Fall detection method and device for household health monitoring radar, and medium

Abstract

The invention relates to the technical field of radar signal processing, in particular to a tumble detection method and device for a home health monitoring radar and a medium, which can accurately, conveniently and quickly image a detection scene and calculate RCS information and height approximation information of a target so as to excavate features to judge whether a person tumble or not. According to the method, the RCS amplitude time sequence of the target and the approximate height time sequence of the target are obtained by using the one-dimensional range profile (HRRP) information, feature mining is performed on the RCS amplitude time sequence and the approximate height time sequence, whether the personnel in the core area of the top-mounted radar fall down or not is judged, the technical complexity is low, engineering implementation is easy, and the cost is low. According to the invention, the method can be oriented to the environments such as homes and hospitals with obvious differences, can obtain the activity information of the monitored environment without professional measurement equipment, detects the abnormal falling behavior of the personnel, and achieves the danger alarm.

Description

Technical Field

[0001] The present invention relates to the technical field of radar signal processing, and particularly to a fall detection method, device and medium for a home health monitoring radar. Background Art

[0002] In recent years, with the gradual increase in the elderly population, a large number of convenient and effective health monitoring auxiliary devices have been introduced in the market. Fall detection, that is, to detect whether the person being cared for has fallen. This function can alarm the abnormal fall behavior of specific elderly people (such as the elderly with limited mobility), and ensure the safety of the elderly's actions. It is an important function of home health monitoring.

[0003] Among the devices currently available on the market that can perform fall detection, relatively typical ones are fall detection devices based on smart watches and bracelets. They integrate sensors such as accelerometers and gyroscopes, and can monitor the motion state of the wearer in real time. Once a fall event is detected, an alarm is immediately issued. However, such products can only be worn by a single person and need to be worn continuously, and the price is also relatively expensive, which is not suitable for monitoring the motion state and fall warning of people in public areas of the home. In addition, there are also fall detection devices based on cameras and artificial intelligence technology. This device can monitor the activity state of people in real time based on video stream information. Once a fall event is detected, an alarm is immediately sent to the caregiver or emergency contact. However, the camera may involve privacy issues in the home environment, and the user acceptance is not high.

[0004] The millimeter-wave radar-based fall detector has received increasing attention due to its advantages such as high detection accuracy, strong privacy protection, and insensitivity to light. Patent (CN118549926A) proposes a human fall detection method based on millimeter-wave radar point clouds. This method uses a millimeter-wave radar to collect the original radar data of a person in different action postures, preprocesses the original radar data to obtain millimeter-wave radar point cloud data, inputs the point cloud data into a human pose deep learning network model for training to obtain a fall detection model, and uses this model to determine whether the target person has fallen. However, this method is relatively complex and requires the pre-construction of a dataset of falls and situations that are easily misjudged as falls, with high requirements for data completeness. In addition, this method captures the local and global structural information of the radar point cloud data of the human target through a deep learning model, and the accuracy of fall detection is limited by the sparsity of the generated point clouds. Patent (CN115343684A) proposes a fall detection method based on target trajectories. This method first obtains the three-dimensional point cloud data of the human target by a millimeter-wave radar; then converts the three-dimensional point cloud data into the position and velocity data of each target, and uses a Kalman filter to predict the position of the target at the next moment. By matching the position of the target at the current moment with the predicted position of the target at the previous moment, the trajectory ID of each target at the current moment is obtained; finally, fall detection and judgment are performed based on the ID, position, and velocity information of the target. However, this method has high requirements for the performance of the millimeter-wave radar, and the radar needs to be able to accurately obtain the three-dimensional point cloud data and velocity information of the target. In addition, the parameter setting and adjustment of the Kalman filter may require certain experience and skills to ensure the best prediction effect. Summary of the Invention

[0005] In view of this, the present invention proposes a fall detection method, device, and medium for a home health monitoring radar, which can accurately, conveniently, and quickly image the detection scene (one-dimensional range profile), and calculate the RCS information and height approximation information of the target, and then mine features to determine whether a person has fallen.

[0006] To achieve the above object, the technical solution of the present invention is as follows: A fall detection method for a home health monitoring radar, including: performing one-dimensional range imaging on the original echo data collected by the radar to obtain the one-dimensional range image of the observation scene; suppressing clutter in the one-dimensional range image of the observation scene; calculating the amplitude of the clutter-removed one-dimensional range image and performing peak search to find the point with the largest amplitude, using the range cell where this point is located as the range cell of the target, and performing iterative filtering based on the amplitude of this point and the historical amplitude to output the amplitude of the target, obtaining the range and amplitude time series of the target; based on the obtained range and amplitude time series of the target, calculating the approximate height value of the target according to the installation height and range resolution of the radar; performing feature mining on the approximate height value and RCS amplitude of the target, respectively statistically setting feature information: designing logical judgment conditions using the statistically set feature information to determine whether a fall event occurs.

[0007] Among them, statistically setting feature information includes: 1) Statistically counting the continuous frame number AccFlag1 when the target height value is less than the set threshold HTh1 and the amplitude is greater than the set threshold RTh1; 2) Statistically counting the continuous frame number AccFlag2 when the current frame target height value is less than the set threshold HTh1 and the amplitude is greater than RTh1 and the difference between the current frame height value H and the previous frame height value is 0. If AccFlag2 is greater than N / 4, where N is the frame number threshold for fall determination, then clear AccFlag1; 3) Statistically counting the continuous frame number AccFlag3 when the difference between the current frame amplitude and the previous frame amplitude is less than the set threshold RTh2.

[0008] Among them, the specific implementation method of designing logical judgment conditions using the statistically set feature information to determine whether a fall event occurs is as follows: The first step, if AccFlag1 > N, it is considered that the height of the target is in a state of being less than the set threshold HTh1 for a long time, and there may be a suspected fall behavior; The second step, based on the first step, if AccFlag1 < N / 2, it is considered that the height value of the target is less than the set threshold HTh1 and there is a micro-motion feature, and the target is determined to be a human target; Finally, based on the first step and the second step, if AccFlag3 < N / 2, the false alarm caused by the RCS tail when the human target moves away from the scene can be excluded; If the above conditions are met simultaneously, it is determined that an abnormal fall behavior occurs, output a fall flag, and issue an alarm.

[0009] Among them, the clutter in the one-dimensional range image of the observation scene is suppressed by using the clutter map method.

[0010] Among them, the echo sampling points of the first Chirp signal in each frame are taken for fast Fourier transform to obtain the one-dimensional range image of the observed scene.

[0011] Among them, complex inter-frame clutter suppression is performed on each range cell of the one-dimensional range image of the current frame by using a first-order recursive iterative calculation method.

[0012] The present invention also provides an electronic device, which includes a processor and a memory for storing executable instructions that can be executed by the processor; the processor is configured to read the executable instructions from the memory and execute the instructions to implement the fall detection method for a home health monitoring radar according to the present invention.

[0013] The present invention also provides a computer-readable storage medium, characterized in that the storage medium stores a computer program, and the computer program is used to execute the fall detection method for a home health monitoring radar according to the present invention. Beneficial effects 1. The present invention uses the one-dimensional range image (HRRP) information to obtain the RCS amplitude time series of the target and the approximate height time series of the target, and performs feature mining on the RCS amplitude time series and the height approximate time series to judge whether the person in the core area (directly below) of the ceiling-mounted radar has fallen. The technical complexity is low, it is easy to be implemented in engineering and the cost is low. The present invention can be applied to environments such as homes and hospitals with obvious differences. Without professional measurement equipment, it can obtain the activity information of the monitored environment and detect the abnormal fall behavior of personnel, and then realize danger warning. 2. The specific implementation of the present invention is to install the radar directly above (ceiling-mounted) in an environment such as a bathroom where falls are likely to occur, and define the detection range in the core area (directly below) of the ceiling-mounted radar. The one-dimensional range image information is used to obtain the RCS and approximate height information of the target and perform feature mining on it, and then judge whether the person has fallen.

[0014] 3. The fall detection method of the present invention can be applied to home environments with obvious differences, and can detect the occurrence of fall events in a timely and accurate manner, and then give an alarm for abnormal fall behavior. 4. According to the modeling of fall behavior and the analysis of false alarm situations, the determination of abnormal fall behavior and the suppression of false alarms in the method of the present invention are mainly implemented by designing logical judgment conditions based on the RCS change characteristics and target height change characteristics. It can adapt to various requirements, can give an alarm for the abnormal fall behavior of specific elderly people, and escort the safety of the elderly's actions.

[0015] 5. The fall detection method of the present invention can be applied to home environments with obvious differences, and can detect the occurrence of fall events in a timely and accurate manner, and then give an alarm for abnormal fall behavior.

[0016] 6. According to the modeling of fall behavior and the analysis of false alarm situations, the determination of abnormal fall behavior and the suppression of false alarms in the method of the present invention are mainly implemented by designing logical judgment conditions based on the RCS change characteristics and target height change characteristics. It can adapt to various requirements, can give an alarm for the abnormal fall behavior of specific elderly people, and escort the safety of the elderly's actions.

[0017] 5. The device of the present invention is used to implement the method of the present invention. It utilizes one-dimensional range profile (HRRP) information to obtain the RCS amplitude time series of the target and the approximate height time series of the target, and performs feature mining on the RCS amplitude time series and the height approximate time series to determine whether a person in the core area (directly below) of the ceiling-mounted radar has fallen. The technical complexity is low, it is easy to implement in engineering, and the cost is low. area) has fallen.

[0018] 6. The medium of the present invention is used to implement the method of the present invention. It utilizes one-dimensional range profile (HRRP) information to obtain the RCS amplitude time series of the target and the approximate height time series of the target, and performs feature mining on the RCS amplitude time series and the height approximate time series to determine whether a person in the core area (directly below) of the ceiling-mounted radar has fallen. The technical complexity is low, it is easy to implement in engineering, and the cost is low. area) has fallen. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic flowchart of the method of the present invention.

[0020] Figure 2 It is a graph of the time-varying curves of RCS, Rbin, target height and the fall detection results when there is no person, normal activities and a fall event occur during the experimental verification of the method of the present invention.

[0021] Figure 3 It is a schematic structural diagram of the electronic device provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] The present invention will be described in detail below with reference to the accompanying drawings and by way of examples.

[0023] The present invention proposes a fall detection method for a home health monitoring radar. The radar can be installed directly above (ceiling-mounted) in an environment where falls are likely to occur, such as a bathroom. The detection range is defined in the core area (directly below) of the ceiling-mounted radar. One-dimensional range profile information is used to obtain the RCS and approximate height information of the target and perform feature mining on it, and then determine whether a person has fallen. In this embodiment, the radar is installed directly above (ceiling-mounted) in an environment where falls are likely to occur, such as a bathroom. The detection range is defined in the core area (directly below) of the ceiling-mounted radar. The center of the antenna beam is vertically irradiated at the center position directly below, and a linear frequency modulated continuous wave (Chirp) signal is transmitted in a TDM-MIMO transmission mode. The intermediate frequency echo data collected by the radar is rearranged for antenna channels, and the data of each receiving antenna is segmented to obtain the corresponding virtual antenna data, and a three-dimensional radar data cube (RadarCube) containing fast time sampling points, slow time Chirp numbers, and virtual antenna channel numbers is obtained. area), and the data of each receiving antenna is segmented to obtain the corresponding virtual antenna data, and a three-dimensional radar data cube (RadarCube) containing fast time sampling points, slow time Chirp numbers, and virtual antenna channel numbers is obtained.

[0024] The flowchart of the method of the present invention is asFigure 1 As shown in the figure. It includes the following steps: Step 1: Perform signal and data processing on the original echo data collected by the radar to obtain the high-resolution range profile (HRRP) of the observed scene; Furthermore, in this embodiment, when performing signal and data processing on the original echo data collected by the radar, the echo sampling points of the first Chirp signal in each frame are taken for fast Fourier transform (i.e., range dimension FFT) to obtain the high-resolution range profile (HRRP) of the observed scene.

[0025] Step 2: Perform clutter suppression on the high-resolution range profile. In this embodiment, the clutter map method is used for clutter suppression. The specific steps are as follows: Perform complex frame-to-frame clutter suppression on each range cell of the high-resolution range profile (HRRP) of the current frame by using the first-order recursive iterative calculation method. The specific clutter suppression method is as follows: (1) In the formula is the iterative factor of clutter estimation, is the observed value of the real part of the current range cell, is the observed value of the imaginary part of the current range cell, is the historical estimated value of the real part of the clutter of this range cell; is the historical estimated value of the imaginary part of the clutter of this range cell. is the estimated value of the real part of the clutter of this range cell; is the estimated value of the imaginary part of the clutter of this range cell; (2) In the formula, RCS is the amplitude after clutter suppression of this range cell; (3) The specific steps for calculating the target RCS and approximate height are as follows: Perform peak search on the high-resolution range profile data after clutter suppression to find the range cell of the point with the largest amplitude and the corresponding amplitude as the range cell (Rbin) where the target is located and the amplitude of the target ( ). According to the following formula (4), use the historical value to filter the amplitude of the target and output the RCS: (4) In the formula is the filtering factor. Assume that the range cell Rbin where the target is located is , and the approximate height information of the target can be obtained according to the installation height of the radar and the range resolution Solve according to Equation (4) (the range resolution is determined by the bandwidth of the radar transmitted signal).

[0026] (5) Step 3: Calculate the amplitude of the clutter-removed one-dimensional range profile (HRRP) and perform peak search to find the point with the maximum amplitude. Use the range cell where this point is located as the range cell (Rbin) of the target. And perform iterative filtering based on the amplitude of this point and the historical amplitude, output the amplitude (RCS) of the target, and obtain the time series of the range (Rbin) and amplitude (RCS) of the target.

[0027] Solve the approximate height information of the target according to Rbin, installation height, and range resolution; Step 4: Based on the obtained time series of the range (Rbin) and amplitude (RCS) of the target, calculate the approximate height value of the target according to the installation height and range resolution of the radar.

[0028] Step 5: Mine the features of the target's height and RCS, and respectively count the following features: 1) Count the continuous frame number AccFlag1 when the target height value is less than the set threshold HTh1 (typical value is 0.3) and the RCS is greater than the set threshold RTh1 (typical value is 150). If the target height H is less than the set threshold HTh1 and the target RCS is greater than the set threshold RTh1, then increment AccFlag1 by 1; otherwise, clear AccFlag1 to zero. This feature is the main feature for fall determination.

[0029] 2) Count the continuous frame number AccFlag2 when the current frame target height value is less than the set threshold HTh1 and the RCS is greater than the set threshold RTh1 and the difference between the current frame height value H and the previous frame height value is 0. If AccFlag2 is greater than N / 4 (N is the frame number threshold for fall determination, typical value is 80 frames), then clear AccFlag1 to zero; if AccFlag2 is greater than N / 4 (N is the frame number threshold for fall determination, typical value is 80 frames), it means that the height value of the current target remains unchanged for a long time and does not have the unique micro-motion characteristics of a stationary human target. Consider the target as a non-human target and clear AccFlag1 to zero.

[0030] 3) Count the continuous frame number AccFlag3 when the difference between the amplitude RCS of the current frame and the amplitude of the previous frame is less than the set threshold RTh2 (typical value is 50); specifically, when the human target moves away from the scene, due to the filtering tailing effect of RCS, the RCS decreases slowly. During the decreasing process, the height value H of the target may remain in a state less than the set threshold HTh1 for a long time, resulting in false alarms. Therefore, the continuous frame number AccFlag3 when the difference between the current frame RCS and the previous frame is less than RTh2 (typical value is 50) can be counted to determine whether the current human target is moving away from the scene.

[0031] Step Six: Design logical judgment conditions using the statistically obtained feature information to determine whether a fall event has occurred.

[0032] Based on the modeling of fall behaviors and the analysis of false alarm situations, in the embodiments of the present invention, the determination of abnormal fall behaviors and the suppression of false alarms are mainly implemented by designing logical judgment conditions based on the RCS change characteristics and target height change characteristics. The specific implementation methods are as follows: (1) If AccFlag1 > N, it is considered that the height of the target has been in a state less than the set threshold HTh1 (typical value is 0.3) for a long time, and there may be a suspected fall behavior.

[0033] (2) Further, on the basis of (1), if AccFlag1 < N / 2, it is considered that the height value of the target is less than the set threshold HTh1 and there is a micro-motion feature, and the target is determined to be a human target.

[0034] (3) Finally, on the basis of (1) and (2), if AccFlag3 < N / 2, the false alarm caused by the RCS tailing when the human target moves away from the scene can be excluded.

[0035] (4) If the above conditions are satisfied simultaneously, it is determined that an abnormal fall behavior has occurred, the fall flag is output, and an alarm is issued. The typical value of N is 80 frames.

[0036] The embodiments of the present application also provide an electronic device. Figure 3The structure of the electronic device provided by the embodiments of the present invention is shown. For example, the electronic device 30 may include a processor 31, a memory 32, and a transmission device 33. The processor is used to execute the fall detection method for the home health monitoring radar mentioned in the above embodiments. The processor and the memory may be connected through a bus or other means. Taking the connection through the bus as an example, the transmission device can be connected to the processor and the memory in a wired or wireless manner. The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the fall detection method for the home health monitoring radar in the embodiments of the present application. By running the non-transitory software programs, instructions, and modules stored in the memory, the processor can execute various functional applications and data processing of the processor, that is, implement the fall detection method for the home health monitoring radar in the above method embodiments. The memory may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely set relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above network include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof. The one or more modules are stored in the memory and, when executed by the processor, execute the fall detection method for the home health monitoring radar in the embodiments.

[0037] As another aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium may be the computer-readable storage medium included in the device in the above embodiments; it may also exist alone and be a computer-readable storage medium not assembled into the device. The computer-readable storage medium may be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium well-known in the technical field. The computer-readable storage medium stores one or more programs, and the one or more programs are used by one or more processors to execute the fall detection method for the home health monitoring radar described in the present application.

[0038] In summary, the above are only the preferred embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A fall detection method for a home health monitoring radar, characterized in that, Including: performing one-dimensional range imaging on the original echo data collected by the radar to obtain the one-dimensional range image of the observed scene; Performing clutter suppression on the one-dimensional range image of the observed scene; calculating the amplitude of the clutter-removed one-dimensional range image and performing peak search to find the point with the maximum amplitude, using the range cell where this point is located as the range cell of the target, and performing iterative filtering with the amplitude of this point combined with the historical amplitude to output the amplitude of the target, obtaining the range and amplitude time series of the target; based on the obtained range and amplitude time series of the target, calculating the approximate height value of the target according to the installation height and range resolution of the radar; Performing feature mining on the approximate height value and RCS amplitude of the target, respectively statistically calculating the set feature information: designing logical judgment conditions using the statistically calculated set feature information to determine whether a fall event occurs.

2. The method according to claim 1, wherein The statistically calculated set feature information includes: 1) Statistically calculating the continuous frame number AccFlag1 when the target height value is less than the set threshold HTh1 and the amplitude is greater than the set threshold RTh1; 2) Statistically calculating the continuous frame number AccFlag2 when the target height value in the current frame is less than the set threshold HTh1 and the amplitude is greater than RTh1 and the difference between the height value H in the current frame and the height value in the previous frame is 0. If AccFlag2 is greater than N / 4, where N is the frame number threshold for fall determination, then clear AccFlag1; 3) Statistically calculating the continuous frame number AccFlag3 when the difference between the amplitude in the current frame and the amplitude in the previous frame is less than the set threshold RTh2.

3. The method according to claim 2, wherein , using the statistically calculated set feature information to design logical judgment conditions to determine whether a fall event occurs. The specific implementation method is as described below: In the first step, if AccFlag1 > N, it is considered that the height of the target has been in a state of being less than the set threshold HTh1 for a long time, and a suspected fall behavior may occur; In the second step, based on the first step, if AccFlag1 < N / 2, it is considered that the height value of the target is less than the set threshold HTh1 and there is a micro-motion feature, and the target is determined to be a human target; Finally, based on the first and second steps, if AccFlag3 < N / 2, the false alarm caused by the RCS tail when the human target moves away from the scene can be excluded; If the above conditions are satisfied simultaneously, it is determined that an abnormal fall behavior occurs, an alarm flag is output, and an alarm is issued.

4. The method according to any one of claims 1 to 3, characterized in that , using the clutter map method to perform clutter suppression on the one-dimensional range image of the observed scene.

5. The method according to any one of claims 1 to 3, characterized in that, Taking the echo sampling points of the first Chirp signal in each frame for fast Fourier transform to obtain the one-dimensional range image of the observed scene.

6. The method according to claim 4, characterized in that, Performing complex inter-frame clutter suppression on each range cell of the one-dimensional range image of the current frame by using the first-order recursive iteration calculation method.

7. An electronic device, characterized in that, The electronic device includes a processor and a memory for storing executable instructions that can be executed by the processor; the processor is used to read the executable instructions from the memory and execute the instructions to implement the fall detection method for a home health monitoring radar as described in any one of claims 1-6 above.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, and the computer program is used to execute the fall detection method for a home health monitoring radar as described in any one of claims 1-6 above.

Images