Fall detection method and apparatus

1. A fall detection apparatus, comprising:

a memory;

a processor coupled to the memory and the processor configured to:

acquire information of radar detection points belonging to a detected object in a predetermined number of consecutive frames, information of a radar detection point among the radar detection points including at least two of information on a position, information on a radial velocity and information on reflection signal strength; and

determine whether the detected object has fallen according to a relationship between time and the information of the radar detection points in the predetermined number of consecutive frames;

wherein the predetermined number of consecutive frames are a first predetermined number of consecutive frames and the processor is further configured to:

calculate average detection point information of a plurality of radar detection points belonging to the detected object in each frame, wherein the average detection point information is an average value of each of the information of the plurality of radar detection points;

calculate a first feature value and a second feature value of the average detection point information in the first predetermined number of consecutive frames,

where the first feature value denotes:

a parameter of the average detection point information related to time variation, and the second feature value denoting a statistical value of average detection point information of a second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames, and/or

a difference between a statistical value of average detection point information of a third predetermined number of consecutive frames located in front of the first predetermined number of consecutive frames and a statistical value of average detection point information of the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames; and

determine whether the detected object has fallen according to the first feature value and the second feature value.

2. The fall detection apparatus according to claim 1 , wherein the information on the position of the radar detection point includes at least one of information on a distance between the radar detection point and a radar and information on a three-dimensional coordinate of the radar detection point.

3. The fall detection apparatus according to claim 1 , wherein when the information of the radar detection point is information of a height in information on a three-dimensional coordinate of the radar detection point, or the information on the reflection signal strength, or the information on the radial velocity, the first feature value is a rate of variation of linear fitting between average detection point information and time, and/or a coefficient of correlation between the average detection point information and time, and the second feature value is an average value of the average detection point information of the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames.

4. The fall detection apparatus according to claim 1 , wherein when the information of the radar detection point is respective information of a horizontal coordinate and a vertical coordinate of a horizontal plane in information on a three-dimensional coordinate of the radar detection point,

the first feature value is a linear fitting error between average detection point information of the horizontal coordinate and average detection point information of the vertical coordinate, and/or a coefficient of correlation between average detection point information of the horizontal coordinate and average detection point information of the vertical coordinate, and

the second feature value is a distance between a coordinate trajectory center of a horizontal plane of the third predetermined number of consecutive frames located in front of the first predetermined number of consecutive frames and a coordinate trajectory center of a horizontal plane of the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames.

5. The fall detection apparatus according to claim 1 , wherein the processor is further configured to:

count a number of radar detection points belonging to the detected object in each frame, and

calculate a third feature value and/or a fourth feature value of the number of radar detection points,

where the third feature value denotes a parameter of the number of detection points related to time variation, and the fourth feature value denotes an average value of the number of detection points of the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames; and

determine whether the detected object has fallen according to the first feature value, the second feature value, the third feature value and/or the fourth feature value.

6. The fall detection apparatus according to claim 5 , wherein the third feature value is a rate of variation of linear fitting between the number of radar detection points of each frame and time, and/or a coefficient of correlation between the number of radar detection points of each frame and time.

7. The fall detection apparatus according to claim 1 , wherein the processor compares the first feature value and the second feature value, or the first feature value, the second feature value, the third feature value and/or the fourth feature value, with respective falling judgment thresholds, and determines whether the detected object has fallen according to a result of comparison.

8. The fall detection apparatus according to claim 7 , wherein the processor is further configured to:

acquire information on detection points of falling training, and obtain a range of the falling judgment thresholds by using a machine learning algorithm.

9. A fall detection apparatus, comprising:

a memory;

a processor coupled to the memory and the processor configured to:

acquire information on radar detection points belonging to a detected object in a predetermined number of consecutive frames, information of a radar detection point among the radar detection points including at least two of information on a position, information on a radial velocity and information on reflection signal strength;

calculate feature values according to the information on radar detection points in the predetermined number of consecutive frames, the feature values including at least two of a feature related to a number of detection points of a radar reflection plane, or a feature related to reflection signal strength of a radar reflection plane, a feature related to a height of a radar detection point, a feature related to a velocity of a radar detection point, and a feature related to a detection point trajectory of a radar reflection plane; and

determine whether the detected object has fallen according to the feature values;

wherein the predetermined number of consecutive frames are a first predetermined number of consecutive frames, and

the feature related to the number of detection points of a radar reflection plane includes a rate of variation (slope) of linear fitting between the number of radar detection points of each frame and time, and a coefficient of correlation between the number of radar detection points of each frame and time, and an average value of the numbers of radar detection points in a second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames,

the feature related to the reflection signal strength of the radar reflection plane includes a rate of variation (slope) of linear fitting between an average reflection signal strength of radar detection points of each frame and time, and a coefficient of correlation between an average reflection signal strength of radar detection points of each frame and time, and an average reflection signal strength of radar detection points in the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames,

the feature related to a height of a radar detection point includes a rate of variation (slope) of linear fitting between an average height of radar detection points of each frame and time, and a coefficient of correlation between an average height of radar detection points of each frame and time, and an average height of radar detection points in the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames,

the feature related to a velocity of a radar detection point includes a rate of variation (slope) of linear fitting between an average Doppler velocity of radar detection points of each frame and time, and a coefficient of correlation between an average Doppler velocity of radar detection points of each frame and time, and an average Doppler velocity of radar detection points in the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames,

the feature related to a detection point trajectory of a radar reflection plane includes a linear fitting error between average detection point information of a horizontal coordinate and average detection point information of a vertical coordinate of each frame, and a coefficient of correlation between average detection point information of the horizontal coordinate and average detection point information of the vertical coordinate of each frame, and a distance between a coordinate trajectory center of a horizontal plane of a third predetermined number of consecutive frames located in front of the first predetermined number of consecutive frames and a coordinate trajectory center of a horizontal plane of a second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames.

10. A fall detection method, comprising:

acquiring information of radar detection points belonging to a detected object in a predetermined number of consecutive frames, information on a radar detection point among the radar detection points including at least two of information on a position, information on a radial velocity and information on reflection signal strength; and

determining whether the detected object has fallen according to a relationship between time and the information of radar detection points in the predetermined number of consecutive frames,

wherein the predetermined number of consecutive frames are a first predetermined number of consecutive frames and the determining of whether the detected object has fallen according to a relationship between time and the information of radar detection points in the predetermined number of consecutive frames includes:

calculating average detection point information of a plurality of radar detection points belonging to the detected object in each frame, wherein the average detection point information is an average value of each the information of the plurality of radar detection points;

calculating a first feature value and a second feature value of the average detection point information in the first predetermined number of consecutive frames,

where the first feature value denotes:

a parameter of the average detection point information related to time variation, and the second feature value denoting a statistical value of average detection point information of a second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames, and/or

a difference between a statistical value of average detection point information of a third predetermined number of consecutive frames located in front of the first predetermined number of consecutive frames and a statistical value of average detection point information of the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames; and

determining whether the detected object has fallen according to the first feature value and the second feature value.

11. The fall detection method according to claim 10 , wherein when the information of the radar detection point is information of a height in information on a three-dimensional coordinate of the radar detection point, or the information on reflection signal strength, or the information on the radial velocity, the first feature value is a rate of variation of linear fitting between the average reflection information and time, and/or a coefficient of correlation between the average reflection information and time, and the second feature value is an average value of the average detection point information of the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames.

12. The fall detection method according to claim 10 , wherein when the information of the radar detection point is information on a horizontal coordinate and a vertical coordinate of a horizontal plane in information on a three-dimensional coordinate of the detection point,

the first feature value is a linear fitting error between average detection point information of the horizontal coordinate and average detection point information of the vertical coordinate, and/or a coefficient of correlation between average detection point information of the horizontal coordinate and average detection point information of the vertical coordinate, and

the second feature value is a distance between a coordinate trajectory center of a horizontal plane of the third predetermined number of consecutive frames located in front of the first predetermined number of consecutive frames and a coordinate trajectory center of a horizontal plane of the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames.

13. The fall detection method according to claim 10 , wherein the method further includes:

counting a number of radar detection points belonging to the detected object in each frame, and calculate a third feature value and/or a fourth feature value of the number of radar detection points,

where the third feature value denotes a parameter of the number of radar detection points related to time variation, and the fourth feature value denotes an average value of the number of radar detection points of the second predetermined number of consecutive frames located behind the first predetermined number of consecutive frames; and

whether the detected object falls is determined according to the first feature value, the second feature value, the third feature value and/or the fourth feature value.