Target detection and point cloud output method, integrated circuit, sensor, terminal equipment

Abstract

Embodiments of the present application relate to the art of target detection and disclose a target detection and point cloud output method, an integrated circuit, a sensor, and terminal equipment. The target detection method includes the steps of acquiring first data, generating second data based on a plurality of the first data, detecting a first target based on the first data, and detecting a second target based on the second data. This is advantageous for improving the accuracy of radar-based target detection.

Description

Technical Field

[0001] This application is based on Chinese patent applications with application numbers "202311127963.3", filing date September 1, 2023, application number "202311127857.5", filing date September 1, 2023, application number "202311198067.6", filing date September 15, 2023, and application number "202410678539.6", filing date May 28, 2024, and claims their priorities. All of its content is incorporated herein by reference.

[0002] Embodiments of this application relate to the technical field of target detection, and in particular, to target detection and point cloud output methods, integrated circuits, sensors, and terminal devices.

Background Art

[0003] A radar is an electronic device that uses electromagnetic waves to detect targets. Its operating principle is that the radar system first emits electromagnetic waves through an antenna, and after these electromagnetic waves hit the target and are reflected back, the radar system receives the echo signal reflected by the target, processes and analyzes the echo signal, and determines information such as the distance, rate of change of distance (radial velocity), azimuth, and altitude of the target.

[0004] However, currently, the accuracy of target detection by radar systems still needs to be improved.

Summary of the Invention

[0005] The target detection and point cloud output methods, integrated circuits, sensors, and terminal devices according to the embodiments of this application are advantageous for improving the accuracy of target detection by radar.

[0006] In other embodiments of the present application, a target detection method according to a first embodiment of the present application includes the steps of: acquiring first data; generating second data based on a plurality of the first data; detecting a first target based on the first data; and detecting a second target based on the second data.

[0007] In other embodiments of the present application, a target detection method according to a second embodiment of the present application includes the steps of: acquiring a plurality of targets based on a target detection result; acquiring characteristic information for each detected target, including noise characteristics and signal characteristics that characterize the target; and determining the confidence level of each target based on the characteristic information.

[0008] In other embodiments of the present application, a target detection method according to a third embodiment of the present application includes the steps of: acquiring a first target and a second target based on the target detection method described in the first embodiment; acquiring characteristic information for each of the detected first target and the second target, including noise characteristics and signal characteristics that characterize the target; and determining the confidence level of each of the first target and each of the second target based on the characteristic information.

[0009] In other embodiments of the present application, a target point cloud output method according to a fourth embodiment of the present application includes the steps of acquiring a first target and a second target based on the target detection method described in the first embodiment or the target detection method described in the third embodiment, and outputting the first target and the second target in a hierarchical manner.

[0010] In other embodiments of the present application, a fifth embodiment of the present application further provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs a target detection method or a target point cloud output method described in any embodiment of the present application.

[0011] In other embodiments of the present application, an integrated circuit according to a sixth embodiment of the present application includes sequentially connected radio frequency modules, analog signal processing modules, and digital signal processing modules, wherein the radio frequency module generates a radio frequency transmission signal and receives a radio frequency reception signal; the analog signal processing module performs frequency reduction processing on the radio frequency reception signal to obtain an intermediate frequency signal; the digital signal processing module performs analog-to-digital conversion on the intermediate frequency signal to obtain digital signal data; and based on the digital signal data, the target detection method described in the first embodiment, the target detection method described in the second embodiment, the target detection method described in the third embodiment, or the target point cloud output method described in the fourth embodiment is implemented.

[0012] In other embodiments of the present application, and in a sensor according to a seventh embodiment of the present application, a sensor includes a carrier, an integrated circuit according to the sixth embodiment provided on the carrier, and an antenna provided on the carrier or integrated with the integrated circuit into a single component provided on the carrier, wherein the integrated circuit is connected to the antenna and processes echo signals received by the antenna.

[0013] In other embodiments of the present application, and in the eighth aspect of the embodiments of the present application, the terminal device includes a device body and a sensor provided on the device body and described in any embodiment of the present application, the sensor being used for target detection and / or communication to provide reference information for the operation of the device body.

[0014] In the technical means according to the embodiment of the present invention, when detecting a second target, the second data for detecting the second target is generated from a plurality of first data, and the plurality of first data corresponds to the observation time length of the plurality of first data. That is, the second target has a longer observation time length than the first target. In this way, it is possible to detect targets that are difficult to detect in a short time and avoid missing the second target. At the same time, since the detection of the first target can be achieved in a short time, the present invention can also achieve timely and efficient detection of the first target. Therefore, the present invention can realize hierarchical detection for different targets, adapt to the characteristics of different targets, and not only improve the accuracy of detection but also ensure the timeliness of detection.

[0015] This invention allows for the identification of the reliability of each detected target based on its characteristic information after the target has been detected. This enables the target to undergo further verification, which is advantageous in improving the reliability and accuracy of target detection. [Brief explanation of the drawing]

[0016] One or more embodiments are illustrated by the corresponding drawings, and these illustrative descriptions are not limiting to the embodiments. Elements having the same reference numerals in the drawings are shown as similar elements, and unless otherwise specified, the drawings are not limiting to proportions.

[0017] [Figure 1] This is flowchart 1 of the target detection method according to an embodiment of the present invention. [Figure 2] This is flowchart 2 of the target detection method according to an embodiment of the present invention. [Figure 3] This is flowchart 3 of the target detection method according to an embodiment of the present invention. [Figure 4] This is flowchart 4 of the target detection method according to an embodiment of the present invention. [Figure 5]This is the flowchart 5 of the target detection method according to the embodiment of the present application. [Figure 6] This is the flowchart 6 of the target detection method according to the embodiment of the present application. [Figure 7] This is the flowchart 7 of the target detection method according to the embodiment of the present application. [Figure 8] This is the flowchart 8 of the target detection method according to the embodiment of the present application. [Figure 9] This is the flowchart 9 of the target detection method according to the embodiment of the present application. [Figure 10] This is the flowchart of the target point cloud output method according to the embodiment of the present application. [Figure 11] This is the schematic configuration diagram of the integrated circuit according to the embodiment of the present application. [Figure 12] This is the schematic diagram of the signal waveform according to the embodiment of the present application.

Embodiments for Carrying out the Invention

[0018] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below with reference to the drawings. However, those skilled in the art can understand that in each embodiment of the present application, many technical details are presented to help readers better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical means for which protection of the present application is sought can be realized.

[0019] The division of the following embodiments is for ease of explanation and should not constitute any limitation to the specific implementation manner of the present application. Each embodiment can be cited in combination with each other as long as there is no contradiction.

[0020] To facilitate those skilled in the art to better understand the target detection and point cloud output methods, media, circuits, sensors, and devices according to the embodiments of the present application, the process of target detection by a conventional radar will be described first below. Specifically, the process is as follows.

[0021] Taking frequency-modulated continuous wave (FMCW) radar as an example, when performing target detection, a chirp signal is transmitted for each frame, and each transmitted signal of each frame contains a plurality of chirp signals. Then, digital signal processing is performed on the echo corresponding to the signal of each frame, and a radar point cloud set is output in a manner of outputting for each frame. The processing process of the echo digital signal of each frame is as follows.

[0022] 1. Perform fast Fourier transform (FFT) within the chirp to obtain the original channel data.

[0023] 2. Calculate the average value of the original channel data of all chirps within the same frame to obtain the zero-Doppler channel data of the frame.

[0024] 3. Subtract the zero-Doppler channel data from the original channel data to obtain the updated channel data.

[0025] 4. Perform in-frame FFT on the updated channel data to obtain range-Doppler (RD) spectrum data.

[0026] 5. Perform coherent processing (or non-correlation processing) and constant false-alarm rate (CFAR) detection on the RD spectrum data to obtain the target.

[0027] 6. Extract the range unit (range bin) and Doppler unit (Doppler bin) for the detected target and perform digital beamforming (DBF) to obtain the DBF spectrum for each target.

[0028] 7. For each target, find the maximum value of the DBF spectrum, calculate the azimuth angle corresponding to the maximum value, and obtain the target's azimuth angle.

[0029] However, the above method, which performs target detection and output on a frame-by-frame basis, cannot be applied to all targets when facing various types of targets. For example, at the time of detection, there may be dynamic targets whose movement is immediately perceived, such as a child running in a room, and there may also be micro-movement targets whose movement is difficult to perceive, such as a person lying in bed with subtle movements like breathing. In the case of dynamic targets, their movement speed and the resulting displacement are generally large, so sufficient detection information to detect changes in the target is accumulated within one frame period, and in this case, the radar can output relatively accurate detection results for dynamic targets. However, in the case of micro-movement targets, their movement speed and the resulting displacement are generally small, so there is a high possibility that sufficient detection information to detect changes in the target will not be accumulated within one frame period, and echo signals formed by reflection from the target are often mixed in with noise. As a result, detecting such micro-movement targets is extremely difficult, and in this case, it is difficult for the radar to output accurate detection results for micro-movement targets.

[0030] To address the above problem, this invention provides different observation times adaptable to different targets, thereby enabling detection of micro-movement targets that do not have clear movement characteristics by accumulating sufficient positional changes for observation over a relatively long time interval. This achieves the objective of enabling more comprehensive detection for different types of targets and more comprehensive sensing and rendering of the scene. At the same time, for dynamic targets with clear movement characteristics, output can be generated efficiently and in a timely manner based on positional changes accumulated over a relatively short time interval.

[0031] To achieve the above target, the target detection method according to the embodiment of the present invention detects different types of targets in a targeted manner. In other embodiments, as shown in Figure 1, the target detection method is Step 101 to obtain the first data, Step 102 involves generating second data based on multiple first data, Step 103 involves detecting a first target based on the first data, The process includes step 104, which involves detecting a second target based on second data.

[0032] Each first data point corresponds to an observation time length T. Detecting a second target using second data generated from multiple first data points is equivalent to detecting a second target using second data acquired within multiple observation time lengths T. This allows for the accumulation of sufficient target features for observation over longer observation times, enabling the detection of second targets that are difficult to detect in short periods, thus preventing missed detections. Simultaneously, first targets with clear movement characteristics do not require the accumulation of multiple first data points for detection; detection can be achieved using a single first data point. Therefore, this invention not only enables timely and efficient detection of first targets but also accurately, reliably, and comprehensively detects second targets, providing hierarchical detection for different targets and guaranteeing accuracy and timeliness for different types of target detection.

[0033] The embodiments of this application do not limit the specific value of T, and it can be set in combination with actual needs, characteristics of the application scene, target characteristics, algorithm characteristics, etc. For example, in some cases, the value of T is the time length of one frame period. In some cases, the movement characteristics of a dynamic target may be very pronounced, and sufficient target information can be accumulated to observe within a time of less than one frame period. In this case, T may be a smaller value, for example, a time length of 0.5 frames, a time length of 1 / 3 of a frame, etc. In some cases, the algorithm needs to acquire more data so that the accuracy of target detection meets the requirements. In this case, T may be a larger value, for example, a time length of 1.5 frames, a time length of 2 frames, a time length of 5 frames, etc. In some cases, the scene may require a more real-time output. If the target detection algorithm supports this, T may be a smaller value, for example, a time length of 0.2 frames, a time length of 0.5 frames, etc. Naturally, the above is merely an explanation of T using "frames" as the unit of reference. In some cases, T may refer to other periods, such as the time length of a chirp period, for example, the time length of multiple consecutive chirp periods, which will not be explained here.

[0034] When detecting human activity in relatively enclosed scenes such as indoors, conventional radar target detection cannot accurately acquire movement information such as velocity and acceleration of the point cloud of a target that only has respiratory activity or a slightly moving human body that is unconsciously swaying (e.g., a seated human body). However, by introducing second data as described in the above embodiment, this invention realizes point cloud output of a second target whose movement characteristics are not clear, such as a target that only has respiratory movement or a slightly moving human body that is unconsciously swaying. This increases the amount of data that the second target relies on when performing target detection against the first target, and by extending the observation time, the second target can have sufficient displacement within the observation time and acquire a point cloud of a slightly moving target.

[0035] This embodiment does not limit the distinction between the first and second targets. In some cases, the first target may be a dynamic target, and the second target may be a micro-movement target. A dynamic target may be a target with a movement speed exceeding V1, and a micro-movement target may be a target with a speed less than V2, and so on. The specific values ​​of V1 and V2 may be determined according to the application scene, the actual situation of the target, etc., and will not be listed here. In some cases, the first target may be a large target, and the second target may be a small target. In some cases, based on the application scene, the first and second targets can be obtained by combining empirical data, big data analysis or clustering analysis, movement trajectories, etc., to divide the target into different movement states. In some scenes, the first target may be defined as a target detected when a person or car is moving relatively fast overall in the target area, and the second target may be defined as another target with a movement speed slower than the first target. In other cases, a type division operation may be performed on dynamic targets by combining dimensions such as movement speed and movement width. Specifically, this can be set by comprehensively considering the application scene, actual needs, and radar monitoring performance, and it is sufficient if a distinction between the first target and the second target can be achieved, or if a hierarchical display can be achieved subsequently. For the sake of explanation, when providing illustrative explanations below, the first target will be a dynamic target and the second target will be a micro-movement target as an example.

[0036] Furthermore, during implementation, there may be additional needs for target detection. For example, if it is desired during implementation that the data obtained by processing one frame of echo signal (i.e., the echo signal for one frame) be treated as one type of first data, if it is desired during implementation that the detection results be more reliable, another type of first data can be set as the echo signal for two frames, and each time target detection is performed, the data obtained by processing the echo signal for two frames can be acquired as another type of first data. In this way, more data can be used during target detection, which is advantageous in ensuring that target detection has higher accuracy and reliability. Also, if it is desired during implementation that the detection results be fed back in real time by the user or other equipment, one type of first data may be set as the data obtained by processing the echo signal for 0.5 frames. If the data of a given unit includes a predetermined number of adjacent chirp data within one frame, the above embodiment can be realized by setting the data obtained by processing the echo signal based on a different number of adjacent chirp data as one type of first data, i.e., the period of the signal processing unit corresponding to different types of data should be different. Thus, not only can processing be performed on echo signals within a shorter time duration, but target detection can also be performed more quickly on the acquired echo signals, and the results of target detection can be fed back. This is advantageous in improving the real-time nature of the feedback of target detection results to the user or other devices, but the explanation is omitted here.

[0037] To facilitate a better understanding of the target detection method according to the above embodiment by those skilled in the art, it will be described below.

[0038] In step 101, first data is acquired. This embodiment does not limit the meaning of the first data. To make it understandable, regarding the radar signal processing process, after different digital signal processing methods are completed, corresponding data is acquired, such as a digital signal acquired by sampling the echo signal, the aforementioned original channel data, RD spectral data, etc. The first data may be data acquired after any of the above processing, and if the data acquired after any of the above processing is the first data, the above effect can be achieved. However, if the meaning of the first data is different, the acquisition method will also be different. The aforementioned echo signal may be an echo signal generated by any type of transmitted signal being reflected by a target, such as an echo signal generated by a signal transmitted from any continuous wave (CW) radar having multiple receiving channels and / or multiple transmitting channels being reflected by a target, or an echo signal generated by a signal transmitted from any frequency modulated continuous wave (FMCW) radar having multiple receiving channels and / or multiple transmitting channels being reflected by a target.

[0039] For example, in some cases, the first data may be an initial digital signal. In this case, the first data may be obtained by converting the echo signal from analog to digital.

[0040] Furthermore, in some cases, the first data may be original channel data. In this case, the acquisition of the first data can be achieved by the step of acquiring the original channel data as described above.

[0041] Furthermore, in some cases, for example, the first data may be the updated channel data. In this case, the acquisition of the first data can be achieved by the step of acquiring the updated channel data as described above.

[0042] Furthermore, in some cases, for example, the first data may be RD spectral data. In this case, the acquisition of the first data can be achieved by the step of acquiring the RD spectral data described above.

[0043] The embodiments of this application do not limit the process for realizing the acquisition of the above data.

[0044] For example, in other embodiments, RD spectral data may be obtained by: performing a distance-dimensional Fourier transform on the echo signal to obtain frequency domain signals corresponding to at least two chirps; generating zero-Doppler channel data for each frame based on the frequency domain signals corresponding to chirps belonging to the same frame; removing the clutter frequency domain component from the frequency domain signal based on the corresponding zero-Doppler channel data to obtain the dynamic frequency domain signal for each chirp; and performing a slow time-dimensional Fourier transform on the data in the dynamic frequency domain signal corresponding to the same distance unit in each chirp, and generating RD spectral data based on the result of the transformation. Thus, RD spectral data is obtained by processing the echo signal to identify the zero-Doppler channel data of the signal for each frame, further identifying the dynamic frequency domain signal for each chirp, and then performing a slow time-dimensional Fourier transform. In other words, providing a processing method that adapts to the dynamic characteristics of a dynamic target to obtain RD spectral data is advantageous in improving the accuracy of indoor dynamic target detection.

[0045] Furthermore, in other embodiments, for example, RD spectral data may be obtained by: performing a distance-dimensional Fourier transform on the echo signal to obtain frequency-domain signals corresponding to at least two chirps; generating zero-Doppler channel data for each frame based on the frequency-domain signals corresponding to chirps belonging to the same frame; and performing a slow-speed time-dimensional Fourier transform on the frequency-domain signals corresponding to the zero-Doppler channel data corresponding to the same distance unit in each frame, and generating RD spectral data based on the transformed result. In this way, RD spectral data is obtained by identifying the zero-Doppler channel data for each frame by processing the echo signal, and then performing a Fourier transform on the zero-Doppler for each frame, i.e., an inter-frame Fourier transform. That is, providing a processing method that adapts to the micro-motion characteristics of a micro-motion target to obtain RD spectral data is advantageous in improving the accuracy of indoor static target detection.

[0046] In other embodiments, before performing a slow time-dimension Fourier transform, the zero-Doppler channel data is updated based on the zero-Doppler channel data of each frame, and the clutter frequency-domain component is removed from the frequency-domain signal based on the corresponding updated zero-Doppler channel data to obtain the frequency-domain signal after each chirp update. Accordingly, performing a slow time-dimension Fourier transform on the frequency-domain signal corresponding to the zero-Doppler channel data corresponding to the same distance unit in each frame can be achieved by performing a slow time-dimension Fourier transform on the updated frequency-domain signal corresponding to the zero-Doppler channel data corresponding to the same distance unit in each frame. In this way, by updating the zero-Doppler channel data based on data from multiple frames to account for the effects of long-term clutter, clutter removal becomes more thorough, which is advantageous in obtaining a more accurate micro-motion frequency-domain signal of the micro-motion target and further detecting the micro-motion target more accurately, i.e., further improving the accuracy of micro-motion target detection.

[0047] Here, each of the above types of processing can be implemented using the corresponding algorithm or process. For example, the generation of zero-Doppler channel data can be implemented using the Moving Target in Idcation (MTI) algorithm, the phasor mean offsetting algorithm, and the like.

[0048] In other embodiments, as shown in Figure 2, obtaining the first data can be achieved by the following steps 201 to 205.

[0049] In step 201, echo data corresponding to a detection signal having a predetermined transmission period of 50 ms (milliseconds) or longer is acquired.

[0050] In step 202, the first data is generated based on the echo data.

[0051] Steps 203 to 205 are almost identical to steps 102 to 104 of the embodiment shown in Figure 1, and therefore will not be explained here.

[0052] In this embodiment, the transmission period is extended, specifically increasing the transmission period from 20ms per frame used in conventional radar to 50ms or more per frame. When target detection is performed using frame data as the basic unit, it is ensured that the data within the frame has sufficient sensitivity to dynamic points. As the transmission period length increases, the velocity resolution is also improved. Furthermore, by outputting a point cloud of many first targets, an effect approximating rough imaging is achieved, and the space of the RD spectral data acquired by spectral analysis (e.g., 2DFFT (2D Fast Fourier Transform)) can be fully utilized.

[0053] This embodiment does not limit the method for extending the transmission period. As can be understood, since a detection signal usually includes multiple chirp signals, the method for extending the transmission period of the detection signal can be implemented based on the configuration of the chirp signals. In other examples, extending the transmission period may be implemented by increasing the number of chirp signals included; that is, a detection signal having a predetermined transmission period is obtained by increasing the number of chirp signals included in the detection signal for each frame. In other examples, extending the transmission period may be implemented by extending the period of each chirp signal; that is, a detection signal having a predetermined transmission period is obtained by extending the period of the included chirp signals, for example, by extending the time length A shown in Figure 12 (i.e., the idle time in the chirp signals). In other examples, extending the transmission period may be obtained by increasing the frame idle time between chirp signals in the detection signal; that is, a detection signal having a predetermined transmission period is obtained by increasing the idle time after a chirp signal in the detection signal, for example, by increasing the time length B shown in Figure 12. Figure 12 illustrates, using only the example of a triangular wave as the chirp signal in an FMCW waveform, where the rising edge appears first, followed by the falling edge. This does not mean that the embodiment of the present invention is applicable only to the FMCW waveform shown in Figure 12, and therefore, further explanation is omitted here.

[0054] Step 102 generates second data based on multiple first data. As mentioned above, the first data can have multiple different situations, and correspondingly, there can be multiple different implementation methods for generating the second data. To facilitate understanding, the following will illustrate this by using channel data and RD spectral data as the first data.

[0055] If the first data is channel data, and the channel data is obtained by subtracting zero-Doppler channel data from the original channel data, then in other embodiments, generating the second data based on multiple first data can be achieved by superimposing the multiple first data to obtain the second data. In this case, since the first data is channel data and no subsequent processing is performed between the first data, there is no need to perform complex calculations, and the second data can be obtained by superimposition, making implementation easier.

[0056] Considering that the second data is generated from multiple first data, the same interference may exist among the multiple data, that is, the same noise component exists among the multiple first data. Based on this, in other embodiments, as shown in Figure 3, generating the second data from multiple first data can be achieved by the following steps 301 to 305.

[0057] In step 302, multiple primary data points are averaged to obtain zero-Doppler channel data.

[0058] In step 303, the second data is obtained by subtracting zero-Doppler channel data from multiple first data points and then superimposing them.

[0059] Steps 301, 304, and 305 are substantially the same as steps 101, 103, and 104 in the previously described embodiment, and are therefore omitted from this explanation.

[0060] Removing the averaged zero-Doppler channel data is equivalent to removing the same noise component between different data. Therefore, in the above embodiment, noise component removal is performed twice on the echo signal. In the first case, the average value is calculated within the first data, so the noise component removed is within the first data. In the second case, the average value is calculated across multiple first data, i.e., the average value is calculated outside the first data, so the noise component removed is between multiple first data. This allows for better removal of different types of noise components from multiple first data, thereby enabling more effective extraction of signal components. In this way, more accurate target detection can be achieved based on more effective signal components, resulting in a better target detection effect.

[0061] To facilitate understanding, we will use frames as processing units in our example. In the first case, the data within a frame is averaged, and noise components within the frame are removed. In the second case, the data from multiple frames is averaged across frames to obtain the average value between frames, and the noise components between multiple frames are removed.

[0062] In some cases, the zero-Doppler channel data of multiple frames may be compared, anomalous data may be removed, and then the average value of the zero-Doppler channel data of the remaining frames may be used as the zero-Doppler channel data in step 302. This is advantageous in avoiding sudden noise interference and further improving the accuracy of the zero-Doppler channel data.

[0063] Thus, by updating the zero-Doppler channel data based on data from multiple frames, taking into account the effects of long-term clutter, clutter removal becomes more thorough. This allows for the acquisition of frequency-domain signals corresponding to more accurate micro-motion targets, and further enables more accurate detection of micro-motion targets, thus improving the accuracy of micro-motion target detection.

[0064] If the first data is RD spectral data, in other embodiments, generating second data based on multiple first data can be achieved by estimating the second data based on multiple first data. In this case, since the RD spectral data is further processed relative to the channel data and the relationships between the data become more complex, it is not possible to simply superimpose and obtain the second data, and some estimation is required. The estimation process is related to the FFT processing performed on the channel data, which will not be explained here. However, compared to the case where the first data is channel data as described above, if the first data is RD spectral data, the processing data in the detection of the first and second targets can be shared before the RD spectrum is obtained, thus simplifying the process and improving efficiency.

[0065] Naturally, the above is merely an illustrative explanation, and in other examples, other data may be constructed as the first data, etc., but such explanations are omitted here.

[0066] Furthermore, for clarity, in the above embodiment, targets are divided into two types: first targets and second targets. Considering that various micro-motion targets may exist in the indoor environment, these micro-motion targets may have different movement characteristics, and when different micro-motion targets are observed, the length of observation time that needs to be accumulated will also differ, and the amount of echo data used for target detection will reflect the length of observation time. For example, head tremors that occur when reading different parts of a page and body tremors when walking slowly indoors, the width and speed of walking slowly indoors are greater, so the length of observation time that needs to be accumulated when the latter is observed is shorter than the length of observation time that needs to be accumulated when the former is observed. Therefore, the number of echo signal frames required to detect slow walking indoors is less than the number of echo signal frames required to detect head tremors that occur when reading different parts of a page. In the target detection process, if the same frame of historical echo signals is used for all micro-motion targets based on the first echo signal, the detection feedback for some micro-motion targets will be delayed relative to their movement process (i.e., not real-time), and / or some micro-motion targets will not be detected (i.e., not accurate). Therefore, target classification can be further refined based on even more granular movement characteristics, and by providing more detailed target detection based on this more granular target classification, more accurate target detection can be achieved.

[0067] Based on this, and taking the further classification of the second target as an example, in other embodiments, as shown in Figure 4, generating second data based on multiple first data is possible. Step 402 involves obtaining i types of second data based on Ni first data, This can be achieved by step 403, which involves obtaining j types of second data based on Nj first data.

[0068] Accordingly, detecting a second target based on the second data is possible. Step 405 involves detecting a second target of type i based on a second data of type i, This can be achieved by step 406, which involves detecting a second target of type j based on a second data of type j.

[0069] i-type and j-type second data belong to the second data category, i and j are integers greater than 1, Ni and Nj are positive integers, and Nj > Ni.

[0070] Thus, by further subdividing the target and intentionally providing different numbers of first data for detection, i.e., providing different observation time lengths, and processing the movement information of the second target accumulated over different observation time lengths, it is possible to present the target detection results of different types of second targets from more time dimensions, more comprehensively, and more in real time, which is advantageous in further improving the real-time and accuracy of target detection.

[0071] Steps 401 and 404 in this embodiment are substantially the same as steps 101 and 103 in the previously described embodiment, so their explanation is omitted here.

[0072] Here, Nj and Ni are not limited and may be fixed values, or they may be values ​​that change based on information such as the total number of targets included in the target detection result of the previously detected first echo signal, the number of first and / or second targets, the velocity of the second target, and the acceleration of the second target. In some cases, even if Nj and Ni are fixed values, the specific numerical values ​​of Nj and Ni can still be set according to the needs, for example, according to the real-time requirement for target detection. As can be understood, if the real-time requirement is high, the values ​​of Nj and Ni cannot be too large, otherwise the amount of data that needs to be processed at the moment tends to increase, which leads to low processing efficiency and affects the target detection efficiency. If the movement range and velocity of the micro-movement target are too small, the values ​​of Nj and Ni cannot be too small, otherwise the micro-movement target may not be observed during target detection, and the problem of missing a target in target detection is likely to occur. Therefore, different numerical values ​​may be set for Nj and Ni depending on the application situation and the need for real-time performance in the application process.

[0073] The classification of different secondary targets may be set according to the needs. For example, the i-th type of secondary target may be defined as a target generated by breathing, heart rate, pulse, etc., when a person is at rest, and the j-th type of secondary target may be defined as a target generated by slow body movements when a person is at rest (for example, the slow, unconscious movement of the body when sleeping while sitting, or the displacement that occurs in the body when concentrating). Alternatively, different types of dynamic point clouds may be divided based on the movement velocity dimension, with secondary targets with a velocity greater than V3 being the i-th type of secondary target, and secondary targets with a velocity of V4 or greater being the j-th type of secondary target. Further explanation is omitted here.

[0074] Naturally, the above is merely an illustrative explanation provided for cases where the second target is further subdivided. In some cases, the first target may also be further subdivided, in which case the first data may be split to support the detection of different first targets, but this explanation is omitted here.

[0075] Step 103 involves detecting the first target based on the first data. As mentioned earlier, the first data can have multiple different possibilities, so the implementation method for target detection based on the first data in different situations will also differ. For example, if the first data is channel data, detecting the first target requires obtaining RD spectral data, but if the first data is RD spectral data, no processing is required to obtain RD spectral data, etc. Different implementation methods for detecting the first target for the first data in different situations will not be described here.

[0076] Furthermore, the FFT, CFAR, and other processes that may be involved in the process of detecting the first target are almost the same as those used in conventional target detection, and therefore will not be explained here.

[0077] In step 104, the second target is detected based on the second data. Similar to the detection of the first target described above, the second data is generated from multiple first data points, and therefore the second data will also differ depending on the first data point. Consequently, the detection of the second target will also differ, but the processing such as FFT and CFAR that may be involved in the target detection process is almost the same as in conventional target detection, and will not be explained here.

[0078] Furthermore, as can be understood, the second target is a target that needs to be observed for a longer period of time, and at this time, the target detection result can be improved by certain digital signal processing. For example, in another embodiment, as shown in Figure 5, detecting the second target based on the second data is possible. Step 504 involves performing a slow time-dimension FFT based on the second data, This can be achieved by step 505, which involves performing a certain level of false alarm detection based on the FFT results so that the detection of a second target is realized.

[0079] Steps 501 to 503 in Figure 5 are almost the same as steps 101 to 103 in the previously described embodiment, and therefore will not be explained here.

[0080] Thus, because a low-speed time-dimension FFT can distinguish between targets in different moving states, such as stationary targets and moving targets, the above embodiment achieves effective filtering of the first target using a low-speed time-dimension FFT and extracts an effective second target, thereby providing more accurate detection for the second target which needs to be observed for a long time, and achieving more accurate target detection results.

[0081] Naturally, when detecting the first target, a slow time-dimension FFT may be used to accurately extract the first target, which is advantageous for more accurate detection of the first target.

[0082] In other words, RD spectral data is acquired by a slow time-dimension FFT, and in practice, a processing method is provided that adapts to the dynamic characteristics of the corresponding first and / or second targets to acquire the RD spectral data, thereby improving the accuracy of the RD spectral data and further improving the accuracy of target detection performed based on said RD spectral data, that is, it is advantageous for improving the accuracy of first and / or second target detection.

[0083] From the FFT result to the detection of a certain level of false alarm, it is usually necessary to perform either coherent integration or non-coherent integration, thereby improving the signal-to-noise ratio by accumulating multiple pulses. Note that the embodiment of this application does not limit the specific implementation methods of coherent or non-coherent integration; coherent or non-coherent integration may be selectively performed depending on the application scene, user needs, etc., and the accumulation algorithm may be selected depending on computing power, resources, scene, and needs, etc., and such details are omitted here.

[0084] Furthermore, as can be understood, in typical CFAR algorithms, a high threshold is usually used, causing many target points to be considered non-targets. Based on this, in other embodiments, when target detection is performed, a certain level of false alarm detection can be performed based on a predetermined threshold, where threshold = k × estimated noise value, k > 1.

[0085] Currently, to reduce clutter interference, the threshold is set to k' × estimated noise value, and while the value of k' is usually in the tens or hundreds, here, by setting a restriction adjustment to k>1, it is possible to set a smaller threshold value when detecting a target, meaning the threshold does not necessarily have to be a high value. Setting a smaller threshold can reduce or avoid false alarm detections that detect many points and mistakenly exclude valid targets as noise points prematurely, which is advantageous in improving the accuracy of indoor target detection.

[0086] In some cases, the range of k values ​​is (1, 4), for example, k=2, k=2.4, k=2.5, k=3, etc. By setting the value of k to an even smaller range in this way, the threshold is further lowered, and a smaller threshold can further reduce or avoid the constant false alarm detection detecting more points and misidentifying and rejecting valid targets as noise points, which is advantageous in further improving the accuracy of indoor target detection.

[0087] Naturally, the above is merely an example of the range of values ​​for parameter k at the threshold, and the value of parameter k at the threshold may be adjusted depending on the specific scenario; however, we will omit further explanation here.

[0088] Furthermore, as can be understood, in relatively enclosed areas such as indoors, the accuracy of target detection is not high compared to relatively open environments such as outdoors, because people, obstacles, etc., are more complex. Therefore, it is necessary to improve the accuracy of target detection in such scenes. In addition, clutter is unavoidable in space, and clutter interferes with target detection, which may result in invalid targets being detected.

[0089] Therefore, the embodiments of the present invention provide a target detection method that not only avoids the problem of misidentification, where targets are removed as non-targets, by providing confidence in the detected targets, but also avoids the problem of not being able to accurately obtain target information. In other embodiments, the process of the target detection method is as shown in Figure 6, Step 601 involves acquiring multiple targets based on the target detection results, Step 602 involves obtaining characteristic information for each detected target, including corresponding noise features and signal features that characterize the target. This includes step 603, which identifies the confidence level of each target based on characteristic information.

[0090] Thus, since the feature information includes noise features and signal features that characterize the target, after detecting and acquiring a target, it is possible to determine whether the target is a valid target and to determine the confidence level of each detected target. This eliminates the error of detecting noise points as targets and allows for further verification of the targets, which is advantageous in improving the reliability and accuracy of target detection, and in particular in improving the accuracy of target detection indoors.

[0091] To facilitate a better understanding of the above embodiments by those skilled in the art, they will be described below.

[0092] In step 601, multiple targets are acquired based on the target detection results. This embodiment does not limit the method of acquiring the target detection results and can be implemented using any target detection method, which will not be explained here.

[0093] In step 602, for each detected target, feature information is obtained, including corresponding noise and signal features that characterize the target. In this embodiment, the feature information is not limited and may be any noise and signal features that can characterize the target. For example, in another example, the feature signal may be any information that can reflect the distribution of the target reflected signal in the received echo signal. For example, the feature information may be an angular spectrum, an entropy value, a signal-to-noise ratio, a feature value identified based on the covariance matrix of the echo signal, and the angular spectrum may be an RD spectrum, a Capon spectrum, etc. These will not be listed here. The angular spectrum can be obtained from the target detection process when obtaining the target detection result, and the algorithm used to generate the angular spectrum may include a beamforming algorithm, a Capon algorithm, a multiplexed-signal classification algorithm, a rotation-invariant subspace algorithm, an orthogonal matching tracking algorithm, etc., which will not be explained here. The entropy value can be obtained directly, and the signal-to-noise ratio can be obtained from the CFAR process when obtaining the target detection result.

[0094] In this context, the method for characterizing the target for signal features and noise features in the feature information is not limited and may vary depending on the feature information, scene, needs, etc.

[0095] For example, if the feature information includes an energy spectrum, the entropy value of the energy spectrum can be used as feature information based on the characteristic that the signal distribution is usually concentrated and the noise distribution is usually relatively scattered. If the feature information includes an angular spectrum, the relationship between the main lobe and side lobes in the angular spectrum can be used as feature information to characterize the target's signal and noise features, based on the characteristic that the signal energy is usually higher than usual and is mainly reflected in the main lobe, and the noise energy is usually lower than usual and is mainly reflected in the side lobes. Furthermore, if the feature information includes an angular spectrum, the feature information may also be related to the target's local noise features and local signal features, or to the target's global noise features and global signal features, but this will not be explained here.

[0096] To make it easier to understand, if the noise-related content in the feature information exceeds the signal-related content, it indicates a higher probability that the signal described by that feature information is a noise signal (i.e., clutter). Therefore, by determining whether the target is a valid target based on the signal and noise features of the target, and whether the target's feature information contains more noise-related content or more signal-related content, it is possible to determine whether the target is a valid target.

[0097] Furthermore, to make it easier to understand, the criteria for determining reliability differ depending on the different scenarios and needs. For example, if the noise in the environment is complex, there may be cases where the noise is large due to multiple superpositions. In such cases, the reference criteria for reliability can be set to broader conditions, thus reducing the risk of misjudging valid targets.

[0098] In step 603, the confidence level of each target is determined based on the feature information. As can be understood, clutter is unavoidable in space, and this clutter interferes with target detection, potentially resulting in the presence of invalid targets among the detected targets. Since the feature information includes noise and signal features that characterize the target, the target can be analyzed using the feature information to determine whether the target is a valid target or not, i.e., to determine the confidence level of the target.

[0099] As mentioned above, there are multiple possible characteristics, and accordingly, there are multiple ways to implement confidence levels. To facilitate understanding for those skilled in the art, the acquisition of target confidence levels will be explained below with examples, where the characteristic characteristics are angular spectrum, entropy value, and signal-to-noise ratio, respectively.

[0100] When the feature information is an angular spectrum, in other embodiments, as shown in Figure 7, the corresponding feature information can be obtained. This can be achieved by step 702, which involves acquiring the angular spectrum generated in the target detection process.

[0101] Accordingly, identifying the confidence level of each target based on characteristic information is Step 703 identifies the maximum value of the angular spectrum corresponding to the target, Step 704 identifies a first effectiveness parameter that characterizes the relative magnitude of the maximum value among the angular spectral maxima and the average value of the other maxima, based on the maximum value among the angular spectral maxima and the average value of the other maxima. This can be achieved by step 705, which identifies the confidence level of each target based on a first effectiveness parameter.

[0102] Step 701 is almost the same as the step in the previously described embodiment, and therefore will not be described here.

[0103] In this embodiment, the specific method for identifying the first effectiveness parameter is not limited. In other examples, the parameter may be the ratio of the maximum value among the maximum values ​​of the angular spectrum to the average value of the other maximum values, or in other examples, it may be an array consisting of the maximum value among the maximum values ​​of the angular spectrum and the average value of the other maximum values.

[0104] As mentioned above, since the maximum value among the angular spectrum maxima can correspond to signal-related content, and other maxima can correspond to noise-related content, the first effectiveness parameter can actually reflect whether the global signal distribution is concentrated or not, and further reflect the relative energy distribution. This allows for a more comprehensive and accurate determination of whether a target is an effective target based on the global relative energy distribution, that is, the reliability of the target can be accurately determined, which is advantageous in further improving the accuracy of target detection.

[0105] Note that the "mean" in the mean of the other maximum values ​​besides the maximum value may also refer to the mean, median, or mode, and we will omit the explanation here.

[0106] In other embodiments, as shown in Figure 8, determining the confidence level of each target based on a first effectiveness parameter can be achieved by the following steps 801 to 807.

[0107] In step 805, a second effectiveness parameter is identified that characterizes the relative magnitude of the area between the maximum and second largest values ​​of the angular spectrum maxima for the corresponding target, based on the maximum and second largest values ​​of the angular spectrum maxima.

[0108] In step 806, based on the second effectiveness parameter, it is determined whether the corresponding target belongs to an invalid target.

[0109] In step 807, if the corresponding target does not belong to an invalid target, the confidence level of the target is determined based on the first validity parameter.

[0110] Steps 801 to 804 are almost the same as steps 701 to 704 in the previously described embodiment, and therefore will not be explained here.

[0111] In this embodiment, the specific method for identifying the second effectiveness parameter is not limited. This parameter may be the ratio of the maximum value to the second largest value among the maximum values ​​of the angular spectrum, or it may be the difference between the maximum value to the second largest value among the maximum values ​​of the angular spectrum, etc.

[0112] As mentioned above, the maximum value among the angular spectrum maxima can correspond to signal-related content, and the other maxima can correspond to noise-related content. Therefore, the second effectiveness parameter actually reflects whether or not the local signal distribution is concentrated, and furthermore, the second effectiveness parameter can also reflect the relative energy distribution from a different angle, in addition to the first effectiveness parameter. This makes it advantageous to combine all relative energy distributions based on the local relative energy distribution to determine whether or not a target is an effective target from different dimensions, thereby further improving the accuracy of confidence and thus improving the accuracy of target detection.

[0113] In other embodiments, obtaining corresponding feature information can be achieved by obtaining the entropy value of the corresponding target. Accordingly, determining the confidence level of each target based on the feature information can be achieved by determining the confidence level of each target based on the entropy value.

[0114] Note that, without limiting the method for determining the entropy value, the entropy value can be obtained using any algorithm that can quantify the degree of congestion in the feature information, and therefore, the explanation is omitted here.

[0115] To understand this, valid targets (i.e., actual targets) should have a low entropy value because their corresponding echo signals should be concentrated, while invalid targets (i.e., false targets such as those detected based on clutter) should have a high entropy value because their corresponding echo signals are relatively less concentrated. In other words, the reliability of a target can be determined by identifying whether a detected target is a valid target or not based on the relative magnitude of its entropy value.

[0116] Note that the range of entropy values ​​corresponding to the target is not limited here and may vary depending on the degree of congestion in the detection environment, detection accuracy, and other requirements; therefore, a detailed explanation is omitted here.

[0117] In other embodiments, obtaining corresponding feature information can be achieved by obtaining the signal-to-noise ratio generated by the corresponding target through a certain false alarm detection in the target detection process. Accordingly, determining the reliability of each target based on the feature information can be achieved by determining the reliability of each target based on the signal-to-noise ratio.

[0118] To understand this, the signal-to-noise ratio is an accurate quantification of the signal and noise components in the feature information corresponding to a target. Based on this signal-to-noise ratio, the relative magnitudes of the signal and noise components in the feature information can be accurately determined, which allows for an accurate determination of whether the detected target is a valid target. This is advantageous for more accurately determining the reliability of the target, thereby achieving accurate target detection.

[0119] Naturally, the above embodiments may be combined, for example, in other embodiments, the target detection method according to the embodiment of the present application is as shown in Figure 9, Step 901 involves performing target detection and acquiring the first and second targets, Step 902 involves obtaining characteristic information for each of the detected first and second targets, including corresponding noise features and signal features that characterize the target. The process includes step 903, which identifies the confidence level of each first and second target based on characteristic information.

[0120] In this embodiment, step 901 is implemented by a target detection method capable of detecting the first and second targets as described above, and steps 902 and 903 are substantially the same as steps 602 to 603 in the previously described embodiment, and are therefore omitted from this explanation.

[0121] In other embodiments, the process of the target detection method according to the embodiment of the present application is as follows: Step 1 is a step of performing target detection by performing a certain false alarm detection on a Doppler spectral matrix based on a predetermined threshold, wherein the threshold is k × noise estimate and k > 1. Step 2 generates corresponding feature information for each detected target, This includes step 3, which involves identifying the confidence level of the target based on the corresponding feature information to further confirm whether the target is a valid target.

[0122] Steps 2 and 3 described above are almost the same as steps 602 to 603 in the previously mentioned embodiment, and therefore will not be explained here.

[0123] Step 1 described above may be replaced with any of the aforementioned solutions that can perform target detection and acquire the detected target. For example, an inter-frame FFT may be introduced in the process of generating RD spectral data, but this will not be explained here.

[0124] Furthermore, to make it clearer, in the above embodiment, a first target and a second target are detected for a given target movement state (i.e., point cloud type). That is, in the time dimension, within the same period, the corresponding target object can be divided into a first target and a second target, and can be further subdivided, for example, into a moving target, a short-motion target, and a long-motion target, depending on the generated movement state. In detecting a moving target, the target movement can be divided into a "moving" state and a "motion" state based on the amplitude of the target movement, that is, the amplitude of "moving" is greater than the amplitude of "motion", and at the same time, the "motion" state can be divided into at least two states, "long-motion" and "short-motion," based on the time dimension, and the time period of "long-motion" is greater than the time period of "short-motion".

[0125] However, conventional systems can generally output only one type of detection target based on predetermined frame data; for example, they can output only dynamic point clouds or micro-motion point clouds based on predetermined frame data. If a radar system needs to output two types of detection targets, they can output them alternately, i.e., one predetermined frame data outputs a dynamic point cloud, and the other predetermined frame data outputs a micro-motion point cloud. Simultaneously, micro-motion point cloud data is generally detected based on a single frame or multiple frames.

[0126] Therefore, embodiments of the present invention further provide a target point cloud output method that displays the above-mentioned different targets in a hierarchical correspondence in order to improve the user experience. In other embodiments, the process is as shown in Figure 10, Step 1001 involves detecting targets and identifying the first and second targets, The process includes step 1002, which outputs the first target and the second target in a hierarchical structure.

[0127] The first and second targets identified in step 1001 may be the first and second targets identified by the target detection method according to the above-described embodiment, or they may be the first and second targets with a certain level of reliability identified by the target detection method according to the above-described embodiment.

[0128] Regarding step 1002, in other embodiments, as described above, the target is subdivided into moving targets, short-motion targets, and long-motion targets depending on the generated movement state, and it is assumed that all three types of targets are acquired based on data from one predetermined group frames. That is, the data from the predetermined group frames can be used as one data unit for long-motion detection, at least two (e.g., two) frames included in the predetermined group frames can be used as one data unit for short-motion detection, and the data from each frame can be used for movement detection. Furthermore, for example, when performing long-motion detection using data from six consecutive frames as one data unit, the data from three consecutive frames can be selected as one data unit for short-motion detection, and the data from two consecutive frames can be selected as one data unit for movement detection. A frame does not necessarily refer to a single frame, but only to the smallest data unit for detecting a target. For example, the data from one frame may be 1 / 2 (or 1 / 3, 0.2, 1, 2.5, 3, 5, etc.) times the data of a conventional general frame.

[0129] Regarding step 1002, in other embodiments, when performing moving point cloud detection based on data from a single frame, point cloud data may be output for each frame corresponding to short, minute moving points using a sliding window method to further supplement the detected dynamic (moving) points. That is, by outputting both dynamic points and short, minute moving points for each frame, the overall posture and partial movements of a single human body can be depicted more accurately and comprehensively, achieving the effect of approximate imaging (imaging with relatively coarse granularity).

[0130] In other words, by the method according to the above embodiment, a scene can be sensed based on echo signals at different time scales, and a corresponding confidence level can be set for the point cloud according to predetermined rules, thereby realizing hierarchical output of point cloud data and fully enabling the display of the scene and detection of omnidirectional state information of the target. At the same time, at least two types of detection target data can be output simultaneously based on the echo signal of one predetermined unit, that is, at least two types of detection target results can be obtained based on the data of the same predetermined unit. The data of the predetermined unit may be a predetermined number of adjacent chirp echo signals within one frame (e.g., half frame, quarter frame, etc.), data from one frame, or echo signals from at least two adjacent frames. For example, if the data for a given unit consists of echo signals from three adjacent frames, that is, if the data from those three adjacent frames is used as the data source to acquire three different types of dynamic point clouds, then, for example, a first type of dynamic point cloud (e.g., short-term micro-motion point cloud) can be output based on the data from two adjacent frames, and a second type of dynamic point cloud (e.g., moving point cloud) can be output based on the data from each frame. Based on this, a third type of dynamic point cloud (e.g., long-term micro-motion point cloud) can be output based on the data from three adjacent frames. Simultaneously, by setting different colors, shapes, etc., in the point cloud display, a hierarchical display of the three types of point cloud data can be achieved.

[0131] To facilitate a better understanding of the methods and combinations thereof according to the above embodiments for those skilled in the art, the following will be explained with examples. For the sake of ease of understanding, the first target will be a moving target, the second target a micro-movement target, and the second target will be further subdivided into a long-duration micro-movement target and a short-duration micro-movement target.

[0132] A multi-channel 4D FMCW millimeter-wave radar mounted in a room (e.g., on the roof or wall) emits electromagnetic waves to a target in the room in units of frames. The signal in each frame may contain multiple chirp signals, and the frame period or the total transmission time of the intra-frame chirps is longer than that used in conventional radars; for example, the total transmission time of the intra-frame chirps can be increased from 20 ms to 50 ms or more. Simultaneously, the radar receives echo signals from the electromagnetic waves emitted in units of frames and performs the following processing on the received signals.

[0133] In step S1101, distance window processing and distance FFT are performed on each chirp signal within the frame to obtain the original channel data.

[0134] In step S1102, the zero-Doppler channel data of the signal in the frame is calculated and stored based on the original channel data in the frame.

[0135] In step S1103, zero-Doppler channel data is subtracted from each chirp in the frame according to the distance bin.

[0136] In step S1104, windowing between chirps and FFT are performed on the data acquired in step S1103.

[0137] In step S1105, the data acquired in all channel steps S1104 is modulo calculated and the results are accumulated.

[0138] In step S1106, target detection is performed on the results from step S1105 using the CFAR algorithm.

[0139] In step S1107, the distance bin and Doppler bin of all azimuthal channels corresponding to the target detected in step S1106 are extracted, DBF is performed, and modulo calculation is performed to obtain the DBF spectrum of the target.

[0140] In step S1108, the maximum value is determined for the DBF spectrum of each target, and the azimuth angle corresponding to the maximum value is calculated as the azimuth angle of the target.

[0141] In step S1109, the confidence level of the target is calculated based on the concentration of the DBF spectrum, and the output of the dynamic point is obtained.

[0142] In step S1110, a single micro-motion frame is constructed using data acquired in step 1103, which is nearly four frames long, and the processing in steps S1104 to S1109 is performed to acquire a short-time micro-motion point.

[0143] Four frames may be replaced with a time length of one second. The time length of one second can be adaptively adjusted according to actual needs and is not limited to the embodiments of this application, and may be 0.8 seconds, 0.5 seconds, or 0.3 seconds, etc.

[0144] In step S1111, a single micro-motion frame is constructed using data acquired in step 1103, which is nearly 16 frames long, and the processing in steps S1104 to S1109 is performed to acquire long-duration micro-motion points.

[0145] The 16 frames may be replaced with a time duration longer than 1 second. The specific time duration can be adaptively adjusted according to actual needs and is not limited in the embodiments of this application, but may be 5 seconds, 3 seconds, or 1 second, etc.

[0146] Step S1112 outputs the dynamic point, short-term micro-motion point, long-term micro-motion point, and their confidence levels. The dynamic point, short-term micro-motion point, and long-term micro-motion point are displayed hierarchically so that the user can observe them more intuitively.

[0147] In other embodiments, instead of recalculating the RD matrix for long-term micro-move points based on the above-described embodiment of the target detection method, the RD matrix for short-term micro-move points can be obtained by cumulatively estimating it. The RD matrix for short-term micro-move points can be obtained using a sliding window method, and output for each frame can be realized. Simultaneously, the RD matrix for long-term micro-move points may be obtained using a sliding window method, or it may be obtained by processing once at predetermined frame intervals, or the RD matrix for long-term micro-move points may be obtained based on the accumulation of the RD matrices for short-term micro-move points. When obtaining the RD matrix for long-term micro-move points based on the accumulation of the RD matrices for short-term micro-move points, phase compensation may be performed on the frame corresponding to each short-term micro-move point RD matrix, and then the RD matrix for long-term micro-move points may be obtained by accumulation.

[0148] In other embodiments, different frame rates can be used to sense scenes at different time scales. For example, the radar can be set to at least two scenes, such as a daytime (or active or living room) scene or a nighttime (or inactive or bedroom) scene, and the divided states can be set based on different time scales corresponding to the different scenes; that is, different scenes and different time scales for the same state. For example, in a daytime living room scene, the time scale for short-range motion can be set to 0.8s and the time scale for long-range motion can be set to 3-5s. In a nighttime bedroom scene, the time scale for short-range motion can be set to 1-2s and the time scale for long-range motion can be set to 10-100s, etc. This allows for adaptive adjustment to the scene and effectively reduces the radar's power consumption.

[0149] In this way, after detecting all possible potential target points in the RD spectral data using a constant false-alarm rate (Noise Reference Constant False-Alarm Rate, NRCFAR) based on extremely low threshold noise, possible target points can be selected from a large number of candidate points using DBF spectroscopy. The difference from conventional RD processing is that the formation of the final point cloud depends not on how the RD spectral data is detected, but on how the point cloud is selected.

[0150] In other words, conventional RD processing involves performing 2DFFT to form RD spectral data, then detecting target points from the RD spectral data based on CFAR, and determining their orientation and pitch angle to form a point cloud. However, the indoor environment is complex and contains a large number of reflective objects (desks, chairs, etc.), leading to a large number of multi-angle echoes at the same distance (reflections from many scenes) and multipath phenomena. That is, in conventional RD processing detection points, some are targets of interest, some are multipath points, and some are false points that have been reflected and synthesized from multiple sources. Therefore, conventional RD processing results have a phenomenon of having many noise points. On the other hand, in order to suppress noise points, conventional RD processing methods use a high detection threshold, and at the same time, suppress points from the actual target with a relatively weak signal-to-noise ratio, resulting in fewer points in the point cloud of the actual target (when combined with noise points, this shows a phenomenon where points are dispersed and not concentrated), or there are no points at targets with a low signal-to-noise ratio.

[0151] For points from actual targets of interest, since the signal component is single, the form of its DBF spectrum is always close to a sinc function, i.e., a clear principal peak exists. For multipath points, because there is a difference between the exit phase and the incident phase, after multiple-input multiple-output (MIMO) processing, the DBF spectrum of a multipath point cannot form a principal peak similar to that of the actual target, and the energy of the DBF spectrum is relatively dispersed. Similarly, for false points distributed in a room, which are reflected and combined from multiple sources, their DBF spectra also cannot form a clear principal peak. Therefore, candidate points can be selected using the DBF spectrum.

[0152] In indoor applications, human targets differ from typical radar targets in that the velocities of different body parts vary, meaning that human body echoes have a rich velocity component. At the same time, human body movement states are diverse, including tangential movement and low radial velocity states such as slight swaying, which pose a challenge for radars that perform target separation depending on radial velocity. On the other hand, the radial velocity of a point cloud of a human body only reflects the radial velocity of a particular body part and may differ significantly from the radial component of the actual movement velocity of the human body. In other words, in this scenario, a clear velocity range does not substantially affect target detection, meaning that the overall accuracy of target detection can be effectively improved by performing high-resolution velocity detection on indoor targets using a long chirp period.

[0153] Furthermore, improving velocity detection resolution has at least the following advantages: 1) to 4).

[0154] 1) By rendering the target with a finer velocity dimension, the signal-to-noise ratio of each velocity unit can be effectively improved.

[0155] 2) By distinguishing between echoes of the same distance from two targets in terms of velocity dimension, simultaneous detection of two targets at the same distance can be achieved.

[0156] 3) The radar's sensitivity to the scene is improved, that is, its ability to detect low radial velocity conditions such as tangential movement is enhanced, giving it high practical value in indoor scenes.

[0157] 4) Make full use of the number of points in the velocity dimension of RD spectral data to avoid wasting memory space.

[0158] Simultaneously, the detection of micro-motion targets is processed using intra-frame zero-Doppler channel data from multiple frames. Because the above dynamic point processing method allows for more appropriate detection of dynamic points based on RD spectral data, the same processing concept can be applied to micro-motion targets. Inter-frame FFT is performed on the intra-frame zero-Doppler channels from multiple frames to form inter-frame RD spectral data, and based on this RD spectral data, the above dynamic point processing method is used to form the final micro-motion point group.

[0159] In other embodiments, to more completely sense the scene, the micro-motion processing is divided into short-time micro-motion processing and long-time micro-motion processing. Short-time micro-motion point processing can use a small number of frames, and compared to in-frame processing, it can effectively detect seated and stationary human bodies. Short-time micro-motion points effectively capture dynamic points within frames, and the two are complementary in real-time scene sensing, allowing human posture and movement to be depicted together by dynamic points and short-time micro-motion points. The number of frames used for long-time micro-motion point processing is larger, allowing for the detection of finer scene changes and assisting short-time micro-motion points in detecting static human bodies.

[0160] In other words, steps a to d are executed in the dynamic point detection process.

[0161] a. Process signals acquired at different time scales to fully sense the motion targets in the scene and simultaneously output dynamic points, short-term micro-motion points, and long-term micro-motion points.

[0162] b. In dynamic point processing, a high-resolution velocity waveform is used to perform more detailed rendering relative to the velocity dimension of the scene, thereby expanding the "pool" that accommodates the target point cloud.

[0163] c) A low-threshold NRCFAR is used to detect in the RD region of the scene echo, detecting "bright spots" in all RD regions and ensuring that no possible target echo points are missed.

[0164] d. The candidate point set detected using DBF spectra is evaluated, and point clouds matching the target features are selected from the candidate point set and output as the final point cloud.

[0165] The above improvements make it possible to achieve the following 1-5.

[0166] 1. It can output a rich point cloud for a human body target, depict the posture of the human body in dynamic and static states, and perform point cloud tracking (coarse imaging) for human body movements (e.g., waving hands).

[0167] 2. Filter out and remove a large number of noise and multipath points that are widely present in the indoor environment, while retaining effective small target points, thereby improving the radar's ability to detect small targets.

[0168] 3. By improving the point cloud cohesion of human body targets, the resolution between human body targets is improved.

[0169] 4. It can output a large number of micro-motion point clouds for stationary human targets, both near and far.

[0170] 5. Dynamic point clouds can be continuously output even for targets moving in the tangential direction.

[0171] In other embodiments, the signal processing process may be divided into dynamic point processing and micro-point processing (i.e., "static" point processing, as described later in the embodiments of this application). Dynamic point processing may be in-frame processing. It mainly detects dynamic points with apparent radial velocity. Micro-point processing is inter-frame processing, and by increasing the observation time, the velocity resolution is further improved to sense minute movements of the human body. Micro-point processing may also be divided into short-time micro-point processing and long-time micro-point processing.

[0172] In other preferred embodiments, the processing of dynamic points may be in-frame processing, i.e., detecting dynamic targets using echo data from multiple chirps within a single frame. Specifically, this includes the following steps S1201 to S1211.

[0173] Step S1201 performs distance window processing and distance FFT.

[0174] In other examples, distance windowing and distance FFT are performed on the ADC (Analog to Digital Converter) data of each received chirp to obtain a one-dimensional distance image. In the above operation, windowing may be performed, or it may not be performed, or different windowing may be performed simultaneously. Furthermore, for pulse-compressed radars, pulse compression is performed on each chirp echo signal to obtain a one-dimensional distance image. In some cases, a stepping frequency radar may be used in this step; that is, the embodiments of this application do not limit the radar mechanism, as long as a one-dimensional distance image can be obtained.

[0175] In step S1202, zero Doppler removal is performed.

[0176] In another example, the results of the one-dimensional distance image for each chirp are accumulated and averaged based on the distance bin to obtain zero-Doppler channel data in the frame, and then the zero-Doppler channel data in the frame is subtracted from each chirp to filter out stationary targets in the frame and retain dynamic targets in the frame.

[0177] In step S1203, velocity window processing and velocity FFT are performed.

[0178] In another example, after zero-Doppler removal based on distance bins, windowing (or not windowing) is performed on the data according to the slow time dimension, a slow time dimension FFT is performed, and echoes of the same distance in the scene are separated by velocity dimension.

[0179] In step S1204, non-coherent accumulation of inter-channel RDM (Distance Doppler Matrix, i.e., the aforementioned RD spectral data) is performed.

[0180] In another example, modulus squaring is performed on the FFT results for each channel velocity, cross-channel accumulation is performed to find the average value, non-coherent gains between channels are obtained, and an RDM is obtained for target detection.

[0181] In some cases, multiple RDMs may be formed by multi-channel non-coherent or coherent accumulation, and then the multiple RDMs may be detected, or detection may be performed by directly calculating the modulus square for one or more channels without accumulation.

[0182] In step S1205, perform NRCFAR.

[0183] In another example, the first 10 distance bin data (within a range of approximately 1m), excluding the 20 Doppler channels near the zero Doppler channel, are extracted from the detected RDM, their average value is calculated, and the noise average value is determined. Each point in the detected RDM is compared with the above noise average value, and if it is greater than a certain threshold, that point is included as a candidate point. To detect all weak echo points, the threshold set here may be obtained by modulo arithmetic or modulus square method, i.e., a relatively low threshold (compared to the conventional threshold) is used to detect as many candidate points as possible, for example, it can be set to 10 times by the modulus square method. Preferably, the noise selection method may be to calculate the noise average value by selecting several distance bins at the furthest distance points, i.e., a more accurate noise floor can be obtained mainly based on which ranges are relatively simple, contain only noise, and do not have a target.

[0184] Step S1206 performs a DOA (Direction of Arrival).

[0185] In another example, DBF is performed on the primary and secondary sub-array channel data of candidate points based on the TRX array configuration to obtain the azimuthal DBF spectrum. Then, the pitch DBF is calculated using a combined method to obtain the pitch angle.

[0186] In step S1207, the confidence level of the point cloud is calculated.

[0187] In other examples, if the energy of a candidate point originates solely from the target and its signal configuration is single, a clear principal peak will be present as the DBF spectrum approaches an ideal sinc function. Therefore, a confidence level based on the candidate point azimuthal DBF spectrum can be used to determine whether a candidate point is the desired target point. The confidence level is divided into confidence level 1 and confidence level 2.

[0188] The formula for calculating confidence level 1 (conf1) is: conf1 = max / peak2 is also acceptable. Here, max is the maximum value of the azimuthal DBF spectrum, peak2 is the second largest extremum of the azimuthal DBF spectrum, and confidence level 1 reflects the highest sidelobe level of the DBF spectrum. A higher confidence level of 1 indicates a lower highest sidelobe, and the more single its signal component, the higher the probability that it originated from a single target.

[0189] The formula for calculating confidence level 2 (conf2) is: conf2 = max / mean(peakoth) is also acceptable. Here, max is the maximum value of the azimuthal DBF spectrum, peakoth is the remaining extrema other than the maximum and second largest extrema of the azimuthal DBF spectrum, and mean() indicates that the mean value is taken. Therefore, a confidence level of 2 reflects the energy concentration at the main peak of the DBF spectrum.

[0190] In the embodiments of this application, the method for calculating reliability is not unique; it is sufficient to measure the energy concentration of the DBF spectrum. That is, the higher the concentration, the purer the echo is and the higher the likelihood that it is an effective target.

[0191] In some cases, the confidence calculation in the embodiments of this application may be based on the azimuthal DBF spectrum, but may also be obtained using the pitch DBF spectrum, or by other methods such as the non-azimuthal non-pitch DBF spectrum.

[0192] In step S1208, a single target point is extracted from the active target.

[0193] In other examples, as mentioned above, the higher the confidence level of a candidate point (1), the more likely it is to be from a single target. Therefore, a certain threshold is set for confidence level 1, and if confidence level 1 is higher than this threshold, the target is output as a single target point. The current criteria for determining a single target point are: conf1 > 2.

[0194] In step S1209, dual target points are extracted from the active target.

[0195] In another example, in an indoor scene, it is unavoidable that two human body targets are located in the same distance unit. When the echoes of two targets are in the same RD unit, two close peaks appear in the azimuthal DBF spectrum when the two targets are sufficiently separated in azimuthal. To avoid candidate points in such situations being selected by the "Single Target Selection" module, confidence levels 1 and 2 are used to determine dual targets. If confidence levels 1 and 2 simultaneously satisfy the following conditions, the main lobe and highest side lobe of the candidate point DBF spectrum are output as dual targets.

[0196] conf1<1.5, conf2>2.5.

[0197] In step S1210, the candidate point set is obtained.

[0198] In other examples, for point clouds that do not satisfy the single target point or dual target point criteria, if all of the following conditions are met, they will be output as candidate point clouds.

[0199] conf1>1.5, conf2>3, SNR>10.

[0200] Step S1211 enhances the point cloud.

[0201] In another example, for "high-quality" point clouds—that is, point clouds with concentrated DBF spectral energy and a high signal-to-noise ratio—the left and right azimuthal points of the main peak of the DBF spectrum are output simultaneously, and more weight is assigned to the "high-quality" point clouds based on their number, thereby enhancing the overall density of the scene point cloud. The current criteria for determining a "high-quality" point cloud are: conf2 > 4.5 and SNR > 30.

[0202] In other embodiments, the processing of micromotion points can be divided into short-time micromotion point processing and long-time micromotion point processing. Both may be interframe processing operations. For example, short-time micromotion points may be processed using 4 frames, and long-time micromotion points may be processed using 16 frames of time-length data (one of two frames is selected from the 16 frames, and the final number of frames used is 8). Except for the difference in the number of frames used, the processing flow of both is basically the same. To avoid interference of dynamic targets with the detection of micromotion targets, both use intraframe zero-Doppler channel data as input. Specifically, this includes the following steps.

[0203] Micromotion point processing is divided into short-term and long-term micromotion point processing. Both are interframe processes. Short-term micromotion points are processed using 4 frames. Long-term micromotion points are processed using 16 frames of time-length data (one of two frames is selected from the 16 frames, and the final number of frames used is 8). Except for the difference in the number of frames used, the processing flow of both is basically the same. To avoid interference from dynamic targets with the detection of micromotion targets, both use intraframe zero-Doppler channel data as input. The micromotion point processing steps are as follows:

[0204] After the zero-Doppler removal step S1202, during step S1203, step 1, which performs inter-frame zero-Doppler removal, and step 2, which performs inter-frame DFT, are executed. In other examples, inter-frame zero-Doppler removal involves taking the inter-frame zero-Doppler average value of multiple frames for each distance bin and subtracting this average value from the distance bin corresponding to the zero-Doppler data of each frame. In other examples, inter-frame DFT, i.e., using a sliding window method (or non-sliding window method), is performed on the zero-Doppler channel data (after subtracting the average value) of each distance bin to obtain a two-dimensional FFT matrix for each channel.

[0205] In other embodiments, when outputting point clouds, the outputting point clouds is used to perform a hierarchical processing, fully representing the scene situation from different dimensions, thereby allowing the user to accurately understand the meaning, application, and usage conditions of each type of point cloud. Specifically, In other embodiments, point cloud data such as dynamic points, short-term micro-motion points, long-term micro-motion points, and static points are displayed in a hierarchical manner, i.e., Dynamic points can reflect the dynamic targets of the scene and may include active human bodies with known radial velocities, as well as slightly swaying human targets and human targets moving mostly tangentially. Overall, the reliability of dynamic points is higher than that of short-term and long-term micro-points. Therefore, they can be used to initiate a track and maintain track reliability.

[0206] Short-time micro-motion points are an effective supplement to dynamic points; that is, when the human body is stationary, short-time micro-motion points can continuously maintain perception of the human body target and reflect the human body target's posture to some extent. Therefore, short-time micro-motion points can be used for continuous tracking of a stationary human body.

[0207] At the same time, because long-duration micro-motion points are detected using a relatively long number of frames, they can detect even weaker micro-motion targets. Due to the long number of frames used, there is a clear delay in long-duration micro-motion points for a single dynamic target; that is, the long-duration micro-motion point remains at position A for some time after the target has moved away from position A. Therefore, long-duration micro-motion points can be used for the continuous retention of stationary targets and the maintenance of tracks. For example, if a stationary track does not have long-duration micro-motion points and the reliability of that stationary track is not high, it may be considered to delete that track.

[0208] In other embodiments, for point clouds such as single targets, dual targets, and candidate point clouds, Dynamic points and long / short-term micro-movement points all include single target point clouds, dual target point clouds, and candidate point clouds. From the confidence angle of the point cloud, You can see the difference between a single-target point cloud, a dual-target point cloud, and a candidate point cloud.

[0209] Therefore, for a single target point cloud, the higher the confidence level of the point cloud (1), the higher the SNR and the greater the reliability of the point cloud. A single dynamic target point cloud can be used to start a track, a dual target point cloud should be used with caution, and a candidate point cloud is not recommended for track initiation. At the same time, when tracking a confirmed track, a dual target point cloud can be used conditionally when a single target point cloud is unavailable. That is, if the other points in the dual target point cloud are definitely within the wave gate of another confirmed track, the confidence level of that dual target point cloud improves, and the current track can be associated using that dual target point cloud. Also, if there are no single target or dual target dynamic points around any confirmed track, it may be possible to use a candidate point cloud.

[0210] To facilitate a better understanding of the above embodiments by those skilled in the art, the following will describe some example target tracking application scenarios.

[0211] Depending on the needs of dynamic target detection, the target detection method is: Step 1401 involves acquiring an echo signal, performing windowing on each chirp in the echo signal, and performing distance-dimensional FFT based on the signal after windowing. Step 1402 involves performing a distance-dimensional FFT on each chirp within the same frame, obtaining the average of each distance unit (bin), identifying the zero-Doppler channel data of the corresponding frame, and subtracting the zero-Doppler channel data from the signal of each chirp to remove clutter. Step 1403 involves windowing the portion of the data with distance bins in the data after removing clutter from the signal of each chirp, in order to perform a slow time-dimension FFT. Step 1404 involves performing a modulo operation on the result of a slow time-dimension FFT and accumulating the results to obtain a Doppler matrix. Step 1405 involves performing target detection on the Doppler matrix using the CFAR algorithm, Step 1406 involves extracting the distance bin and Doppler bin of the detected target, performing digital beamforming (DBF) processing, performing modulo calculations, and obtaining the DBF spectrum of the target. Step 1407 involves finding the maximum value for the target's DBF spectrum, calculating the azimuth angle corresponding to the maximum value as the target's azimuth angle, and obtaining the target's angle measurement result. Step 1408 calculates all maxima in the target DBF spectrum, Step 1409 involves sorting the local maximum values ​​obtained by calculating the same target in descending order, Step 1410 calculates the ratio of the maximum value to the second largest value in the sorting results as the first effectiveness parameter of the target, Step 1411 involves calculating the mean of all local maxima other than the target maximum value, and calculating the ratio of the maximum value to the mean value as the second effectiveness parameter. Step 1412 classifies targets whose first effectiveness parameter is greater than a first predetermined threshold as effective targets in trace pool 1, targets whose second effectiveness parameter is greater than a second predetermined threshold among the remaining target points as potential effective targets in trace pool 2, and the remaining targets as invalid targets in trace pool 3. The target cluster tracking process may include step 1413, which uses the target in tracepool 1 if the target in tracepool 1 satisfies the tracking request, and uses the targets in both tracepool 1 and tracepool 2 if the target in tracepool 1 does not satisfy the tracking request.

[0212] Depending on the needs for detecting micro-motion targets, the target detection method may be: Step 1501 involves acquiring an echo signal, performing windowing on each chirp in the echo signal, and performing distance-dimensional FFT based on the signal after windowing. Step 1502 involves performing a distance-dimensional FFT on each chirp within the same frame and identifying the average value of each distance bin obtained as the zero-Doppler channel data of the corresponding frame. Step 1503 continues to acquire zero-Doppler channel data corresponding to at least one frame of echo signal, Step 1504 involves performing an FFT on each distance bin of the acquired zero-Doppler channel data, Step 1505 involves performing a modulo operation on the FFT result, adding the channels together, and forming a Doppler matrix. Step 1506 involves detecting the Doppler matrix using k(k>1) × noise estimate as a threshold, and obtaining the distance bin and Doppler bin of the candidate points. Step 1507 involves extracting the distance bin and Doppler bin of the detected target corresponding to all azimuthal channels, performing DBF processing, performing modulo calculation, and obtaining the DBF spectrum of the target. Step 1508 involves finding the maximum value for the target's DBF spectrum, calculating the azimuth angle corresponding to the maximum value as the target's azimuth angle, and obtaining the target's angle measurement result. Step 1509 calculates all maxima in the target DBF spectrum, Step 1510 involves sorting the local maximum values ​​obtained by calculating the same target in descending order, Step 1511 calculates the ratio of the maximum value to the second largest value in the sorting results as the first effectiveness parameter of the target, Step 1512 involves calculating the mean of all local maxima other than the target maximum, and calculating the ratio of the maximum to the mean as the second effectiveness parameter. Step 1513 classifies targets whose first effectiveness parameter is greater than a first predetermined threshold as effective targets in trace pool 1, targets whose second effectiveness parameter is greater than a second predetermined threshold among the remaining target points as potential effective targets in trace pool 2, and the remaining targets as invalid targets in trace pool 3. The target cluster tracking process may include step 1514, which uses the target in tracepool 1 if the target in tracepool 1 satisfies the tracking request, and uses the targets in both tracepool 1 and tracepool 2 if the target in tracepool 1 does not satisfy the tracking request.

[0213] In this embodiment, after detecting a target using echo signals received by multiple radar receiving channels, corresponding feature information is generated for the detected target. Based on this feature information, it is possible to determine whether the detected target is a valid target. By detecting a target and then determining its validity, false detections where noise points are detected as targets are reduced, thereby improving the accuracy of indoor target detection.

[0214] In this embodiment, it is permissible to set the threshold to a small value, and the threshold does not necessarily have to be a high value. This is advantageous in improving the accuracy of indoor target detection by reducing or avoiding the detection of more points in a certain false alarm detection, which would cause valid targets to be mistakenly identified and excluded as noise points.

[0215] In this embodiment, by providing a processing method that adapts to the dynamic characteristics of a dynamic target and obtaining a Doppler matrix, the accuracy of the Doppler matrix can be improved, and furthermore, the accuracy of target detection performed based on the Doppler matrix can be improved, which is advantageous in improving the accuracy of indoor dynamic target detection.

[0216] The above division of steps into various methods is merely for the purpose of clarifying the explanation, and may be merged into a single step or divided into multiple steps during implementation, as long as they contain the same logical relationships, and may also involve adding non-essential modifications to the algorithm or process or introducing non-essential designs, as long as they do not alter the algorithm and process core designs, and are all within the scope of the patent.

[0217] Another embodiment of the present invention further provides a computer-readable storage medium in which a computer program is stored. The above embodiment of the method is realized when the computer program is executed by a processor.

[0218] As those skilled in the art will understand, the implementation of all or part of the steps in the methods of the above embodiments can be completed by instructing the relevant hardware by program, which is stored in a storage medium and includes several instructions for causing a device (which may be a single-chip microcomputer, or a chip) or processor to perform all or part of the steps of the methods of each embodiment of the present application. The aforementioned storage medium includes a variety of media capable of storing program code, such as U disks, removable hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0219] Another embodiment of the present invention further provides an integrated circuit including sequentially connected radio frequency module 1101, analog signal processing module 1102, and digital signal processing module 1103, as shown in Figure 11. The radio frequency module 1101 generates a radio frequency transmission signal and receives a radio frequency reception signal. The analog signal processing module 1102 performs frequency reduction processing on the radio frequency received signal to obtain an intermediate frequency signal. The digital signal processing module 1103 performs analog-to-digital conversion on an intermediate frequency signal to acquire digital signal data, acquires first data, generates second data based on a plurality of first data, detects a first target based on the first data, detects a second target based on the second data, and / or, the digital signal processing module 1103 performs analog-to-digital conversion on an intermediate frequency signal to acquire digital signal data, acquires a plurality of targets based on the target detection result, acquires corresponding feature information for each detected target, including noise features and signal features that characterize the target, and identifies the confidence level of each target based on the feature information.

[0220] In other embodiments, the integrated circuit may further include a data processing module that processes digital signals to enable target detection and / or wireless communication.

[0221] In other embodiments, the integrated circuit may be a millimeter-wave chip.

[0222] In other embodiments, the radio frequency received signal is an echo signal formed by the target transmission and / or scattering of the radio frequency transmitted signal, and the integrated circuit is a sensor chip.

[0223] It is readily apparent that this embodiment is a circuit embodiment corresponding to the method embodiment, and that this embodiment can be implemented in combination with the method embodiment. Details of the related technologies mentioned in the method embodiment are still valid in this embodiment and, to reduce redundancy, are omitted here. Accordingly, details of the related technologies mentioned in this embodiment can also be applied to the method embodiment.

[0224] Furthermore, in order to emphasize the novel aspects of the present invention, this embodiment does not introduce units that are less relevant to solving the technical problems related to the present invention, but this embodiment does not imply that other units do not exist.

[0225] Another embodiment of the present application relates to a sensor comprising a carrier, an integrated circuit provided on the carrier, and an antenna provided on the carrier or integrated with the integrated circuit as a single component and provided on the carrier. The integrated circuit is connected to the antenna and processes echo signals received by the antenna. The integrated circuit is the integrated circuit according to the above embodiment. The integrated circuit is the integrated circuit according to any embodiment of the present application.

[0226] If the antenna and integrated circuit are not integrated into a single component, the integrated circuit is connected to the antenna via a first transmission line, which may be a printed circuit board (PCB) wiring. The carrier may be a printed circuit board (PCB) such as a development board, data acquisition board, or main board of equipment, and will not be described further here.

[0227] The structure and operating principle of the integrated circuit included in the sensor have been described in detail in the above embodiment, so a detailed explanation is omitted here.

[0228] The terminal device according to the embodiment of the present application may include a device body and a sensor as described above provided on the device body, the sensor being used for target detection and / or communication to provide reference information for the operation of the device body.

[0229] In other embodiments, the sensor may be located outside the main body of the device; in other embodiments, the sensor may be located inside the main body of the device; and in other embodiments, part of the sensor may be located inside the main body of the device and part of it may be located outside the main body of the device. The embodiments of the present application are not limited to these, and will be specifically determined depending on the circumstances.

[0230] The sensor, by transmitting and receiving wireless signals, enables functions such as target detection, providing measurement information of the detected target to the main unit, thereby assisting and ultimately controlling the operation of the main unit. The measurement information includes, for example, at least one of relative distance, relative velocity, and relative angle.

[0231] In other embodiments, the device body may be components and products applied in fields such as transportation, consumer electronics, surveillance, indoor detection, and healthcare. For example, the device body may be intelligent transportation equipment (e.g., automobiles, motorcycles, ships, subways, trains, etc.), security equipment (e.g., cameras), liquid level / flow rate detection equipment, smart wearable devices (e.g., bracelets, glasses, etc.), smart home equipment (e.g., cleaning robots, door locks, televisions, air conditioners, smart lamps, etc.), various communication devices (e.g., mobile phones, tablet computers, etc.), or it may be gates, smart traffic signals, smart signs, traffic cameras, and various industrialized mechanical arms (or robots), or various measuring instruments for detecting vital sign parameters, and various devices that incorporate such measuring instruments, such as in-car detection equipment, indoor occupancy monitoring equipment, smart medical devices, consumer electronics, etc.

[0232] In other embodiments, when the above-mentioned device is applied to an Advanced Driving Assistance System (ADAS), the electromagnetic wave sensor as an in-vehicle sensor can provide the ADAS system with various functional safety assurances, such as Autonomous Emergency Braking (AEB), Blind Spot Detection (BSD), Lane Changing Assist (LCA), and Rear Cross Traffic Alert (RCTA).

[0233] Furthermore, the examples mentioned in the above embodiments can be freely combined, and any combination can be understood as a single embodiment. The terms "embodiment" or "example" appearing in each part of this specification do not necessarily refer to the same embodiment, nor are they exclusive, independent, or alternative embodiments to other embodiments. Those skilled in the art will understand that the embodiments described herein can be combined with other embodiments.

[0234] Those skilled in the art will understand that the above embodiments are specific examples of realizing the present invention, and that in actual applications, various modifications can be made to the form and details without departing from the spirit and scope of the present invention.

Claims

1. Steps to obtain the first data, A step of generating second data based on a plurality of first data, A step of detecting a first target based on the first data, The process includes the step of detecting a second target based on the second data, Target detection method.

2. The second data includes i types of second data and j types of second data, where i and j are integers greater than 1. The step of generating second data based on a plurality of first data is: A step of obtaining i types of second data based on Ni first data, The process includes the step of obtaining the j types of second data based on the Nj first data, Ni and Nj are positive integers, and Nj > Ni. The step of detecting a second target based on the second data is: The steps include: calculating a second target of type i based on the aforementioned i-type second data; The step of detecting a second target of type j based on the aforementioned second data of type j, The target detection method according to claim 1.

3. The first data is channel data, The step of generating second data based on a plurality of first data is: The step of obtaining the second data by superimposing multiple first data, The target detection method according to claim 1 or 2.

4. The step of generating second data based on a plurality of first data is: A step of averaging multiple first data to obtain zero-Doppler channel data, The step of obtaining the second data by subtracting the zero-Doppler channel data from a plurality of the first data and then superimposing them, The target detection method according to claim 1 or 2.

5. The step of detecting a second target based on the second data is: The steps include performing a low-speed time-dimension FFT based on the second data, The process includes the step of performing a certain level of false alarm detection based on the FFT results so that the detection of the second target is realized, A target detection method according to any one of claims 1 to 4.

6. The first data is RD spectral data, The step of generating second data based on a plurality of first data is: The step includes estimating the second data based on a plurality of the first data, The target detection method according to claim 1 or 2.

7. The step of obtaining the first data is: A step of acquiring echo data corresponding to a detection signal having a predetermined transmission period of 50 ms or more, The step of generating the first data based on the echo data includes, A target detection method according to any one of claims 1 to 6.

8. The aforementioned detection signal includes a plurality of chirp signals, The detection signal having the predetermined transmission period is obtained by at least one of the following: increasing the number of chirp signals included in the detection signal, increasing the period of the chirp signals, or increasing the frame idle time between chirp signals. The target detection method according to claim 7.

9. The aforementioned detection is The step includes detecting a certain false alarm based on a predetermined threshold, wherein the threshold = k × estimated noise value, and k > 1. A target detection method according to any one of claims 1 to 7.

10. Steps include acquiring multiple targets based on the target detection results, For each detected target, the process involves obtaining characteristic information including corresponding noise features and signal features that characterize the target. The step of identifying the confidence level of each target based on the aforementioned characteristic information, Target detection method.

11. The step of obtaining the corresponding feature information is: The process includes the step of obtaining the angular spectrum generated in the target detection process, The step of identifying the confidence level of each target based on the aforementioned characteristic information is: A step of identifying the maximum value of the angular spectrum corresponding to the target, A step of identifying a first effectiveness parameter that characterizes the relative magnitude between the maximum value among the maximum values ​​of the angular spectrum and the average value of the other maximum values, based on the maximum value among the maximum values ​​of the angular spectrum and the average value of the other maximum values. The steps include: determining the confidence level of each target based on the first effectiveness parameter; The target detection method according to claim 10.

12. The step of identifying the confidence level of each target based on the first effectiveness parameter is: A step of identifying a second effectiveness parameter that characterizes the relative magnitude of the relationship between the maximum and second largest values ​​of the angular spectrum for the corresponding target, based on the maximum and second largest values ​​of the angular spectrum. A step of determining whether the corresponding target belongs to an invalid target based on the second effectiveness parameter, If the corresponding target does not belong to an invalid target, the step of determining the confidence level of the target based on the first validity parameter includes: The target detection method according to claim 11.

13. The step of obtaining the corresponding feature information is: This includes the step of obtaining the entropy value of the corresponding target, The step of identifying the confidence level of each target based on the aforementioned characteristic information is: The step includes determining the confidence level of each target based on the aforementioned entropy value, The target detection method according to claim 10.

14. The step of obtaining the corresponding feature information is: The process includes the step of obtaining the signal-to-noise ratio generated by a certain false alarm detection in the target detection process for the corresponding target, The step of identifying the confidence level of each target based on the aforementioned characteristic information is: The step includes determining the reliability of each target based on the signal-to-noise ratio, The target detection method according to claim 10.

15. A step of acquiring a first target and a second target based on the target detection method described in any one of claims 1 to 9, For each of the detected first and second targets, the step of acquiring characteristic information including corresponding noise features and signal features that characterize the target, The step of determining the confidence level of each first target and each second target based on the aforementioned feature information, Target detection method.

16. A step of acquiring a first target and a second target based on the target detection method described in any one of claims 1 to 9 or the target detection method described in claim 15, The step of outputting the first target and the second target in a hierarchical structure, Target point cloud output method.

17. It includes sequentially connected radio frequency modules, analog signal processing modules, and digital signal processing modules, The aforementioned radio frequency module generates a radio frequency transmission signal and receives a radio frequency reception signal. The analog signal processing module performs frequency reduction processing on the radio frequency received signal to obtain an intermediate frequency signal. The digital signal processing module performs analog-to-digital conversion on an intermediate frequency signal to acquire digital signal data, and based on the digital signal data, implements the target detection method described in any one of claims 1 to 9, the target detection method described in any one of claims 10 to 14, the target detection method described in claim 15, or the target point cloud output method described in claim 16. Integrated circuit.

18. Carrier and, The integrated circuit according to claim 17 provided on the carrier, An antenna provided on the carrier, or integrated into a single component together with the integrated circuit and provided on the carrier, The aforementioned integrated circuit is connected to the antenna and processes the echo signal received by the antenna. Sensor.

19. The main unit of the device, The device body includes the sensor described in claim 18, The sensor is used for target detection and / or communication to provide reference information for the operation of the device body. Terminal equipment.

Images