Target detection methods, target point cloud output method, integrated circuit, sensor, and terminal device

Abstract

Target detection methods, a target point cloud output method, an integrated circuit, a sensor, and a terminal device, which relate to the technical field of target detection. One of the target detection method comprises: acquiring first data (101); on the basis of a plurality of pieces of first data, generating second data (102); on the basis of the first data, detecting a first target (103); and on the basis of the second data, detecting a second target (104). The method facilitates an improvement in the accuracy of radar target detection.

Description

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

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application is based on the Chinese patent application with application number "202311127963.3" and application date of September 1, 2023, the Chinese patent application with application number "202311127857.5" and application date of September 1, 2023, the Chinese patent application with application number "202311198067.6" and application date of September 15, 2023, and the Chinese patent application with application number "202410678539.6" and application date of May 28, 2024, and claims the priority of the above-mentioned Chinese patent applications. The entire contents of the above-mentioned Chinese patent applications are hereby incorporated into this application by introduction. Technical Field

[0003] The embodiments of the present application relate to the field of target detection technology, and in particular to a target detection and point cloud output method, integrated circuit, sensor, and terminal device. Background Art

[0004] Radar is an electronic device that uses electromagnetic waves to detect targets. Its working principle is as follows: the radar system first transmits electromagnetic waves through the antenna. These electromagnetic waves will be reflected back after encountering the target. The radar system receives the echo signal generated by the reflection of the target, and determines the target's distance, distance change rate (radial velocity), direction, altitude and other information by processing and analyzing the echo signal.

[0005] However, the accuracy of target detection by current radar systems still needs to be improved.

[0006] Summary of the Invention

[0007] The embodiments of the present application provide a target detection and point cloud output method, integrated circuit, sensor, and terminal device, which are conducive to improving the accuracy of radar target detection.

[0008] According to some embodiments of the present application, a first aspect of the embodiments of the present application provides a target detection method, including: acquiring first data; generating second data based on multiple first data; detecting a first target based on the first data; and detecting a second target based on the second data.

[0009] According to some embodiments of the present application, the second aspect of the embodiments of the present application also provides a target detection method, including: obtaining several targets based on the target detection results; obtaining corresponding feature information for each detected target, the feature information including noise features and signal features for characterizing the target; and determining the confidence of each target based on the feature information.

[0010] According to some embodiments of the present application, a third aspect of the embodiments of the present application further provides a target detection method, including: obtaining a first target and a second target according to the target detection method as described in the first aspect; obtaining corresponding feature information for each of the first target and the second target detected, the feature information including noise features and signal features for characterizing the target; and determining the confidence of each of the first target and the second target based on the feature information.

[0011] According to some embodiments of the present application, the fourth aspect of the embodiments of the present application also provides a target point cloud output method, including: obtaining a first target and a second target according to the target detection method as described in the first aspect or the target detection method as described in the third aspect; and outputting the first target and the second target in a hierarchical manner.

[0012] According to some embodiments of the present application, the fifth aspect of the embodiments of the present application further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the target detection method or target point cloud output method as described in any embodiment of the present application.

[0013] According to some embodiments of the present application, the sixth aspect of the embodiments of the present application further provides an integrated circuit, comprising a radio frequency module, an analog signal processing module and a digital signal processing module connected in sequence; wherein the radio frequency module is used to generate a radio frequency transmit signal and receive a radio frequency receive signal; the analog signal processing module is used to down-convert the radio frequency receive signal to obtain an intermediate frequency signal; the digital signal processing module is used to perform analog-to-digital conversion on the intermediate frequency signal to obtain digital signal data, and based on the digital signal data, implement the target detection method as described in the first aspect, or implement the target detection method as described in the second aspect, or implement the target detection method as described in the third aspect, or implement the target point cloud output method as described in the fourth aspect.

[0014] According to some embodiments of the present application, the seventh aspect of the embodiments of the present application further provides a sensor, including: a carrier; the integrated circuit as described in the sixth aspect, which is arranged on the carrier; an antenna, which is arranged on the carrier, or the antenna and the integrated circuit are integrated into a single device and arranged on the carrier; wherein the integrated circuit is connected to the antenna for processing the echo signal received by the antenna.

[0015] According to some embodiments of the present application, the eighth aspect of the embodiments of the present application also provides a terminal device, including: a device body; and a sensor as described in any embodiment of the present application arranged on the device body; wherein the sensor is used for target detection and / or communication to provide reference information for the operation of the device body.

[0016] The technical solution provided by the embodiment of the present application is that when detecting the second target, the second data used to detect the second target is generated by multiple first data, and the multiple first data correspond to the observation time of the multiple first data, that is, the second target has a longer observation time relative to the first target, so that it is possible to detect targets that are difficult to detect in a short time, avoiding missing the second target. At the same time, the detection of the first target can be achieved in a short time, so the present application can also achieve timely and efficient detection of the first target. Therefore, the present application can implement hierarchical detection for different targets, adapt to the characteristics of different targets, and can not only improve the accuracy of detection, but also ensure the timeliness of detection.

[0017] After detecting the target, the present application can also determine the confidence level of each detected target through the target's characteristic information, so that the target can be further verified, which is conducive to improving the reliability and accuracy of target detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] One or more embodiments are exemplarily illustrated by pictures in the corresponding drawings. These exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings are represented as similar elements. Unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0019] FIG1 is a flowchart of a target detection method provided in an embodiment of the present application;

[0020] FIG2 is a second flow chart of the target detection method provided in an embodiment of the present application;

[0021] FIG3 is a flowchart of a target detection method provided in an embodiment of the present application;

[0022] FIG4 is a fourth flow chart of the target detection method provided in an embodiment of the present application;

[0023] FIG5 is a flowchart of a target detection method provided in an embodiment of the present application;

[0024] FIG6 is a flowchart of a target detection method provided in an embodiment of the present application;

[0025] FIG7 is a flowchart of the target detection method provided in an embodiment of the present application;

[0026] FIG8 is a flowchart of the target detection method provided in an embodiment of the present application;

[0027] FIG9 is a flowchart of a target detection method provided in an embodiment of the present application;

[0028] FIG10 is a flow chart of a target point cloud output method provided in an embodiment of the present application;

[0029] FIG11 is a schematic diagram of the structure of an integrated circuit provided in an embodiment of the present application;

[0030] FIG12 is a schematic diagram of signal waveforms involved in an embodiment of the present application. DETAILED DESCRIPTION

[0031] To make the purpose, technical solutions, and advantages of the embodiments of the present application more clear, each embodiment of the present application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will appreciate that many technical details are provided in each embodiment of the present application to help readers better understand the present application. However, even without these technical details and various variations and modifications based on the following embodiments, the technical solutions claimed in the present application can still be implemented.

[0032] The following embodiments are divided for the convenience of description and should not constitute any limitation on the specific implementation of the present application. The various embodiments can be combined with each other and referenced to each other without contradiction.

[0033] To help those skilled in the art better understand the target detection and point cloud output methods, media, circuits, sensors, and devices provided in the embodiments of this application, the following first describes the target detection process of existing radars.

[0034] For example, a frequency modulated continuous wave (FMCW) radar transmits chirp signals frame by frame when detecting a target. Each frame of the transmitted signal contains multiple chirp signals. Digital signal processing is then performed on the echo corresponding to each frame, and radar point clouds are output frame by frame. The processing of each frame of the digital echo signal is as follows:

[0035] 1. Perform Fast Fourier Transform (FFT) within the chirp to obtain the original channel data.

[0036] 2. Calculate the average value of the original channel data of all chirps in the same frame to obtain the zero Doppler channel data of the frame.

[0037] 3. Subtract the zero Doppler channel data from the original channel data to obtain the updated channel data.

[0038] 4. Perform intra-frame FFT on the updated channel data to obtain Range-Doppler (RD) spectrum data.

[0039] 5. Perform coherent processing (or non-correlated processing) and constant false alarm rate (CFAR) detection on the RD spectrum data to obtain the target.

[0040] 6. For each detected target, extract the range unit (range bin) and Doppler unit (Doppler bin) to perform digital beam forming (DBF) and obtain the DBF spectrum of each target.

[0041] 7. For each target, find the maximum value of the DBF spectrum and calculate the azimuth corresponding to the maximum value to obtain the azimuth of the target.

[0042] However, the above-mentioned frame-based target detection and output method is not suitable for all targets when facing a wide variety of targets. For example, during detection, there may be dynamic targets whose movement can be detected quickly, such as a child running around a room, and there may also be slightly moving targets whose movement is difficult to detect, such as a person lying in bed, who may make small movements due to breathing. For dynamic targets, due to their generally large movement speed and displacement, sufficient detection information for sensing target changes will be accumulated within a frame period. In this case, the radar can output relatively accurate detection results for dynamic targets. However, for slightly moving targets, due to their generally small movement speed and displacement, sufficient detection information for sensing target changes may not be accumulated within a frame period. Moreover, the echo signal generated and detected by the target reflection is often mixed with noise, making the detection of such slightly moving targets very difficult. In this case, the radar cannot output accurate detection results for such slightly moving targets.

[0043] To address the above issues, this application further provides different observation times that are suitable for different targets. This allows for the detection of micro-moving targets without obvious motion characteristics by accumulating sufficient position changes to be observed over a relatively long time interval, thereby achieving more comprehensive detection of different types of targets and a more comprehensive perception and characterization of the scene. At the same time, for dynamic targets with obvious motion characteristics, the position changes accumulated over a relatively short time period can be output efficiently and timely.

[0044] To achieve the above objectives, embodiments of the present application provide a target detection method for performing targeted detection on different types of targets. In some embodiments, as shown in FIG1 , the target detection method includes the following steps:

[0045] Step 101: Acquire first data.

[0046] Step 102: Generate second data based on a plurality of first data.

[0047] Step 103: Detect the first target according to the first data.

[0048] Step 104: Detect the second target based on the second data.

[0049] Among them, each first data corresponds to an observation time length T, and the detection of the second target using the second data generated by multiple first data is equivalent to detecting the second target using the second data obtained within multiple observation time lengths T, thereby accumulating target features sufficient to be observed through a longer observation time, so that the second target that is difficult to detect in a short time can also be detected without missing detection. At the same time, the first target with obvious motion characteristics does not need to wait until multiple first data are accumulated before it can be detected. It can be detected using a single first data. Therefore, this application can not only detect the first target in a timely and efficient manner, but also detect the second target accurately, reliably and without omission, provide hierarchical detection for different targets, and ensure the accuracy and timeliness of detection of different types of targets.

[0050] It should be noted that the embodiments of the present application do not limit the specific value of T, which can be set in combination with actual needs, application scenario characteristics, target characteristics, algorithm characteristics, etc. For example, in some cases, the value of T is one frame period. In some cases, the motion characteristics of a dynamic target may be very prominent, and sufficient information about the observed target can be accumulated in less than one frame period. In this case, T can be taken as a smaller value, such as 0.5 frame period, 1 / 3 frame period, etc. In some cases, because the algorithm needs to obtain more data to ensure that the accuracy of target detection meets the requirements, T can be taken as a larger value, such as 1.5 frame period, 2 frame period, 5 frame period, etc. In some cases, the scenario may require a more real-time output. In this case, if the target detection algorithm supports it, T can also be taken as a smaller value, such as 0.2 frame period, 0.5 frame period, etc. Of course, the above is only an explanation of T provided with "frame" as the reference unit. In some cases, T may also refer to other periods, such as the chirp period duration. For example, T may be the duration of multiple consecutive chirp periods, etc., which will not be described in detail here.

[0051] Taking a relatively closed scene such as an indoor environment as an example, when detecting human activities, conventional radar target detection cannot accurately obtain the speed, acceleration and other motion information of the point cloud of a target that only has breathing activities or a slightly moving human body that shakes unconsciously (for example, a sitting human body). However, this application introduces the second data through the above embodiment to achieve the point cloud output of a second target with unclear motion characteristics, such as a slightly moving human body that only has breathing activities or shakes unconsciously. Thus, by increasing the amount of data based on which the second target is detected relative to the first target, the observation time is lengthened so that the second target has sufficient displacement within the observation time, thereby being able to obtain the point cloud of the slightly moving target.

[0052] It should also be noted that this embodiment does not limit the classification of first and second targets. In some cases, the first target can be a dynamic target, and the second target can be a slightly moving target. A dynamic target can be a target with a speed exceeding V1, and a slightly moving target can be a target with a speed less than V2. The specific values ​​of V1 and V2 can be determined based on the application scenario, the actual conditions of the target, and so are not listed here. In some cases, the first target can be a large target, and the second target can be a small target. In some cases, based on the application scenario, combined with empirical data, big data analysis, or cluster analysis, combined with motion trajectories, etc., targets can be classified according to their different motion states to obtain first and second targets. In some scenarios, the first target can be defined as a target detected when a person or vehicle moves relatively quickly within the target area, and the second target can be defined as any other target moving slower than the first target. In other cases, dynamic targets can be classified based on dimensions such as motion speed and motion amplitude. The specific classification operation can be based on a comprehensive consideration of the application scenario, actual needs, and radar monitoring performance, as long as the distinction between the first and second targets can be achieved, or a hierarchical display can be achieved later. For the sake of convenience, in the following exemplary description, the first target will be a dynamic target and the second target will be a micro-motion target.

[0053] It should also be noted that, during implementation, further requirements may exist for target detection. For example, taking data obtained after processing one frame of echo signals (i.e., one frame of echo signals) as a type of first data, if, during implementation, a more reliable detection result is desired, another type of first data may be set to an echo signal of two frames. Each time target detection is performed, data obtained after processing two frames of echo signals is obtained as another type of first data. In this way, more data can be used for target detection, which helps to ensure higher accuracy and reliability of target detection. For example, during implementation, if it is desired that the detection result can be fed back to a user or other device in a more real-time manner, the first type of data may be set to data obtained after processing an echo signal of 0.5 frames. Alternatively, when the data of the preset unit includes a preset number of adjacent chirp data within a frame, data obtained by processing echo signals of different numbers of adjacent chirp data may be used as the first type of data to implement the above embodiment. That is, as long as the periods of the signal processing units corresponding to different types of data are different, the above embodiment can be implemented. In this way, not only can the echo signals within a shorter time period be processed, but the target detection can also be performed more quickly on the obtained echo signals and the target detection results can be fed back, which is conducive to improving the real-time feedback of the target detection results to users or other devices. I will not go into details here.

[0054] To help those skilled in the art better understand the target detection method provided by the above embodiment, it will be explained below.

[0055] Step 101: Obtain first data. This embodiment does not limit the meaning of the first data. It is understood that, for the radar signal processing process, after completing different digital signal processing methods, corresponding data will be obtained, such as a digital signal obtained by sampling the echo signal, the aforementioned raw channel data, RD spectrum data, etc. The first data can be data obtained after any of the above processing methods, and the above effects can be achieved by using any of the above processing methods as the first data. However, when the meaning of the first data is different, its acquisition method will also be different. The echo signal can also be an echo signal generated by any type of transmitted signal reflected by a target, such as an echo signal generated by a signal transmitted by any continuous wave (CW) radar having multiple receiving channels and / or multiple transmitting channels and reflected by a target, or an echo signal generated by a signal transmitted by any frequency modulated continuous wave (FWCW) radar having multiple receiving channels and / or multiple transmitting channels and reflected by a target, etc.

[0056] For example, in some cases, the first data may be an initial digital signal. In this case, the first data may be obtained by performing analog-to-digital conversion on the echo signal.

[0057] For another example, in some cases, the first data may be original channel data. In this case, the first data may be obtained by the aforementioned step of obtaining the original channel data.

[0058] For example, in some cases, the first data may also be updated channel data. In this case, the acquisition of the first data can be achieved through the aforementioned step of acquiring the updated channel data.

[0059] For example, in some cases, the first data may also be RD spectrum data. In this case, the first data can be obtained by the aforementioned step of obtaining the RD spectrum data.

[0060] The embodiments of the present application do not limit the process of obtaining the above data.

[0061] For example, in some embodiments, RD spectrum data can be obtained by performing a range-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 the chirps belonging to the same frame; removing clutter frequency domain components from the frequency domain signal based on the corresponding zero-Doppler channel data to obtain a dynamic frequency domain signal for each chirp; and performing a slow-time Fourier transform on the data in the dynamic frequency domain signal corresponding to the same range unit in each chirp, and generating RD spectrum data based on the transform result. In this way, RD spectrum data is obtained by processing the echo signal to determine the zero-Doppler channel data for each frame of the signal, and then determining the dynamic frequency domain signal for each chirp, and performing a slow-time Fourier transform on the dynamic frequency domain signal for each chirp. In other words, providing a processing method that adapts to the dynamic characteristics of dynamic targets to obtain RD spectrum data is beneficial for improving the accuracy of indoor dynamic target detection.

[0062] For example, in some embodiments, RD spectrum data can also be obtained by performing a range-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 the chirps belonging to the same frame; performing a slow-time Fourier transform on the frequency domain signals corresponding to the zero-Doppler channel data corresponding to the same range unit in each frame, and generating RD spectrum data based on the transformed results. In this way, the RD spectrum data is obtained by processing the echo signal to determine the zero-Doppler channel data for each frame, and then further performing a Fourier transform on the zero-Doppler channel data of each frame, i.e., performing an inter-frame Fourier transform. In other words, providing a processing method that is adapted to the micro-motion characteristics of micro-motion targets to obtain RD spectrum data is beneficial for improving the accuracy of indoor static target detection.

[0063] In some embodiments, before performing a slow-time Fourier transform, the zero-Doppler channel data is updated based on the zero-Doppler channel data of each frame. Based on the corresponding updated zero-Doppler channel data, the clutter frequency domain component is removed from the frequency domain signal to obtain an updated frequency domain signal for each chirp. Accordingly, performing a slow-time Fourier transform on the frequency domain signal corresponding to the zero-Doppler channel data corresponding to the same range unit in each frame can be achieved by performing a slow-time Fourier transform on the updated frequency domain signal corresponding to the zero-Doppler channel data corresponding to the same range unit in each frame. In this way, by taking into account the effects of long-term clutter and updating the zero-Doppler channel data based on data from multiple frames, clutter removal is more thorough, thereby obtaining a more accurate frequency domain signal of a micro-motion target, and thus more accurately detecting the micro-motion target, thereby further improving the accuracy of micro-motion target detection.

[0064] The above-mentioned various types of processing can be implemented by using corresponding algorithms or processing. For example, the generation of zero Doppler pass data can be implemented by using a moving target identification (MTI) algorithm, a phasor mean cancellation algorithm, and the like.

[0065] In some embodiments, as shown in FIG2 , obtaining the first data can be achieved by the following steps:

[0066] Step 201: Acquire echo data corresponding to a detection signal having a preset transmission period, where the preset transmission period is not less than 50 ms (milliseconds).

[0067] Step 202: Generate first data according to the echo data.

[0068] Among them, steps 203 to 205 are substantially the same as steps 102 to 104 of the embodiment shown in FIG1 , and are not described in detail here.

[0069] In this embodiment, the duration of the transmission cycle is extended. Specifically, the duration of the transmission cycle is increased from approximately 20 ms per frame used by existing radars to 50 ms or more per frame. This ensures that when target detection is performed based on frame data as the basic unit, the data within the frame has sufficient sensitivity to dynamic points. In addition, since the duration of the transmission cycle is longer, the velocity resolution is also improved. It can also further output more point clouds of the first target, achieving an effect of approximate coarse imaging, and can fully utilize the space of the RD spectrum data obtained through spectrum analysis (such as 2DFFT (two-dimensional fast Fourier transform, 2D Fast Fourier Transform)).

[0070] It should be noted that this embodiment does not limit the method of extending the transmission period. It is understandable that the detection signal usually includes multiple chirp signals. Therefore, the method of extending the transmission period of the detection signal can be achieved by configuring the chirp signal. In some examples, extending the transmission period can be achieved by increasing the number of chirp signals included, that is, the detection signal with a preset transmission period is obtained by increasing the number of chirp signals included in each frame of the detection signal. In some examples, extending the transmission period can also be achieved by increasing the period of each chirp signal, that is, the detection signal with a preset transmission period is obtained by increasing the period of the included chirp signals, such as increasing the duration A (i.e., the idle time in the chirp signal) as shown in Figure 12. In some examples, extending the transmission period can also be achieved by increasing the frame idle time between the chirp signals in the detection signal, that is, the detection signal with a preset transmission period is obtained by increasing the idle time after the chirp signal in the detection signal, such as increasing the duration B as shown in Figure 12, etc. Among them, Figure 12 only uses the chirp signal in the FMCW waveform as a triangular wave, and the rising edge appears first and then the falling edge as an example for illustration, but it does not mean that the embodiment of the present application is only applicable to the FMCW waveform shown in Figure 12, and will not be described one by one here.

[0071] In step 102, second data is generated based on multiple first data. As previously mentioned, there are many different scenarios for the first data, and correspondingly, there are also many different implementations for generating the second data. For ease of understanding, the following example uses the first data as channel data and RD spectrum data as an example.

[0072] When the first data is channel data, and the channel data is obtained by subtracting zero-Doppler channel data from the original channel data, in some 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 has not undergone subsequent processing, the second data can be obtained by superimposing, without complex calculations, and thus is easy to implement.

[0073] Considering that the second data is generated from multiple first data, the multiple data may have the same interference, that is, the multiple first data may have the same noise component. Based on this, in some embodiments, as shown in FIG3 , generating the second data based on the multiple first data can be achieved by the following steps:

[0074] Step 302: Averaging the plurality of first data to obtain zero Doppler channel data.

[0075] Step 303 : Subtract zero Doppler channel data from a plurality of first data and then superimpose the data to obtain second data.

[0076] Among them, step 301, step 304, and step 305 are substantially the same as step 101, step 103, and step 104 in the aforementioned embodiment, and are not described in detail here.

[0077] Removing the averaged zero-Doppler channel data is equivalent to removing the same noise components between different data. Therefore, in the above embodiment, noise component removal is performed twice on the echo signal. The first time, because the average value is calculated within the first data, the noise components within the first data are removed. The second time, because the average value is calculated for multiple first data, that is, the average value is calculated outside the first data, the noise components between the multiple first data are removed. This can better remove different types of noise components from multiple first data, thereby more effectively extracting signal components. This allows for more accurate target detection based on more effective signal components, resulting in better target detection results.

[0078] For ease of understanding, this example uses frames as the processing unit. The first average is performed on the data within a frame, removing noise components within the frame. The second average is performed across multiple frames, obtaining an average between frames and removing noise components that extend across multiple frames.

[0079] In some cases, multiple frames of zero Doppler channel data can be compared to eliminate abnormal data, and then the average of the retained multiple frames of zero Doppler channel data is used as the zero Doppler channel data in step 302, which is conducive to avoiding sudden noise interference and further improving the accuracy of the zero Doppler channel data.

[0080] In this way, considering the influence of long-term clutter, the zero-Doppler channel data is updated according to the data of multiple frames, so that the clutter removal is more thorough, thereby obtaining a more accurate frequency domain signal corresponding to the micro-motion target, and then being able to detect the micro-motion target more accurately, which is conducive to further improving the accuracy of micro-motion target detection.

[0081] When the first data is RD spectrum data, in some embodiments, generating the second data based on multiple first data can be achieved by inferring the second data based on the multiple first data. In this case, because the RD spectrum data undergoes further processing relative to the channel data, the relationship between the data is complicated. Therefore, the second data cannot be obtained through simple superposition, but requires certain inferences. The inference process is related to the FFT processing performed on the channel data and will not be detailed here. However, compared to the aforementioned case where the first data is channel data, when the first data is RD spectrum data, the processing data for the detection of the first and second targets before obtaining the RD spectrum can be shared, thereby simplifying the process and improving efficiency.

[0082] Of course, the above is only an exemplary description. In some embodiments, other data may be configured as the first data, etc., which will not be described in detail here.

[0083] It should also be understood that the above embodiments categorize targets into two main categories: first targets and second targets. However, considering that various micro-movement targets may exist in indoor environments and have different motion characteristics, the observation time required to detect different micro-movement targets also varies. The amount of echo data used for target detection reflects this observation time. For example, micro-movements of the head generated when reading different parts of a book page and micro-movements of the body when walking slowly indoors have larger amplitudes and higher speeds, so the observation time required to detect the latter is shorter than the observation time required to detect the former. Therefore, the number of echo signal frames required to detect micro-movements of the head generated when reading different parts of a book page is less than the number of echo signal frames required for micro-movements of the head generated when reading different parts of a book page. If, during the target detection process, historical echo signals of the same frame are combined with the first echo signal for all micro-movement targets, the detection feedback of some micro-movement targets will be delayed relative to their motion (i.e., low real-time performance) and / or some micro-movement targets will not be detected (i.e., low accuracy). Therefore, the classification of the target can be further refined according to the further refined motion features. In this way, more refined target detection can be provided based on the more refined target classification, thereby achieving more accurate target detection.

[0084] Based on this, taking the further classification of the second target as an example, in some embodiments, as shown in FIG4 , generating the second data based on the plurality of first data can be achieved through the following steps:

[0085] Step 402: Obtain i types of second data based on Ni first data.

[0086] Step 403: Obtain j types of second data according to Nj first data.

[0087] Accordingly, detecting the second target according to the second data can be achieved through the following steps:

[0088] Step 405: Detect the i-th type of second target based on the i-th type of second data.

[0089] Step 406: Detect the j-th type of second target based on the j-th type of second data.

[0090] Among them, the i-type second data and the j-type second data both belong to the second data, i and j are integers greater than 1, Ni and Nj are positive integers, and Nj>Ni.

[0091] In this way, by further subdividing the target and providing different amounts of first data for detection in a targeted manner, that is, providing different observation time lengths, it is possible to process the motion information of the second target accumulated over different observation time lengths, and present the target detection results of different types of second targets from more time dimensions, more comprehensively, and more in real time, which is conducive to further improving the real-time and accuracy of target detection.

[0092] It should be noted that step 401 and step 404 in this embodiment are substantially the same as step 101 and step 103 in the aforementioned embodiment, and are not described in detail here.

[0093] It should also be noted that Nj and Ni are not limited herein and can be fixed values ​​or values ​​that vary based on information such as the total number of targets included in the target detection result of the last detected first echo signal, the number of first and / or second targets, the speed of the second target, and the acceleration of the second target. In some cases, even if Nj and Ni are fixed values, their specific values ​​can still be set based on requirements, such as the real-time requirements for target detection. It is understood that when real-time requirements are high, the values ​​of Nj and Ni cannot be too large, otherwise it will easily result in a large amount of data to be processed, resulting in low processing efficiency and affecting target detection efficiency. In cases where the movement amplitude and speed of a slightly moving target are too small, the values ​​of Nj and Ni cannot be too small, otherwise it will easily result in the target being unable to be observed during the target detection process, resulting in missed targets. Therefore, different values ​​for Nj and Ni can be set based on the application scenario and the real-time requirements during the application.

[0094] Among them, the distinction between different second targets can be set according to needs. For example, the i-th type of second target is defined as the target generated by the breathing, heartbeat, pulse, etc. of a person at rest, and the j-th type of second target is the target generated by the slow movement of the body when the person is still (such as the unconscious slow movement of the body when sitting and falling asleep, or the displacement of the body when concentrating, etc.). Alternatively, different types of dynamic point clouds can be divided based on the motion speed dimension, and the second target with a speed greater than V3 is regarded as the i-th type of second target, and the second target with a speed not less than V4 is regarded as the j-th type of second target, etc. I will not go into details here.

[0095] Of course, the above is only an exemplary explanation for the case where the second target is further refined. In some cases, the first target can also be further refined. At this time, the first data can be divided to support the detection of different first targets, etc., which will not be repeated here.

[0096] In step 103, the first target is detected based on the first data. As previously described, there are many different possible scenarios for the first data. Therefore, the implementation methods for target detection based on the first data in different scenarios also vary. For example, when the first data is channel data, the detection of the first target also requires the acquisition of RD spectrum data, but when the first data is RD spectrum data, no processing is required to acquire RD spectrum data. The different implementation methods for first target detection based on the first data in different scenarios will not be elaborated here.

[0097] It should be noted that the FFT, CFAR and other processing that may be involved in the process of detecting the first target are roughly the same as those of existing target detection, and will not be described in detail here.

[0098] Step 104: Detect the second target based on the second data. Similar to the detection of the first target described above, since the second data is generated from multiple first data, the second data will vary depending on the first data. Therefore, the detection of the second target will also vary. However, the FFT, CFAR, and other processing involved in the target detection process are generally similar to those of existing target detection and will not be further described here.

[0099] It is also understandable that since the second target is a target that requires a longer observation time, some specific digital signal processing can be used to improve the target detection results. For example, in some embodiments, as shown in FIG5 , based on the second data, the second target can be detected by the following steps:

[0100] Step 504: Perform slow time dimension FFT based on the second data.

[0101] Step 505 : Perform constant false alarm detection based on the FFT result to achieve detection of the second target.

[0102] Among them, steps 501 to 503 in Figure 5 are roughly the same as steps 101 to 103 in the above embodiment, and will not be described in detail here.

[0103] In this way, through the slow time dimension FFT, targets with different motion states, such as stationary targets and moving targets, can be distinguished. Therefore, the above embodiment can effectively filter out the first target through the slow time dimension FFT, extract the effective second target, and thus provide more precise detection for the second target that requires long-term observation, thereby achieving more accurate target detection results.

[0104] Of course, when detecting the first target, a slow time dimension FFT may also be used, so as to accurately extract the first target, thereby facilitating more accurate detection of the first target.

[0105] That is to say, obtaining RD spectrum data through slow time dimension FFT actually provides a processing method that is adapted to the dynamic characteristics of the corresponding first target and / or second target to obtain RD spectrum data, which can improve the accuracy of RD spectrum data, and thereby improve the accuracy of target detection based on the RD spectrum data, that is, it is beneficial to improve the accuracy of detection of the first target and / or second target.

[0106] Among them, from the FFT result to the constant false alarm detection, coherent integration (Coherent Integration) or non-coherent integration (Non-Coherent Integration) is usually required, so as to improve the signal-to-noise ratio by accumulating multiple pulses. It should be noted that the embodiment of the present application does not limit the specific implementation method of coherent accumulation and non-coherent accumulation. It can selectively perform coherent accumulation or non-coherent accumulation according to application scenarios, user needs, etc., and can also select the accumulation algorithm according to computing power, resources, scenarios and needs, etc., which will not be described in detail here.

[0107] It is also understandable that a typical CFAR algorithm typically uses a higher threshold, which can result in more target points being considered non-targets. Based on this, in some embodiments, the above embodiment can perform constant false alarm detection based on a preset threshold when performing target detection, where the threshold = k * noise estimate, where k > 1.

[0108] It should be noted that, compared to the current practice of setting the threshold to k'*noise estimate to reduce clutter interference, where k' is typically set to tens or hundreds, the k>1 limit adjustment allows for a smaller threshold value during target detection, eliminating the requirement for a higher threshold. This smaller threshold allows CFAR detection to detect more points, reducing or avoiding the premature exclusion of valid targets as clutter, thereby improving the accuracy of indoor target detection.

[0109] In some cases, the value range of k is (1, 4), such as k=2, k=2.4, k=2.5, k=3, etc. In this way, the value of k is further set to a smaller range, thereby further lowering the threshold. Therefore, with a smaller threshold, the constant false alarm detection can detect more points, further reducing or avoiding the misidentification of valid targets as noise points for elimination, which is conducive to further improving the accuracy of indoor target detection.

[0110] Of course, the above is only an example of the value range of the parameter k in the threshold. The value of the parameter k in the threshold can also be adjusted according to specific scenarios, which will not be described in detail here.

[0111] It's also understandable that in relatively enclosed areas, such as indoors, target detection accuracy is low due to the greater complexity of people and obstacles compared to more open outdoor environments. Improving target detection accuracy in such scenarios is crucial. Furthermore, clutter inevitably occurs in space, interfering with target detection and potentially detecting invalid targets.

[0112] To this end, embodiments of the present application also provide a target detection method that provides confidence levels for detected targets, thereby avoiding both the problem of misjudging targets as non-targets and the problem of inability to accurately obtain target information. In some embodiments, the target detection method process is shown in Figure 6 and includes the following steps:

[0113] Step 601: obtain several targets based on the target detection results.

[0114] Step 602: For each detected target, corresponding feature information is obtained, where the feature information includes noise features and signal features used to characterize the target.

[0115] Step 603: Determine the confidence level of each target based on the feature information.

[0116] In this way, since the feature information includes noise features and signal features used to characterize the target, after the target is detected, it is possible to determine whether the target is a valid target and determine the confidence level of each detected target through the feature information, thereby eliminating the erroneous situation of detecting noise as a target, so that the target can be further verified, which is conducive to improving the reliability and accuracy of target detection, especially improving the accuracy of indoor target detection.

[0117] In order to facilitate those skilled in the art to better understand the above embodiments, they will be explained below.

[0118] In step 601, a number of targets are obtained based on the target detection results. This embodiment does not limit the method for obtaining the target detection results, which can be achieved by any target detection method, and will not be described in detail here.

[0119] Step 602: For each detected target, corresponding feature information is obtained. The feature information includes noise and signal characteristics used to characterize the target. In this embodiment, the feature information is not limited and can be any information that can characterize the noise and signal characteristics of the target. For example, in some examples, the feature signal can also be any information that can reflect the distribution of the signal reflected by the target in the received echo signal. For example, the feature information can include an angle spectrum, entropy value, signal-to-noise ratio, eigenvalues ​​determined based on the covariance matrix of the echo signal, etc. The angle spectrum can also be an RD spectrum, a Capon spectrum, etc. These are not listed here. The angle spectrum can be obtained during the target detection process when the target detection results are obtained. The algorithms used to generate the angle spectrum may include: beamforming algorithms, Capon algorithms, multiple signal classification algorithms, rotationally invariant subspace algorithms, orthogonal matching pursuit algorithms, etc., which are not detailed here. The entropy value can be obtained directly, and the signal-to-noise ratio can be obtained during the CFAR process when the target detection results are obtained.

[0120] It should be noted that the way in which the signal features and noise features in the feature information represent the target is not limited here, and may vary according to different feature information, and may also vary according to different scenarios, requirements, etc.

[0121] For example, when the characteristic information includes an energy spectrum, the entropy of the energy spectrum can be used as characteristic information, based on the characteristic that signals are generally concentrated while noise is generally scattered. When the characteristic information includes an angle spectrum, based on the characteristic that signals generally have greater energy, mainly reflected in the main lobe, while noise generally has less energy, mainly reflected in the side lobes, the relevant information of the main lobe and side lobes in the angle spectrum can be used as characteristic information to characterize the signal and noise characteristics of the target. For another example, when the characteristic information includes an angle spectrum, the characteristic information can be related to the local noise characteristics and local signal characteristics of the target, or it can be related to the global noise characteristics and global signal characteristics of the target, etc., which will not be elaborated here.

[0122] It is understood that when the feature information includes more noise-related content than signal-related content, it indicates that the signal described by the feature information is more likely to be a noise signal (i.e., clutter). Therefore, based on the signal and noise features of the target, it can be determined whether the target's feature information contains more noise-related content or more signal-related content, thereby determining whether the target is a valid target.

[0123] It can also be understood that the standards for determining confidence levels may vary depending on the scenario, requirements, etc. For example, when the noise in the environment is complex, there may be multiple superpositions resulting in high noise levels. In this case, the reference standard for confidence levels may be set to a broader condition. This can reduce the risk of misjudging valid targets.

[0124] Step 603 determines the confidence level of each target based on the feature information. It is understood that clutter is inevitable in space, and this clutter can interfere with target detection, potentially causing invalid targets to be detected. Feature information includes both noise and signal characteristics used to characterize a target. Therefore, the target can be analyzed based on this feature information to determine whether it is a valid target, i.e., to determine its confidence level.

[0125] It should be noted that, as previously mentioned, there are many different possible scenarios for feature information, and accordingly, there are also many different ways to implement confidence. To facilitate a better understanding for those skilled in the art, the following examples illustrate how to obtain target confidence when the feature information is angle spectrum, entropy value, and signal-to-noise ratio.

[0126] When the characteristic information is an angle spectrum, in some embodiments, as shown in FIG7 , obtaining the corresponding characteristic information can be achieved by the following steps:

[0127] Step 702: Acquire the angle spectrum generated during the target detection process.

[0128] Accordingly, the confidence level of each target can be determined based on the feature information by the following steps:

[0129] Step 703: Determine the maximum value in the angle spectrum corresponding to the target.

[0130] Step 704: Determine a first validity parameter based on the maximum value among the maximum values ​​of the angle spectrum and the average value of the other maximum values ​​except the maximum value. The first validity parameter is used to characterize the relative size between the maximum value among the maximum values ​​of the angle spectrum and the average value of the other maximum values ​​except the maximum value.

[0131] Step 705: Determine the confidence level of each target based on the first validity parameter.

[0132] Among them, step 701 is substantially the same as that of the aforementioned embodiment, and will not be described in detail here.

[0133] It should be noted that this embodiment does not limit the specific method for determining the first effective parameter. In some examples, it can be the ratio of the maximum value among the maximum values ​​of the angle spectrum and the average of other maximum values ​​except the maximum value; in some examples, it can also be an array composed of the maximum value among the maximum values ​​of the angle spectrum and the average of other maximum values ​​except the maximum value, etc.

[0134] As previously mentioned, the maximum value among the maxima in the angle spectrum can correspond to signal-related content, while other maxima besides the maximum value can correspond to noise-related content. Therefore, the first validity parameter actually reflects whether the global signal distribution is concentrated, and thus can reflect the relative energy distribution. Based on this global relative energy distribution, whether a target is a valid target can be more comprehensively and accurately determined, that is, the target confidence level can be accurately determined, which helps to further improve the accuracy of target detection.

[0135] It should be noted that the "mean" in the mean of other maximum values ​​except the maximum value can be the average value, median or mode, etc., which will not be described one by one here.

[0136] In some embodiments, as shown in FIG8 , determining the confidence level of each target based on the first validity parameter can be achieved by the following steps:

[0137] Step 805: Determine a second validity parameter of the corresponding target based on the maximum value and the second maximum value in the angular spectrum. The second validity parameter is used to characterize the relative size between the maximum value and the second maximum value in the angular spectrum.

[0138] Step 806: Determine whether the corresponding target is an invalid target based on the second validity parameter.

[0139] Step 807: When the corresponding target is not an invalid target, the confidence level of the target is determined according to the first validity parameter.

[0140] Among them, steps 801 to 804 are substantially the same as steps 701 to 704 in the aforementioned embodiment, and are not described in detail here.

[0141] It should be noted that this embodiment does not limit the specific method of determining the second validity parameter. It can be the ratio of the maximum value and the second largest value among the maximum values ​​of the angle spectrum, or the difference between the maximum value and the second largest value among the maximum values ​​of the angle spectrum, etc.

[0142] As previously mentioned, the maximum value among the maxima of the angle spectrum can correspond to signal-related content, and other maxima besides the maximum value can correspond to noise-related content. Therefore, the second validity parameter actually reflects whether the local signal distribution is concentrated. Furthermore, compared to the first validity parameter, the second validity parameter can also reflect the relative energy distribution from another perspective. Based on this local relative energy distribution, combined with the overall relative energy distribution, whether the target is a valid target can be determined from different dimensions, which is conducive to further improving the accuracy of the confidence level, thereby improving the accuracy of target detection.

[0143] In some embodiments, obtaining the 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.

[0144] It should be noted that the method for determining the entropy value is not limited here. Any algorithm that can quantify the degree of confusion of feature information can be used to obtain the entropy value, and will not be described in detail here.

[0145] It can be understood that for a valid target (i.e., a real target), the corresponding echo signal should be concentrated, so the entropy value is small, while for an invalid target (i.e., a false target, such as a target detected based on clutter), the corresponding echo signal is relatively scattered, so the entropy value is large. In other words, the relative size of the entropy value can be used to determine whether the detected target is a valid target, thereby determining the target's confidence.

[0146] It should be noted that the entropy value range corresponding to the target is not limited here, and it may vary according to the degree of chaos in the detection environment, the requirements of detection accuracy, etc., and will not be elaborated here.

[0147] In some embodiments, obtaining the corresponding feature information can be achieved by obtaining a signal-to-noise ratio (SNR) of the corresponding target generated by constant false alarm detection during target detection. 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 SNR.

[0148] It can be understood that the signal-to-noise ratio is an accurate quantification of the signal part and the noise part in the characteristic information corresponding to the target. Based on the signal-to-noise ratio, the relative size of the signal part and the noise part in the characteristic information can be accurately determined, so that it can be accurately judged whether the detected target is a valid target, which is conducive to further accurately determining the confidence level of the target, thereby achieving accurate target detection.

[0149] Of course, the above embodiments can also be combined. For example, in some embodiments, the present application provides a target detection method, the process of which is shown in FIG9 and includes the following steps:

[0150] Step 901: Perform target detection to obtain a first target and a second target.

[0151] Step 902: For each of the detected first target and second target, corresponding feature information is obtained, where the feature information includes noise features and signal features for characterizing the target.

[0152] Step 903: Determine the confidence level of each first target and second target based on the feature information.

[0153] It should be noted that, in this embodiment, step 901 is implemented by any of the above target detection methods that can detect the first target and the second target, and steps 902 and 903 are roughly the same as steps 602-603 in the aforementioned embodiment, and will not be repeated here.

[0154] In some embodiments, the present application also provides a target detection method, the process of which includes the following steps:

[0155] 1. Perform constant false alarm detection on the Doppler spectrum matrix according to a preset threshold to detect the target. The threshold is k*noise estimate, where k>1.

[0156] 2. Generate corresponding feature information for each detected target.

[0157] 3. Determine the confidence level of the target based on the corresponding feature information to further confirm whether the target is a valid target.

[0158] Among them, the above steps 2 and 3 are roughly the same as steps 602 and 603 in the above embodiment, and will not be described in detail here.

[0159] It should be noted that the above step 1 can also be replaced by any of the aforementioned solutions that can perform target detection and obtain the detected target, for example, introducing inter-frame FFT in the process of generating RD spectrum data, etc., which will not be described here one by one.

[0160] It can also be understood that in the aforementioned embodiment, a first target and a second target are detected for the target motion state (i.e., point cloud type). That is, in the time dimension, for the same period of time, the corresponding target object can be divided into a first target and a second target, and can even be further subdivided, for example, according to the motion state generated, it can be subdivided into a motion target, a short micro-motion target, and a long micro-motion target. Among them, in the detection of motion targets, based on the amplitude of the target motion, it can be divided into a "motion" state and a "micro-motion" state, that is, the amplitude of "motion" is greater than the amplitude of "micro-motion"; at the same time, the "micro-motion" state can be divided into at least two states, "long micro-motion" and "short micro-motion" based on the time dimension, and the time period of "long micro-motion" is greater than the time period of "short micro-motion".

[0161] However, existing systems based on preset frame data generally only output one type of detection target, such as a dynamic point cloud or a micro-motion point cloud. When the radar system needs to output two types of detection targets, it can only output them alternately, that is, one preset frame data is used to output a dynamic point cloud, and the other preset frame data is used to output a micro-motion point cloud. Furthermore, micro-motion point data is generally detected based on a single frame or multiple frames.

[0162] Therefore, the present application also provides a method for outputting a target point cloud, which is used to display the above-mentioned different targets in a hierarchical manner to improve the user experience. In some embodiments, the process is shown in Figure 10 and includes the following steps:

[0163] Step 1001: Perform target detection to determine a first target and a second target.

[0164] Step 1002: Output the first target and the second target in a hierarchical manner.

[0165] Among them, the first target and the second target determined in step 1001 can be the first target and the second target determined according to the target detection method provided in the aforementioned embodiment, and the first target and the second target determined in step 1001 can also be the first target and the second target with confidence determined according to the target detection method provided in the aforementioned embodiment.

[0166] Regarding step 1002, in some embodiments, as described above, targets are subdivided into moving targets, short-micromotion targets, and long-micromotion targets based on the generated motion states. Assuming that these three types of targets are all acquired based on data from a preset group of frames, the data from the preset group of frames is used as a data unit for long-micromotion detection, while at least two (e.g., two) frames contained in the preset group of frames are used as a data unit for short-micromotion detection, and the data from each frame can be used for motion detection. Furthermore, when, for example, six consecutive frames of data are used as a data unit for long-micromotion detection, three consecutive frames of data can be selected as a data unit for short-micromotion detection, and two consecutive frames of data can be selected as a data unit for motion detection. The term "frame" refers only to the smallest data unit used for target detection, and does not explicitly refer to a single frame. For example, the data from one frame can be 1 / 2 (or 1 / 3, 0.2, 1, 2.5, 3, 5, etc.) times the data of a typical frame.

[0167] Regarding step 1002, in some embodiments, when detecting motion point clouds based on single-frame data, point cloud data can be output for each frame using a sliding window method for short micro-motion points, further supplementing the detected dynamic (motion state) points. Specifically, by outputting dynamic points and short micro-motion points for each frame, the overall posture and partial movements of a human body can be more accurately and comprehensively portrayed, achieving an approximate imaging effect (relatively coarse-grained imaging).

[0168] That is to say, through the method provided by the above embodiment, the scene can be perceived based on the echo signals at different time scales. At the same time, by setting the corresponding confidence level for the point cloud according to preset rules, the hierarchical output of the point cloud data can be achieved, so as to achieve a more comprehensive scene display and the detection of the target's all-round status information. At the same time, based on the echo signal of a preset unit, at least two types of detection target data can be output simultaneously, that is, at least two types of detection target results can be obtained based on the data of the same preset unit. The data of the preset unit can be a preset number of adjacent chirp echo signals within a frame (such as half a frame, 1 / 4 frame, etc.), or it can be a frame of data or at least two adjacent frames of echo signals. For example, when the data of the preset unit is three adjacent frames of echo signals, three set types of dynamic point clouds are obtained based on the three adjacent frames of data as the data source. For example, the first type of dynamic point cloud (such as a short-term micro-motion point cloud) can be output based on two adjacent frames of data, and the second type of dynamic point cloud (such as a motion point cloud) can be output based on each frame of data. In addition, the third type of dynamic point cloud (such as a long-term micro-motion point cloud) can be output based on the three adjacent frames of data. At the same time, different colors, shapes, etc. are set in the point cloud presentation to achieve a hierarchical display of the three types of point cloud data.

[0169] To facilitate those skilled in the art to better understand the methods and combinations provided in the above embodiments, the following examples are provided. For ease of understanding, the first target is a motion target, and the second target is a micro-motion target. The second target can be further refined into a long-term micro-motion target and a short-term micro-motion target.

[0170] For a multi-channel 4D FMCW millimeter-wave radar installed in a room (such as a roof or wall), the radar transmits electromagnetic waves to targets in the room in frames. Each frame signal can contain several chirp signals, and the frame period or the total transmission time of the chirp within a frame is longer than that used by traditional radars. For example, the total transmission time of the chirp within a frame can be increased from 20ms to 50ms or more. At the same time, the radar receives the echo signals generated by the transmitted electromagnetic waves in frames and processes the received signals as follows:

[0171] Step S1101 : performing range windowing and range FFT on each chirp signal in the frame to obtain original channel data.

[0172] Step S1102 : Calculate zero Doppler channel data of the signal within the frame based on the original channel data within the frame and save the data.

[0173] Step S1103: Subtract the zero Doppler channel data from each chirp in the frame according to the range bin.

[0174] Step S1104: performing windowing and FFT between chirps on the data obtained in step S1103.

[0175] In step S1105, the data obtained in step S1104 of all channels are modulo-calculated and the results are accumulated.

[0176] Step S1106: Use the CFAR algorithm to perform target detection on the result in step S1105.

[0177] In step S1107 , the range bins and Doppler bins of all azimuth channels corresponding to the target detected in step S1106 are extracted, and DBF is performed and modulo is obtained to obtain the DBF spectrum of the target.

[0178] Step S1108 : Find the maximum value of the DBF spectrum of each target, and calculate the azimuth angle corresponding to the maximum value as the azimuth angle of the target.

[0179] Step S1109 : Calculate the confidence of the target by the concentration degree of the DBF spectrum to obtain the output of the dynamic point.

[0180] In step S1110 , the data obtained in step 1103 for nearly four frames are used to form a micro-motion frame, and steps S1104 to S1109 are performed to obtain short-term micro-motion points.

[0181] Among them, 4 frames can also be replaced by a duration of 1 second. It should be noted that the duration of 1 second can be adaptively adjusted according to actual needs, and the embodiment of the present application is not limited to it, and can also be 0.8 seconds, 0.5 seconds or 0.3 seconds, etc.

[0182] In step S1111, the data obtained in step 1103 for nearly 16 frames are used to form a fine motion frame, and the processing of steps S1104 to S1109 is performed to obtain long-term fine motion points.

[0183] Among them, 16 frames can also be replaced by a duration greater than 1 second. The specific duration can be adaptively adjusted according to actual needs and is not limited in the embodiment of the present application, such as 5 seconds, 3 seconds or 1 second.

[0184] Step S1112: Output the dynamic points, short-term fine-motion points, long-term fine-motion points and their confidence levels. The dynamic points, short-term fine-motion points and long-term fine-motion points are displayed in a hierarchical manner to facilitate more intuitive observation by the user.

[0185] In some embodiments, based on the above-mentioned target detection method embodiment, the RD matrix for long-term micro-motion points may not be recalculated, but instead may be obtained by cumulatively extrapolating the RD matrix of short-term micro-motion points. The RD matrix of short-term micro-motion points is obtained using a sliding window method, thereby enabling output for each frame. At the same time, the RD matrix of long-term micro-motion points may also be obtained using a sliding window method, or may be obtained by processing once every preset number of frames, or may be obtained based on the accumulated RD matrix of short-term micro-motion points. When obtaining the RD matrix of long-term micro-motion points based on the accumulated RD matrix of short-term micro-motion points, the RD matrix of long-term micro-motion points may be obtained by performing phase compensation on the frame corresponding to the RD matrix of each short-term micro-motion point, and then accumulating the RD matrix.

[0186] In some embodiments, different time scales of the scene can also be perceived by using different numbers of frames. For example, the radar is set to at least two scenes such as daytime (or active or living room) scene or nighttime (or inactive or bedroom) scene. The above-mentioned divided states can be set based on different time scales for different scenes, that is, the time scales of different scenes and the same state are different. For example, corresponding to the daytime living room scene, the time scale of short micro-motion can be set to 0.8s, and the time scale of long micro-motion can be set to 3~5s. Corresponding to the night bedroom scene, the time scale of short micro-motion can be set to 1~2s, and the time scale of long micro-motion can be set to 10~100s, etc., so as to achieve adaptive adjustment of the scene while effectively reducing the power consumption of the radar.

[0187] In this way, all possible potential target points in the RD spectrum data can be detected using a very low threshold noise reference constant false-alarm rate (NRCFAR) method. The DBF spectrum is then used to filter possible target points from a large number of candidate points. This method differs from conventional RD processing in that the formation of the final point cloud depends on how the point cloud is filtered rather than how the RD spectrum data is detected.

[0188] That is, conventional RD processing forms RD spectrum data after performing 2DFFT, and then detects target points from the RD spectrum data based on CFAR, and calculates their azimuth and pitch angles to form a point cloud. However, the indoor environment is relatively complex, with a large number of reflective objects (walls, tables, chairs, furniture, etc.), which leads to a large number of multi-angle echoes at the same distance (mostly from reflections of the scene) and the existence of multipath phenomena. That is, among the points detected by conventional RD processing, some are targets of interest, some are multipath points, and some are false points after the synthesis of multi-source reflections. Therefore, the results of conventional RD processing have a high number of noise points. On the other hand, in order to suppress noise points, conventional RD processing methods use a higher detection threshold, which also suppresses points with relatively weak signal-to-noise ratios from real targets, resulting in fewer point clouds of real targets (when put together with noise points, the points appear scattered and not clustered) or no points on targets with low signal-to-noise ratios.

[0189] For a point originating from a real target of interest, due to its single signal component, its DBF spectrum inevitably resembles a sinc function, with a distinct main peak. However, for multipath points, due to the difference between the outgoing and incident phases, after Multiple-Input Multiple-Output (MIMO) processing, the DBF spectrum of the multipath point fails to form a main peak similar to that of a real target, resulting in a relatively dispersed DBF spectrum. Similarly, for a false point resulting from the synthesis of multiple reflection sources scattered throughout a room, its DBF spectrum also fails to form a distinct main peak. Therefore, the DBF spectrum can be used to screen candidate points.

[0190] In indoor applications, human targets differ from typical radar targets in that the velocities of different parts of the body vary, meaning that the human echo contains a variety of velocity components. Furthermore, the human body exhibits a variety of motion states, including tangential motion, slight swaying while standing, and other low radial velocity states. These motion states pose a challenge to radars that rely on radial velocity for target separation. Furthermore, the radial velocity of a point cloud of the human body only reflects the radial velocity of a specific part of the body and may differ significantly from the radial component of the person's actual velocity. In this scenario, the unambiguous velocity range has no substantial impact on target detection. Therefore, using a long chirp period for high-resolution velocity detection of indoor targets can effectively improve the comprehensiveness of target detection.

[0191] In addition, improving the speed detection resolution has at least the following advantages:

[0192] 1) A more refined speed dimension characterization of the target can effectively improve the signal-to-noise ratio of each speed unit.

[0193] 2) The echoes from two targets at the same distance are differentiated more finely in the velocity dimension, which enables simultaneous detection of two targets at the same distance.

[0194] 3) Improve the radar's sensitivity to the scene, that is, improve the detection capability of low radial velocity states such as tangential motion, which has high practical value in indoor scenes.

[0195] 4) Make full use of the number of points in the velocity dimension of RD spectrum data to avoid wasting storage space.

[0196] At the same time, the detection of slightly moving targets is processed using multiple frames of intra-frame zero-Doppler channel data. Since the above-mentioned dynamic point processing method can effectively detect dynamic points based on RD spectrum data, a similar processing approach can be used for slightly moving targets. The intra-frame zero-Doppler channel of multiple frames is subjected to inter-frame FFT to generate inter-frame RD spectrum data. Based on this RD spectrum data, the above-mentioned dynamic point processing method is used to form the final slightly moving point cloud.

[0197] In some embodiments, in order to achieve a more complete perception of the scene, micro-motion processing is divided into short-term micro-motion processing and long-term micro-motion processing. Short-term micro-motion point processing can use fewer frames. Compared with in-frame processing, short-term micro-motion processing can effectively detect sitting and standing still people. Short-term micro-motion points are an effective supplement to dynamic points in the frame. The two complement each other in the real-time perception of the scene. The posture and movement of the human body can be jointly portrayed by dynamic points and short-term micro-motion points. Long-term micro-motion point processing uses more frames to perceive more subtle scene changes, thereby assisting short-term micro-motion points in detecting static people.

[0198] That is, in the process of dynamic point detection, through:

[0199] a. Use signals obtained at different time scales for processing to fully perceive the moving and static targets in the scene, and simultaneously output dynamic points, short-term micro-motion points, and long-term micro-motion points.

[0200] b. Use high-resolution velocity waveforms in dynamic point processing to more precisely characterize the scene in the velocity dimension and expand the "pool" that accommodates the target point cloud.

[0201] c. Use low-threshold NRCFAR to detect in the RD domain of the scene echo, detect all "bright spots" in the RD domain, and do not miss any possible target echo point.

[0202] d. Use the DBF spectrum to evaluate the detected candidate point set, and select the point cloud that meets the target features from the candidate point set as the final point cloud output.

[0203] Through the above improvements, it is possible to achieve:

[0204] 1. For human targets, the system outputs a rich point cloud that can depict the posture of the human body in both static and dynamic states, and perform point cloud tracking (coarse imaging) of human movements (such as waving).

[0205] 2. Filter out a large number of clutter points and multipath points that are widely present in indoor environments, while retaining effective weak target points to improve the radar's weak target detection capability.

[0206] 3. Improve the point cloud cohesion of human targets, thereby improving the ability to distinguish between human targets.

[0207] 4. A large number of micro-motion point clouds can be output for both distant and near stationary human targets.

[0208] 5. It can still continuously output dynamic point clouds for tangentially moving targets.

[0209] In some embodiments, the signal processing process can be divided into dynamic point processing and micro-motion point processing (i.e., the "static" point processing described later in the embodiments of this application). Dynamic point processing can be intra-frame processing. It is mainly used to detect dynamic points with significant radial velocity. Micro-motion point processing is inter-frame processing. By extending the observation time, the velocity resolution is further improved to perceive subtle movements of the human body. Micro-motion point processing is further divided into short-term micro-motion point processing and long-term micro-motion point processing.

[0210] In some optional embodiments, the processing of dynamic points can be intra-frame processing, that is, using echo data of multiple chirps in one frame to detect dynamic targets. Specifically, the following steps may be included:

[0211] Step S1201: distance windowing and distance FFT.

[0212] In some examples, the ADC (Analog to Digital Converter) data of each received chirp is subjected to range windowing and range FFT to obtain a one-dimensional range profile. This operation may or may not be performed with windowing, and different windows can be used for processing. Furthermore, for pulse compression radars, pulse compression is performed on each chirp echo signal to obtain a one-dimensional range profile. In some cases, a stepped frequency radar can also be used in this step. In other words, the embodiments of the present application are not limited to the radar mechanism, as long as it can obtain a one-dimensional range profile.

[0213] Step S1202, remove zero Doppler.

[0214] In some examples, the one-dimensional range image results of each chirp are accumulated by range bin and averaged to obtain the zero-Doppler channel data within the frame. The zero-Doppler channel data within the frame is then subtracted from each chirp to filter out stationary targets within the frame and retain moving targets within the frame.

[0215] Step S1203: velocity windowing and velocity FFT.

[0216] In some examples, the Doppler-de-zeroed data is windowed (or not windowed) along the slow time dimension by range bin, and a slow time dimension FFT is performed to separate echoes of the same range in the scene in the velocity dimension.

[0217] Step S1204 : Incoherent accumulation of RDM (Range-Doppler Matrix, also known as the aforementioned RD spectrum data) between channels.

[0218] In some examples, the modulus square of the velocity FFT results of each channel is calculated, and the results are accumulated and averaged across the channels to obtain incoherent gains between the channels, thereby obtaining an RDM for target detection.

[0219] In some cases, multiple RDMs can be formed through incoherent accumulation of multiple channels or coherent accumulation, and then tested on the multiple RDMs. It is also possible to directly square the modulus of one or several channels for testing without accumulation.

[0220] Step S1205: perform NRCFAR.

[0221] In some examples, the first 10 distance bin data (within a range of about 1m) of the 20 Doppler channels near the zero Doppler channel are extracted from the detection RDM, averaged, and the noise mean is calculated. Each point on the detection RDM is compared with the above noise mean, and if it is greater than a certain threshold, the point is included in the candidate point. In order to detect all weak echo points, the threshold set here can be obtained by modulus or modulus square, that is, a relatively low threshold (compared to the threshold value of the traditional method) is used to detect as many candidate points as possible, for example, it can be set to 10 times by modulus square. Optionally, the noise can also be selected by selecting several distance bins at the farthest distance to calculate the noise mean, that is, it is mainly based on which range is relatively clean, with only noise but no target, so as to obtain a more accurate noise floor.

[0222] Step S1206: perform DOA (Direction of Arrival) estimation.

[0223] In some examples, based on the TRX array, DBF is performed on the master and slave sub-array channel data of the candidate point to obtain the azimuth DBF spectrum. The elevation DBF is then calculated by combining them to obtain the elevation angle.

[0224] Step S1207: Calculate the point cloud confidence.

[0225] In some cases, if the energy of a candidate point comes solely from the target, its signal composition is relatively simple, and the DBF spectrum approximates an ideal sinc function, with a distinct main peak. Therefore, a confidence level based on the candidate point's orientation DBF spectrum can be used to determine whether the candidate point is the desired target point. Confidence levels are categorized as Confidence Level 1 or Confidence Level 2.

[0226] Among them, the calculation method of confidence 1 (conf1) can be:

[0227] conf1=max / peak2,

[0228] Here, max is the maximum value of the azimuth DBF spectrum, and peak2 is the second-largest value of the azimuth DBF spectrum. That is, confidence level 1 reflects the highest sidelobe level of the DBF spectrum. A higher confidence level 1 indicates a lower highest sidelobe, a more uniform signal component, and a higher likelihood of originating from a single target.

[0229] The confidence level 2 (conf2) can be calculated as:

[0230] conf2=max / mean(peakoth),

[0231] Here, max is the maximum value of the azimuthal DBF spectrum, peakoth is the maximum and second-largest extreme values ​​of the azimuthal DBF spectrum, and mean() represents the mean. Therefore, confidence level 2 reflects the degree of energy concentration in the main peak of the DBF spectrum.

[0232] It should be noted that in the embodiment of the present application, the confidence calculation method is not unique, as long as it can measure the energy concentration of the DBF spectrum, that is, the higher the concentration, the purer the echo is and the more likely it is to be a valid target.

[0233] In some cases, the confidence calculation in the embodiments of the present application may be based on the azimuth DBF spectrum, but may also be obtained using the elevation DBF spectrum, or using non-azimuth and non-elevation DBF spectra.

[0234] Step S1208: extracting a single target point from a valid target.

[0235] In some examples, as mentioned above, the higher the confidence 1 of a candidate point, the more likely it is from a single target. Therefore, a certain threshold is set for the confidence 1. If it is higher than the threshold, the target is output as a single target point cloud. The current judgment conditions for a single target point are:

[0236] conf1>2.

[0237] Step S1209: extracting dual target points from valid targets.

[0238] In some cases, in indoor scenes, it is inevitable that two human targets will be in the same distance unit. If the echoes of the two targets fall into the same RD unit, when the two targets are sufficiently separated in azimuth, two close peaks will appear in the azimuth DBF spectrum. To prevent the candidate points in this situation from being screened out by the "Single Target Point Screening" module, confidence 1 and confidence 2 are used to determine the dual targets. If confidence 1 and confidence 2 meet the following conditions at the same time, the main lobe and highest sidelobe of the candidate point DBF spectrum are output as dual targets:

[0239] conf1<1.5,conf2>2.5。

[0240] Step S1210, supplementing the point cloud.

[0241] In some examples, for point clouds that do not meet the single or dual target point judgment, if the following conditions are met, they will be output as supplementary point clouds.

[0242] conf1>1.5, conf2>3, SNR>10.

[0243] Step S1211: point cloud enhancement.

[0244] In some cases, for "high-quality" point clouds, i.e., point clouds with concentrated DBF spectrum energy and high signal-to-noise ratio, the left and right azimuth points of the main peak of the DBF spectrum are output simultaneously, giving "high-quality" point clouds more weight in the number of point clouds, thereby enhancing the concentration of scene point clouds. The current criteria for judging "high-quality" point clouds are:

[0245] conf2>4.5 and SNR>30.

[0246] In some embodiments, in the processing of micro-motion points, the micro-motion point processing can be divided into short-term micro-motion points and long-term micro-motion points. Both can be cross-frame processing operations. For example, short-term micro-motion points are processed using 4 frames, and long-term micro-motion points can be processed using 16 frames of duration data (16 frames are extracted from two frames, and the final number of frames used is 8 frames). Except for the difference in the number of frames used, the processing procedures of the two are basically the same. In order to avoid the interference of moving targets on the detection of micro-motion targets, both use zero Doppler channel data within the frame as input. Specifically, the following steps may be included:

[0247] Micro-motion point processing is divided into short-term micro-motion point processing and long-term micro-motion point processing. Both are cross-frame processing. Short-term micro-motion point processing uses 4 frames. Long-term micro-motion point processing uses 16 frames of time data (16 frames are divided into 2 and 1, and the final number of frames used is 8 frames). Apart from the difference in the number of frames used, the processing flow of the two is basically the same. To avoid interference of moving targets on micro-motion target detection, both use zero Doppler channel data within the frame as input. The processing steps of micro-motion points are as follows:

[0248] After zero-Doppler removal in step S1202 and before step S1203, the following steps are performed: 1. Inter-frame zero-Doppler removal; 2. Inter-frame DFT. In some examples, inter-frame zero-Doppler removal involves taking the inter-frame zero-Doppler mean of several frames for each range bin and subtracting this mean from the range bin corresponding to each frame's zero-Doppler data. In some examples, inter-frame DFT involves performing an inter-frame DFT on the zero-Doppler channel data (after subtracting the mean) for each range bin using a sliding window (or non-sliding window) approach to obtain a two-dimensional FFT matrix for each channel.

[0249] In some embodiments, when outputting point clouds, the output point clouds are processed hierarchically in order to fully represent the scene situation from different dimensions, thereby enabling users to accurately understand the meaning, purpose, and usage conditions of various point clouds. Specifically:

[0250] In some embodiments, point cloud data such as dynamic points, short-time micro-motion points, long-time micro-motion points, and static points are displayed hierarchically, namely:

[0251] Because dynamic points reflect moving objects in the scene, they can include moving human bodies with significant radial velocity, as well as those with slight sway and mostly tangential motion. Overall, dynamic points have higher confidence than short-term and long-term micro-motion points. Therefore, they can be used for track initiation and track confidence maintenance.

[0252] Short-term micro-motion points are an effective complement to dynamic points. When a person is in a non-moving state, these points can maintain continuous awareness of the human target and, to a certain extent, reflect the target's posture. Therefore, short-term micro-motion points can be used for continuous tracking of non-moving people.

[0253] At the same time, because long-term micro-motion points use a relatively long frame rate to detect micro-motion in the scene, they can detect even weaker micro-motion targets. Because of the long frame rate, there will be a significant hysteresis in long-term micro-motion points for a dynamic target. That is, after the target moves away from position A, the long-term micro-motion points can still be seen at position A for a period of time. Therefore, long-term micro-motion points can be used to maintain the continuous tracking of stationary targets and track maintenance. For example, if a stationary track no longer has long-term micro-motion points and the confidence level of the stationary track is not high, you may consider deleting the track.

[0254] In some embodiments, for single-target, dual-target, and supplementary point clouds:

[0255] Since dynamic points and long / short time micro-motion points are evenly divided into single target point cloud, dual target point cloud and supplementary point cloud, based on the point cloud confidence angle, we can know that:

[0256] Single target point cloud > dual target point cloud > supplementary point cloud.

[0257] Therefore, for single-target point clouds, the higher the confidence level 1 and the higher the SNR, the more reliable the point cloud. Dynamic single-target point clouds can be used for track initiation, while dual-target point clouds should be used with caution. Supplementary point clouds are not recommended for track initiation. Furthermore, when tracking confirmed tracks, if a single-target point cloud is unavailable, a dual-target point cloud can be conditionally used. That is, if another point in a dual-target point cloud is indeed within the gate of another confirmed track, the confidence level of the dual-target point cloud increases, and the dual-target point cloud can be used to associate the current track. Furthermore, when there are no dynamic points for either single or dual targets around a confirmed track, the use of supplementary point clouds can be considered.

[0258] To facilitate those skilled in the art to better understand the above embodiments, an example will be given below in combination with a target tracking application scenario.

[0259] When dynamic target detection is required, the target detection method may include the following steps:

[0260] Step 1401: Acquire an echo signal, perform windowing on each chirp in the echo signal, and perform range-dimensional FFT based on the windowed signal.

[0261] Step 1402: Determine the mean of each range bin obtained by performing range-dimensional FFT on each chirp in the same frame as the zero-Doppler channel data of the corresponding frame, and subtract the zero-Doppler channel data from the signal of each chirp to remove clutter.

[0262] Step 1403: Window the portion of the range bin in the data after removing clutter from each chirp signal to perform a slow time dimension FFT.

[0263] Step 1404 , modulo the slow time dimension FFT result, and accumulate the results to obtain a Doppler matrix.

[0264] Step 1405: Use the CFAR algorithm to perform target detection on the Doppler matrix.

[0265] In step 1406 , the range bins and Doppler bins of the detected target are extracted, subjected to digital beam forming (DBF) processing, and modulo calculation to obtain the DBF spectrum of the target.

[0266] Step 1407 : Find the maximum value of the DBF spectrum of the target, and calculate the azimuth angle corresponding to the maximum value as the azimuth angle of the target, to obtain the target angle measurement result.

[0267] Step 1408: Calculate all maximum values ​​of the target in the DBF spectrum.

[0268] Step 1409: sort the maximum values ​​calculated for the same target from largest to smallest.

[0269] Step 1410: Calculate the ratio of the maximum value to the second maximum value in the sorting result as the first validity parameter of the target.

[0270] Step 1411, calculate the mean of all maximum values ​​of the target except the maximum value, and calculate the ratio of the maximum value to the mean as the second validity parameter.

[0271] Step 1412: the targets whose first validity parameter is greater than the first preset threshold are classified as valid targets and put into point trace pool 1; the targets whose second validity parameter is greater than the second preset threshold among the remaining target points are classified as potential valid targets and put into point trace pool 2; the remaining targets are classified as invalid targets and put into point trace pool 3.

[0272] Step 1413, during the cluster tracking process of the target, if the target in trace pool 1 meets the tracking requirements, the target in trace pool 1 is used; if the target in trace pool 1 does not meet the tracking requirements, the targets in trace pool 1 and trace pool 2 are used.

[0273] When there is a need for micro-motion target detection, the target detection method may include the following steps:

[0274] Step 1501: Acquire an echo signal, perform windowing on each chirp in the echo signal, and perform range-dimensional FFT based on the windowed signal.

[0275] Step 1502: Determine the mean of each range bin obtained by performing range-dimensional FFT on each chirp in the same frame as the zero-Doppler channel data of the corresponding frame.

[0276] Step 1503: Continue to acquire zero-Doppler channel data corresponding to at least one frame of echo signal.

[0277] Step 1504: Perform FFT on each range bin of the acquired zero Doppler channel data.

[0278] Step 1505 , modulus the FFT result and perform summation between channels to form a Doppler matrix.

[0279] Step 1506: Use k*noise estimation value as a threshold to detect the Doppler matrix and obtain the distance bin and Doppler bin of the candidate point, where k>1.

[0280] In step 1507 , the range bins and Doppler bins of the detected targets corresponding to all azimuth channels are extracted, subjected to DBF processing, and modulo calculation to obtain the DBF spectrum of the target.

[0281] Step 1508 : Find the maximum value of the DBF spectrum of the target, and calculate the azimuth angle corresponding to the maximum value as the azimuth angle of the target to obtain the target angle measurement result.

[0282] Step 1509: Calculate all maximum values ​​of the target in the DBF spectrum.

[0283] Step 1510: sort the maximum values ​​calculated for the same target from largest to smallest.

[0284] Step 1511: Calculate the ratio of the maximum value to the second maximum value in the sorting result as the first validity parameter of the target.

[0285] Step 1512: Calculate the mean of all maximum values ​​of the target except the maximum value, and calculate the ratio of the maximum value to the mean as the second validity parameter.

[0286] Step 1513: the targets whose first validity parameter is greater than the first preset threshold are classified as valid targets and put into point trace pool 1; the targets whose second validity parameter is greater than the second preset threshold among the remaining target points are classified as potential valid targets and put into point trace pool 2; the remaining targets are classified as invalid targets and put into point trace pool 3.

[0287] Step 1514, during the cluster tracking process of the target, if the target in trace pool 1 meets the tracking requirements, the target in trace pool 1 is used; if the target in trace pool 1 does not meet the tracking requirements, the targets in trace pool 1 and trace pool 2 are used.

[0288] In this embodiment, after the target is detected by the echo signal received by the multiple receiving channels of the radar, corresponding feature information is further generated for the detected target, so that it can be determined whether the detected target is a valid target based on the feature information. On the basis of detecting the target, the validity of the target is further judged to reduce the false detection of noise as a target, thereby improving the accuracy of indoor target detection.

[0289] In this embodiment, the threshold is allowed to be set to a smaller value, and the threshold is no longer required to be a higher value. Therefore, a smaller threshold can be passed, so that the constant false alarm detection can detect more points, reduce or avoid valid targets being mistaken for miscellaneous points and excluded, which is conducive to improving the accuracy of indoor target detection.

[0290] In this embodiment, a processing method adapted to the dynamic characteristics of dynamic targets is provided to obtain a Doppler matrix, which can improve the accuracy of the Doppler matrix and further improve the accuracy of target detection based on the Doppler matrix, that is, it is beneficial to improve the accuracy of indoor dynamic target detection.

[0291] The steps of the various methods above are divided only for the purpose of clear description. During implementation, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this patent. Adding insignificant modifications or introducing insignificant designs to the algorithm or process without changing the core design of the algorithm and process are all within the scope of protection of this patent.

[0292] Another aspect of the present application is to provide a computer-readable storage medium storing a computer program that implements the above method embodiment when executed by a processor.

[0293] That is, those skilled in the art will understand that all or part of the steps in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a program, which is stored in a storage medium and includes a number of instructions for causing a device (which may be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., various media that can store program code.

[0294] On the other hand, the embodiment of the present application further provides an integrated circuit, as shown in FIG11 , comprising a radio frequency module 1101 , an analog signal processing module 1102 , and a digital signal processing module 1103 connected in sequence;

[0295] The RF module 1101 is used to generate RF transmit signals and receive RF receive signals.

[0296] The analog signal processing module 1102 is used to down-convert the RF received signal to obtain an intermediate frequency signal;

[0297] The digital signal processing module 1103 is used to perform analog-to-digital conversion on the intermediate frequency signal to obtain digital signal data, and obtain first data, generate second data based on multiple first data, detect the first target based on the first data, and detect the second target based on the second data; and / or, the digital signal processing module 1103 is used to perform analog-to-digital conversion on the intermediate frequency signal to obtain digital signal data, and obtain several targets based on the target detection results; for each detected target, obtain corresponding feature information, the feature information including noise features and signal features used to characterize the target; and determine the confidence level of each target based on the feature information.

[0298] In some embodiments, the integrated circuit may further include a data processing module configured to process digital signals to achieve target detection and / or wireless communication.

[0299] In some embodiments, the integrated circuit may be a millimeter wave chip.

[0300] In some embodiments, the radio frequency receiving signal is an echo signal formed when the radio frequency transmitting signal is transmitted and / or scattered by a target, and the integrated circuit is a sensor chip.

[0301] It is not difficult to find that this embodiment is a circuit embodiment corresponding to the method embodiment, and this embodiment can be implemented in conjunction with the method embodiment. The relevant technical details mentioned in the method embodiment are still valid in this embodiment, and to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the method embodiment.

[0302] In addition, in order to highlight the innovative part of this application, this embodiment does not introduce units that are not closely related to solving the technical problems raised by this application, but this does not mean that there are no other units in this embodiment.

[0303] Another embodiment of the present application relates to a sensor comprising: a carrier, an integrated circuit disposed on the carrier, and an antenna disposed on the carrier, or the antenna and the integrated circuit are integrated into a single device disposed on the carrier. The integrated circuit is connected to the antenna and is configured to process echo signals received by the antenna. The integrated circuit is the integrated circuit provided in the preceding embodiment. The integrated circuit is the integrated circuit provided in any embodiment of the present application.

[0304] When the antenna and integrated circuit are not integrated into a single device, the integrated circuit is connected to the antenna via a first transmission line, which can be a printed circuit board (PCB) trace. The carrier can be a printed circuit board (PCB), such as a development board, a data acquisition board, or a device motherboard, and the details are not detailed here.

[0305] Since the structure and working principle of the integrated circuit included in the sensor have been described in detail in the above embodiments, they will not be repeated here.

[0306] An embodiment of the present application provides a terminal device, which may include: a device body; and a sensor as described above, which is arranged on the device body; wherein the sensor is used for target detection and / or communication to provide reference information for the operation of the device body.

[0307] In some embodiments, the sensor can be disposed outside the device body, in other embodiments, the sensor can be disposed inside the device body, and in other embodiments, the sensor can be disposed partially inside the device body and partially outside the device body. This embodiment of the present application is not limited to this, and the specific determination depends on the specific situation.

[0308] It should be noted that sensors can perform functions such as target detection by transmitting and receiving radio signals, providing measurement information of the detected target to the device itself, thereby assisting or even controlling the operation of the device itself. Examples of measurement information include at least one of relative distance, relative speed, and relative angle.

[0309] In some embodiments, the device body may be a component or product used in fields such as transportation, consumer electronics, monitoring, in-cabin detection, and health care. For example, the device body may be intelligent transportation equipment (such as cars, motorcycles, ships, subways, trains, etc.), security equipment (such as cameras), liquid level / flow rate detection equipment, smart wearable devices (such as bracelets, glasses, etc.), smart home devices (such as sweeping robots, door locks, televisions, air conditioners, smart lights, etc.), various communication devices (such as mobile phones, tablets, etc.), as well as gates, smart traffic lights, smart signs, traffic cameras, and various industrial robotic arms (or robots). It can also be various instruments for detecting vital signs and various devices equipped with the instruments, such as in-cabin detection in cars, indoor personnel monitoring, smart medical equipment, consumer electronic devices, etc.

[0310] In some embodiments, when the above-mentioned device body is applied to an advanced driving assistance system (ADAS), the electromagnetic wave sensor as a vehicle-mounted sensor can provide the ADAS system with various functional safety guarantees such as automatic emergency braking (AEB), blind spot detection warning (BSD), lane changing assist (LCA), and rear cross traffic alert (RCTA).

[0311] In addition, the examples mentioned in the above embodiments can be freely combined, and any combination can be understood as an embodiment. The "embodiment" or "example" appearing in various places in the specification does not necessarily refer to the same embodiment, nor does it mean that it is an independent or alternative embodiment that is mutually exclusive with other embodiments. It will be understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0312] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and that in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A target detection method, comprising: Acquire first data; Generate second data according to the plurality of the first data; Detecting a first target according to the first data; A second target is detected according to the second data.

2. The target detection method according to claim 1, wherein: The second data includes i-type second data and j-type second data, where i and j are integers greater than 1; The step of generating second data according to the plurality of first data comprises: According to Ni pieces of the first data, obtain the i-type second data; Obtaining the j types of second data according to Nj pieces of the first data; Wherein, Ni and Nj are positive integers, and Nj>Ni; The detecting the second target according to the second data includes: Detecting the i-th type of second target according to the i-th type of second data; The j-th type of second target is detected according to the j-th type of second data.

3. The target detection method according to claim 1 or 2, wherein: The first data is channel data; The step of generating second data according to the plurality of first data comprises: The second data is obtained by superimposing a plurality of the first data.

4. The target detection method according to claim 1 or 2, wherein: The step of generating second data according to the plurality of first data comprises: Averaging a plurality of the first data to obtain zero Doppler channel data; The zero Doppler channel data is subtracted from a plurality of the first data and the resultant data are superimposed to obtain the second data.

5. The target detection method according to any one of claims 1 to 4, wherein: The detecting the second target according to the second data includes: Performing a slow time dimension FFT according to the second data; A constant false alarm detection is performed according to the FFT result to achieve the detection of the second goal.

6. The target detection method according to claim 1 or 2, wherein: The first data is RD spectrum data; The step of generating second data according to the plurality of first data comprises: The second data is calculated based on the plurality of the first data.

7. The target detection method according to any one of claims 1 to 6, wherein: The obtaining of the first data comprises: Acquire echo data corresponding to a detection signal having a preset transmission period, wherein the preset transmission period is not less than 50 ms; The first data is generated according to the echo data.

8. The target detection method according to claim 7, wherein: The detection signal includes multiple chirp signals, and the detection signal with the preset transmission period is obtained by at least one of the following methods: increasing the number of chirp signals included in the detection signal, increasing the period of the chirp signal, and increasing the frame idle time between the chirp signals.

9. The target detection method according to any one of claims 1 to 7, wherein: The detection comprises: According to a preset threshold, constant false alarm detection is performed, wherein the threshold=k*noise estimation value, k>1.

10. A target detection method, comprising: According to the target detection results, several targets are obtained; For each detected target, corresponding feature information is obtained, where the feature information includes noise features and signal features used to characterize the target; The confidence level of each target is determined based on the feature information.

11. The target detection method according to claim 10, wherein: The obtaining of corresponding feature information includes: Obtain the angle spectrum generated during target detection; Determining the confidence level of each target according to the feature information includes: Determine a maximum value in the angle spectrum corresponding to the target; Determine a first validity parameter according to a maximum value among the maximum values ​​of the angle spectrum and an average value of other maximum values ​​except the maximum value, wherein the first validity parameter is used to characterize a relative size between the maximum value among the maximum values ​​of the angle spectrum and an average value of other maximum values ​​except the maximum value; The confidence level of each target is determined based on the first validity parameter.

12. The target detection method according to claim 11, wherein: Determining the confidence level of each target according to the first validity parameter includes: Determine a second validity parameter of the corresponding target according to the maximum value and the second maximum value among the maximum values ​​of the angle spectrum, wherein the second validity parameter is used to characterize the relative size between the maximum value and the second maximum value among the maximum values ​​of the angle spectrum; Determining whether the corresponding target is an invalid target according to the second validity parameter; In the case that the corresponding target is not an invalid target, the confidence of the target is determined according to the first validity parameter.

13. The target detection method according to claim 10, wherein: The obtaining of corresponding feature information includes: Get the entropy value of the corresponding target; Determining the confidence level of each target according to the feature information includes: The confidence level of each target is determined based on the entropy value.

14. The target detection method according to claim 10, wherein: The obtaining of corresponding feature information includes: Obtain the signal-to-noise ratio of the corresponding target generated by constant false alarm detection during the target detection process; Determining the confidence level of each target according to the feature information includes: The confidence level of each target is determined according to the signal-to-noise ratio.

15. A target detection method, comprising: According to any one of claims 1 to 9, obtaining a first target and a second target; For each of the first target and the second target detected, obtain corresponding feature information, where the feature information includes a noise feature and a signal feature for characterizing the target; The confidence level of each of the first target and the second target is determined according to the feature information.

16. A target point cloud output method, comprising: According to the target detection method according to any one of claims 1 to 9 or the target detection method according to claim 15, acquiring a first target and a second target; The first target and the second target are output in a hierarchical manner.

17. An integrated circuit comprising a radio frequency module, an analog signal processing module and a digital signal processing module connected in sequence; in, The radio frequency module is used to generate radio frequency transmission signals and receive radio frequency reception signals; The analog signal processing module is used to down-convert the RF received signal to obtain an intermediate frequency signal; The digital signal processing module is used to perform analog-to-digital conversion on the intermediate frequency signal to obtain digital signal data. The digital signal processing module implements the target detection method as described in any one of claims 1 to 9, or implements the target detection method as described in any one of claims 10 to 14, or implements the target detection method as described in claim 15, or implements the target point cloud output method as described in claim 16 based on the digital signal data.

18. A sensor comprising: Carrier; The integrated circuit as claimed in claim 17, disposed on a carrier; An antenna is arranged on the carrier, or the antenna and the integrated circuit are integrated into one device and arranged on the carrier; The integrated circuit is connected to the antenna and is used to process the echo signal received by the antenna.

19. A terminal device, comprising: Equipment body; and a sensor as claimed in claim 18 disposed on the device body; The sensor is used for target detection and / or communication to provide reference information for the operation of the device body.

Images