will provide a convenient guide for those skilled in the art to implement the one or more described embodiments. It should be understood that various changes can be made to the function and arrangement of the elements without departing from the scope defined by the claims, which includes known and foreseeable equivalents at the time of filing this patent application.

Claims

1. An automotive radar system, characterized by comprising: at least one transmitter and at least one receiver, wherein the at least one transmitter and the at least one receiver are configured to transmit and receive radar signals, wherein the at least one transmitter and the at least one receiver are coupled to a vehicle; and a processor configured to: receive received radar signals from the at least one receiver, process the received radar signals to generate a first sub-section of a radar cube, process the first sub-section of the radar cube to detect a first set of candidate peaks, wherein a candidate peak in the first set of candidate peaks is associated with a location in a range-doppler matrix, process the received radar signals to generate a second sub-section of the radar cube, process the second sub-section of the radar cube to detect a second set of candidate peaks, wherein each candidate peak in the second set of candidate peaks is associated with a location in the range-doppler matrix, combine the locations of the first set of candidate peaks and the locations of the second set of candidate peaks into a set of candidate peak data, determine a set of locations in the set of candidate peak data using the candidate peak data, wherein each location in the set of locations is associated with a candidate peak in both the first set of candidate peaks and the second set of candidate peaks, and estimate a direction of arrival of an object using the candidate peaks associated with the set of locations.

2. The automotive radar system of claim 1, wherein the processor is configured to process the range-doppler matrix from the first sub-section to detect a first set of candidate peaks using a constant false alarm rate (CFAR) peak detection algorithm.

3. The automotive radar system of claim 1, wherein, the processor is configured to zero-pad the first sub-section in the Doppler dimension before processing the first sub-section to detect the first set of candidate peaks.

4. The automotive radar system of claim 3, wherein, the processor is configured to zero-pad the first sub-section such that the first sub-section includes values for all chirp signals encoded into the received radar signals.

5. An automotive radar system, characterized by comprising: at least one transmitter and at least one receiver, wherein the at least one transmitter and the at least one receiver are configured to transmit and receive radar signals, wherein the at least one transmitter and the at least one receiver are coupled to a vehicle; and a processor configured to: receive received radar signals from the at least one receiver, generate a first sub-section of a radar cube using the received radar signals, detect a first set of candidate peaks in the first sub-section of the radar cube, generate a second sub-section of the radar cube using the received radar signals, detect a second set of candidate peaks in the second sub-section of the radar cube, determine a set of locations in a set of candidate peak data, wherein each location in the set of locations is associated with a candidate peak in both the first set of candidate peaks and the second set of candidate peaks, and estimate a direction of arrival of an object using the candidate peaks associated with the set of locations.

6. The automotive radar system of claim 5, wherein, the processor is configured to zero-pad the first sub-section of the radar cube in the Doppler dimension before processing the first sub-section to detect the first set of candidate peaks.

7. The automotive radar system of claim 6, wherein, The processor is configured to zero-pad the first sub-section such that the first sub-section includes values for all chirp signals encoded into the received radar signal.

8. A method characterized by, Comprising: receiving a received radar signal from at least one receiver, processing the received radar signal to generate a first sub-section of a radar cube, processing the first sub-section of the radar cube to detect a first set of candidate peaks, wherein a candidate peak in the first set of candidate peaks is associated with a location in a range-Doppler matrix, processing the received radar signal to generate a second sub-section of the radar cube, processing the second sub-section of the radar cube to detect a second set of candidate peaks, wherein each candidate peak in the second set of candidate peaks is associated with a location in the range-Doppler matrix, combining the locations of the first set of candidate peaks and the locations of the second set of candidate peaks into a set of candidate peak data, determining a set of locations in the set of candidate peak data using the set of candidate peak data, wherein each location in the set of locations is associated with a candidate peak in both the first set of candidate peaks and the second set of candidate peaks, and estimating a direction of arrival of an object using the candidate peaks associated with the set of locations.

9. The method of claim 8, wherein, Also comprising zero-padding the first sub-section in the Doppler dimension prior to processing the first sub-section to detect the first set of candidate peaks.

10. The method of claim 9, wherein, Also comprising zero-padding the first sub-section such that the first sub-section includes values for all chirp signals encoded into the received radar signal.

Images