A fall detection method for home health monitoring radar

Abstract

This invention proposes a fall detection method for home health monitoring radar, capable of determining falls in various scenarios and under different installation conditions. The method includes: installing the radar in a location within the scene that covers the entire area and sending point cloud data and target trajectory data to a data center in real time; generating activity levels corresponding to the target's movement state at different times based on the point cloud data of day n; dividing the time period into active and inactive periods based on the target trajectory and activity level of day n, and projecting the two periods separately to obtain an activity area projection map corresponding to the daily activity area and a rest area projection map corresponding to the rest area; obtaining a height distribution map of target activity based on the trajectory of day n; obtaining the activity channels for the day based on the activity area projection map and the rest area projection map of day n; accumulating the activity channels over n days, setting an activity unit threshold, and obtaining an activity distribution map after the determination is completed, etc.

Description

Technical Field

[0001] This invention belongs to the field of radar signal processing technology, specifically relating to a fall detection method for home health monitoring radar. Background Technology

[0002] With the improvement of my country's economic and social development level, population aging has become the norm in Chinese society, and the proportion of elderly people in my country is constantly rising. How to ensure the health and safety of the elderly has become a key focus of social concern. Considering the new characteristics of my country's society, such as low birth rate and high mobility, traditional multi-generational families are gradually decreasing, and the phenomenon of empty-nest families among the elderly is becoming increasingly prominent. Therefore, the home-based care model for the elderly living independently will become the mainstream in my country.

[0003] Falls are the most common dangerous behavior in the daily lives of the elderly. Studies show that about 30% of people aged 65 and over will fall accidentally each year. If not detected in time, falls can lead to injuries such as fractures, head injuries, sprains and strains, and in severe cases, even death. This poses a serious threat to the physical and mental health of the elderly.

[0004] Therefore, in recent years, various sectors have introduced many fall detection methods. One mainstream approach is to determine a fall by detecting changes in the height and speed of the target's movement. However, the target height and speed measured by radar are closely related to the installation angle, height, and scene of the equipment. Under the same fall determination conditions, the performance of the fall detection algorithm may significantly decrease after changes in installation conditions and scene. Summary of the Invention

[0005] This invention proposes a fall detection method for home health monitoring radar, which can determine the fall of the target under different scenarios and installation conditions.

[0006] The present invention is achieved through the following technical solution.

[0007] A fall detection method for home health monitoring radar includes:

[0008] Step 1: Install the radar in a location within the scene that can cover the entire area, and send point cloud data and target trajectory data to the data center in real time;

[0009] Step 2: Based on the point cloud data of day n, generate the activity level corresponding to the target's motion state at different time periods;

[0010] Step 3: Based on the target flight track and activity level on day n, divide the time period into active and leisure periods according to the activity level, and project the two time periods separately to obtain the activity area projection map corresponding to the daily activity area and the rest area projection map corresponding to the rest area; and obtain the altitude distribution map of the target activity based on the flight track on day n.

[0011] Step 4: Based on the activity area projection map and rest area projection map for day n, obtain the activity passage for that day;

[0012] Step 5: Accumulate the activity channels over n days, set the activity unit threshold, and obtain the activity distribution map after completing the judgment;

[0013] Step 6: Perform weighted processing on the altitude distribution map of n days to obtain the latest altitude distribution map;

[0014] Step 7: Extract three features of the specified track in the spatial and temporal dimensions based on the activity distribution map and altitude distribution map, namely, track extension, proportion of activity area along the track, and proportion of low altitude value of the track.

[0015] Step 8: If the track meets all three characteristics, change the current target state to "fall". Then check if the track number has been recorded. If not, issue a fall alarm and record the current track number.

[0016] Beneficial effects of this invention:

[0017] 1. This invention uses radar to accumulate point cloud and trajectory information of targets in a scene by deploying it routinely, and generates activity channel map and height distribution map of the scene respectively, and uses them to assist in the determination of target fall.

[0018] 2. This invention achieves specific scene perception by establishing activity distribution maps and height distribution maps for different scenarios, effectively solving the problem of the impact of height threshold on the performance of fall detection algorithms, thus enabling reliable fall determination in various scenarios and robustly detecting fall events;

[0019] 3. This invention is adaptive. When the installation angle, height, etc. of the equipment change, this method can adapt to the changes in installation conditions by updating the activity distribution map and height distribution map, so as to ensure the performance of the fall detection algorithm.

[0020] 4. The fall detection method provided by this invention can be applied to different home environments, conveniently detect the elderly’s condition in a non-contact manner, and provide alarms for fall behavior. Attached Figure Description

[0021] Figure 1 Flowchart of a fall detection method for home health monitoring radar;

[0022] Figure 2 The activity level on day n;

[0023] Figure 3 Projected images of the activity area and rest area;

[0024] Figure 4 A distribution map of activities accumulated over multiple days;

[0025] Figure 5 The elevation distribution map for day n and the elevation map after one week of iteration;

[0026] Figure 6 This is a map showing the target trajectory and activity distribution during the actual fall. Detailed Implementation

[0027] Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described in the drawings are merely exemplary and are intended to illustrate the principles and spirit of the present invention, and are not intended to limit the scope of the present invention.

[0028] like Figure 1 As shown, a fall detection method for home health monitoring radar according to the present invention specifically includes the following steps:

[0029] Step 1: Install the radar in a location within the scene that can cover the entire area, and send the point cloud (PointSet) and target track to the data center in real time;

[0030] In this embodiment, the point cloud PointSet is a set of points generated by the radar through processing the echo, containing multi-dimensional information such as target angle, distance, and signal-to-noise ratio. Specifically:

[0031] Poigntset = [range i ,azimuth i ,elevation i doppler i ,snr i ], i = 1, 2,, N (1)

[0032] Among them, range i ,azimuth i ,elevation i Doppler represents the radial distance, azimuth, and elevation angle of the i-th point, respectively. i Let snr be the velocity at the i-th point. i Let N be the signal-to-noise ratio of the i-th point, and N be the number of points in the point cloud.

[0033] In this embodiment, the target track is the set of coordinates of the target's motion trajectory, specifically:

[0034] track = [x k ,y k ,z k ,t k ]T k = 1, 2, ... (2)

[0035] Where, x k y k , z k Let x, y, and z represent the x, y, and z coordinates of the k-th trajectory point in the radar coordinate system, respectively. k Let L represent the time of the k-th trajectory, and L be the number of trajectory points.

[0036] Step 2: Based on the point cloud data of day n, generate the ActiveLevel corresponding to the target's motion state at different time periods; specifically:

[0037] ActiveLevel(k) = CntNum k / maxCntNum,k=1,2,,N (3)

[0038] Where ActiveLevel(k) is the activity level in the k-th time period, and CntNum k is the number of point clouds in the k-th time period, maxCntNum is the maximum number of point clouds in all time periods of the day, and N is the number of time periods. In this embodiment, the value is 48, that is, the whole day is divided into 48 time periods with half-hour intervals.

[0039] Step 3: Based on the target flight track and activity level on day n, divide the time period into active and leisure periods according to the activity level, and project the two periods separately to obtain an ActiveMap corresponding to the daily activity area and a RestMap corresponding to the rest area; and obtain an altitude distribution map HMap of the target activity based on the flight track on day n; specifically:

[0040] First, the radar's field of view is divided into a grid, and the spatial coordinate axes are defined as follows:

[0041] x = x i i = 1, 2, ..., N

[0042] y = y j ,j=1,2,,M (4)

[0043] The x-axis direction is divided into N spatial units, covering an area of ​​x. min ~x max The grid spacing is x bin The y-axis is divided into M spatial units, covering an area of ​​y. min ~y max The grid spacing is y bin .

[0044] like Figure 2As shown, based on the indoor scene, the target is in a resting state with low activity between 12:00 AM and 6:00 AM, and has higher activity at other times. These two time periods correspond to the target's sleep state and daily activity state, respectively. By extracting the trajectory for the corresponding time period and projecting it, the corresponding area projection map can be obtained. The target trajectory contains the coordinate set of the target's movement path. The specific projection method of the activity area is shown in the following formula:

[0045] track = [x k ,y k ,z k ,t k ] T k = 1, 2, ...

[0046] t k ∈T active

[0047]

[0048]

[0049] ActiveMap(i,j)=ActiveMap(i,j)+1 (5)

[0050] Where track represents the selected trajectory during the high-activity period, i represents the index in the x-axis direction of the grid, j represents the index in the y-axis direction of the grid, and ActiveMap(i,j) represents the number of trajectory points in the grid with index (i,j).

[0051] The track contains the target's spatial location information at every moment. Therefore, for an active track, its time is within the active period. By traversing the position of each frame of the track and calculating its corresponding projected position, the position indication information is incremented by 1. After completion, the active area projection map (ActiveMap) can be obtained.

[0052] Similarly, when t k ∈T rest At that time, the Rest Map projection can be obtained.

[0053] In this embodiment, the height distribution map HMap of the target activity is obtained using the following method:

[0054] track = [x k ,y k ,z k ,t k ] T k = 1, 2, ...

[0055]

[0056]

[0057] HMap(i,j)=α·HMap(i,j)+(1-α)·z k (6)

[0058] Where 'a' is the iteration coefficient used to smooth the height value of the cell, and HMap(i,j) is the height value of the cell with grid index (i,j).

[0059] Step 4: Based on the ActiveMap and RestMap of the activity area on day n, obtain the activity channel A for that day. n Specifically:

[0060] Based on the activity area and rest area projection diagrams for day n, increment the number of cells enabled by the activity projection diagram by 1 and decrement the number of cells enabled by the rest area projection diagram by 1 to obtain the activity channel A for that day. n :

[0061]

[0062] Among them, A n (i,j) represents the value of the activity channel with grid index (i,j) on day n. A value of -1 indicates a rest area, and a value of 1 indicates an activity area.

[0063] Step 5: Accumulate the activity channels over n days, set the activity unit threshold γ, and obtain the activity distribution map B after completing the judgment. n Specifically:

[0064] By accumulating the activity channels over n days, we obtain the activity distribution map B. n For use in subsequent fall detection:

[0065]

[0066] Among them, B n (i,j) represents the activity distribution value of the grid index (i,j) after n days of accumulation. If a cell is identified as an active area γn times in n days, the cell is set as an active cell and assigned a value of 1; otherwise, it is assigned a value of 0. Here, γ is a scaling factor.

[0067] Step 6: Perform weighted processing on the n-day altitude distribution map to obtain the latest altitude distribution map H. n Specifically:

[0068] H n =βH n-1 +(1-β)·HMap n (9)

[0069] Among them, H n HMap is used to generate a cumulative height distribution map over n days. n This is the altitude distribution map for day n, where β is the iteration coefficient used to smooth the altitude values ​​over multiple days.

[0070] Step 7: Extract three features of the specified track in the spatial and temporal dimensions based on the activity distribution map and altitude distribution map: track extension K0, proportion of activity area along the track K1, and proportion of low track altitude K2.

[0071] In this embodiment, the designated trajectory is determined when, in a single-person scenario, the trajectory interruption time exceeds T0, or the trajectory stationary time exceeds T1, indicating a possibility of a fall.

[0072] In this embodiment, the track extension K0 is determined in the following way:

[0073] For track = [x i ,y i ,z i ] T i = 1, 2, ..., N, x i ,y i ,z i Let x, y, and z be the x, y, and z coordinates of the i-th trajectory point, respectively. Its extension K0 describes the spatial span of the trajectory. This constraint eliminates false alarms where trajectories are clustered within a very small area. The formula is as follows:

[0074] W = max(x) i )-min(x i )

[0075] L = max(y i )-min(y i )

[0076] K0 = max(W, L)(10)

[0077] Where W is the width extension of the track and L is the length extension of the track, the maximum value of the two is taken as the spatial extension K0.

[0078] In this embodiment, the proportion of the activity zone along the track, K1, describes the percentage of the track located within the activity zone. Tracks with a high proportion outside the activity channel can be excluded, thus focusing only on fall behavior within this space during daily activities. It is determined using the following method:

[0079]

[0080]

[0081] Among them, T iThe value of the activity enable for the i-th trajectory point is 1 if the point is located within the activity channel, and 0 otherwise.

[0082] In this embodiment, the low altitude percentage K2 is used to describe the percentage of frames in the test flight set that are lower at the current altitude compared to the altitude at which daily activities take place. Specifically, it is defined as:

[0083]

[0084]

[0085] Among them, h k This indicates that the low altitude value of the k-th waypoint is enabled. If the difference between the altitude distribution value of this waypoint and the altitude distribution value of this cell is less than the threshold thr, the value is 1; otherwise, the value is 0.

[0086] Step 8: If the track simultaneously meets all three characteristics, change the current target state to "fall," then check if the track number has been recorded. If not, issue a fall alarm and record the current track number. The specific formula is as follows:

[0087]

[0088] Among them, ThrExtend is the spatial extension threshold, ThrAcArea is the activity area proportion threshold, and ThrLowHgt is the low track altitude proportion threshold.

[0089] The following is an experimental test example using equipment installed in a real bedroom setting to demonstrate the effectiveness of the invention; specifically:

[0090] The device is routinely deployed within the scenario, and data is accumulated. Based on the point cloud accumulated on day n, the activity level for that day is obtained, such as... Figure 2 As shown. Based on the flight path and activity level on day n, the projected maps of the activity area and rest area for that day are obtained, as follows. Figure 3 As shown. After several days of data accumulation, an activity distribution map that can be used for fall detection was obtained, as shown. Figure 4 As shown. Based on the flight path on day n, the altitude distribution map for that day can be obtained, as shown. Figure 5 As shown in the left figure, after several days of iteration (7 days in total), a height distribution map suitable for fall detection is obtained. Figure 5 As shown in the right figure. After several days of data accumulation and iteration, an actual fall test was conducted in this scenario. The device remained stationary on the ground for more than T1, and then issued a fall alarm normally. The fall trajectory and activity distribution map are shown below. Figure 6 As shown, the asterisk indicates the location where the target fell.

[0091] This experiment verified the effectiveness of using the method proposed in this invention to generate activity distribution maps and height distribution maps through data accumulation, thereby assisting in fall detection technology for targets. The experimental results show that the fall detection technology provided by this invention can be applied to different home environments, conveniently detecting the elderly's condition in a non-contact manner and providing alerts for fall behavior.

[0092] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A fall detection method for home health monitoring radar, characterized in that, include: Install the radar in a location within the scene that can cover the entire area, and send point clouds and target tracks to the data center in real time; The activity level of the target's motion state at different time periods is generated based on the point cloud data of day n. Based on the target flight track and activity level on day n, the time period is divided into active and leisure periods according to the activity level. The two time periods are projected separately to obtain the activity area projection map corresponding to the daily activity area and the rest area projection map corresponding to the rest area. Based on the flight track on day n, the altitude distribution map of the target activity is obtained. Based on the activity area projection map and rest area projection map for day n, the activity passage for that day can be obtained; Accumulate the activity channels over n days, set the activity unit threshold, and obtain the activity distribution map after completing the judgment; The latest height distribution map is obtained by weighting the height distribution maps of n days. Based on the activity distribution map and altitude distribution map, three features of the space and time dimensions of the specified flight path are extracted: flight path extension, proportion of the activity area along the path, and proportion of low altitude values ​​of the flight path. If the trajectory simultaneously meets the above three characteristics, the current target state is changed to fall. Then check whether the trajectory number has been recorded. If it has not been recorded, issue a fall alarm and record the current trajectory number. Based on the activity area projection map on day n and rest area projection map Get access to the day's activities Specifically: Based on the activity area and rest area projection diagrams for day n, increment the number of cells with enabled activity projection diagrams by 1 and decrement the number of cells with enabled rest area projection diagrams by 1 to obtain the activity channels for that day. : in, The grid index for day n is The activity channel of the unit has a value of -1, which indicates a rest area and a value of 1, which indicates an activity area. track extension The following method shall be used to determine: track , The first trajectory points 3D coordinates, its extensibility This constraint is used to describe the size of the space span of a flight path, and it eliminates false alarms caused by flight paths clustering in a very small area. Percentage of activity areas along tracks The proportion of flight tracks located within the activity area is determined in the following manner: (11) in, For the first The activity enable value for each trajectory point is set to 1 if the point is located within the activity channel, and 0 otherwise. Activity distribution map; Low altitude percentage of flight paths This describes the percentage of frames in the test flight set that are relatively low compared to those at the current altitude, and are commonly used for daily activities at this altitude. Specifically, it is defined as: (12) in, Indicates the first Enable low altitude values ​​for each waypoint if the difference between that waypoint and the altitude distribution value of this cell is less than a threshold. The value is 1 if the value is 1, otherwise the value is 0. This is a cumulative altitude distribution map over n days. Indicates the first The z-dimensional coordinates of each trajectory point in the radar coordinate system; (3) in, It represents the activity level in time period k. It is the number of point clouds in the k-th time period. It is the maximum number of point clouds in all time periods of the day, where N is the number of points.

2. The fall detection method for home health monitoring radar as described in claim 1, characterized in that, Point cloud The radar processes the echoes to generate a set of points containing multi-dimensional information about the target's angle, range, and signal-to-noise ratio. Specifically: in, Let be the radial distance, azimuth angle, and elevation angle of the i-th point, respectively. Let be the velocity of the i-th point. Let be the signal-to-noise ratio at the i-th point. The number of points in the point cloud.

3. A fall detection method for home health monitoring radar as described in claim 1 or 2, characterized in that, Target track The set of coordinates for the target's trajectory is as follows: in, , , They represent the first The x, y, z coordinates of a trajectory point in the radar coordinate system. Indicates the first The moment of a trajectory The number of trajectory points.

Images