CN-121987186-A - Non-contact type bed-leaving detection method based on millimeter wave radar

CN121987186ACN 121987186 ACN121987186 ACN 121987186ACN-121987186-A

Abstract

The invention belongs to the technical field of millimeter wave radar non-contact vital sign monitoring, and particularly relates to a non-contact type off-bed detection method based on a millimeter wave radar. The method comprises the steps of screening radar data windows potentially containing human targets through a self-adaptive energy-entropy dual gating mechanism, detecting OS-CFAR (operation, configuration and condition) by adopting ordered statistics constant false alarm rate aiming at the screened windows, detecting effective human targets and bed targets, further screening by combining morphological filtering and a self-adaptive area threshold method, finally, generating radar point cloud for the screened targets, extracting radar point cloud only containing the human targets, and accurately distinguishing real human targets from dynamic interference through joint evaluation of point cloud clustering and space-time consistency.

Inventors

ZENG XIAOLU
WANG ZIXIN
YANG XIAOPENG
HAO HUIMIN
XING CHENGJIAN
QU XIAODONG

Assignees

北京理工大学

Dates

Publication Date: 20260508
Application Date: 20260113

Claims (10)

1. A non-contact bed-leaving detection method based on millimeter wave radar is characterized by comprising the following specific processes: Firstly, screening radar data windows potentially containing human targets through a self-adaptive energy-entropy double gating mechanism; Secondly, detecting an OS-CFAR by adopting ordered statistics of constant false alarm rate aiming at the screened window, detecting effective human targets and bed targets, and further screening by combining morphological filtering and a self-adaptive area threshold method; Finally, radar point cloud generation is carried out on the screened targets, the radar point cloud only comprising human targets is extracted, and the real human targets and dynamic interference are accurately distinguished through joint evaluation of point cloud clustering and space-time consistency.
2. The non-contact bed-leaving detection method based on millimeter wave radar according to claim 1, wherein the step of screening out radar data windows potentially containing human targets by a self-adaptive energy-entropy dual gating mechanism comprises the following steps: processing millimeter wave radar echo data in each sliding window, and calculating total energy of a distance-Doppler diagram in the region of interest; Calculating the total energy mean value and standard deviation of noise of the distance-Doppler graph, setting a strong signal energy threshold according to the total energy mean value and standard deviation, and screening out a high energy window which is obviously higher than the background by utilizing the strong signal energy threshold; for echo windows with weaker energy, renyi entropy is introduced to quantify spectral ordering Setting Reiyi entropy threshold values to screen out low entropy windows; the high-energy window and the low-entropy window are radar data windows potentially containing human targets.
3. The method for non-contact bed exit detection based on millimeter wave radar according to claim 2, wherein the calculating total energy of the range-doppler plot in the region of interest comprises the following steps: the first 60 seconds of setting an empty scene to acquire radar echo data, acquiring millimeter wave radar bed body part echo original data, performing distance dimension FFT and Doppler dimension FFT on each sliding window, and calculating total energy of a distance-Doppler diagram in a region of interest (ROI); The strong signal energy threshold , For the first 60s of each window background noise mean, To set a constant.
4. The millimeter wave radar-based non-contact bed exit detection method of claim 2, wherein the Renyi entropy quantifies spectral ordering: Wherein, the Representing the energy duty cycle of each range-doppler cell, Representing the order of the rayleigh entropy, Representing the total number of range-doppler cells; The Reiyi entropy threshold is set to Wherein The average Reiyi entropy for each window for the first 60 s.
5. The method for non-contact bed leaving detection based on millimeter wave radar according to claim 1, wherein the output radar point cloud is set as a speed threshold value And performing micro Doppler screening to obtain the final radar point cloud only containing the human body target.
6. The non-contact bed exit detection method based on millimeter wave radar according to claim 1, wherein the space-time consistency joint evaluation is specifically: First, a three-dimensional geometric centroid vector is defined And introducing a cluster quality factor To quantify the compactness of the structure: Wherein, the Standard deviation vectors representing cluster points along Cartesian coordinate axes; Second, the latest buffer is cached by adopting a first-in first-out buffer Sequence of individual window centroid trajectories Corresponding cluster quality sequences ; Finally, setting double constraints of space mass center stability and structural compactness, which are required to be met simultaneously in verification of human body targets: The motion space constraint is that the maximum Euclidean norm between the mass centers of each window is smaller than the physiological motion threshold value ; Secondly, morphological consistency constraint that clustering quality is continuously higher than an empirically determined threshold ; Only when the dual conditions of space centroid stability and structural compactness are satisfied, the target is judged to stabilize the human body.
7. The millimeter wave radar-based non-contact bed exit detection method of claim 1, wherein the physiological motion threshold value The maximum Euclidean norm between the window mass centers is smaller than the physiological motion threshold The method comprises the following steps: 。
8. The millimeter wave radar-based non-contact bed exit detection method of claim 1, wherein the cluster quality of the three consecutive windows is continuously higher than an empirically determined threshold The method comprises the following steps: 。
9. the non-contact bed leaving detection method based on millimeter wave radar according to claim 1, wherein two parameters including a neighborhood radius Eps and a core point threshold Pts are required for DBSCAN through DBSAN for point cloud clustering, wherein the neighborhood radius is set to be 0.2m, and the core point threshold is set to be 5.
10. A non-contact type bed leaving detection device based on millimeter wave radar is characterized by comprising an energy-entropy screening module, an OS-CFAR detection module and a time-space consistency evaluation module, wherein, The energy-entropy screening module is used for screening radar data windows potentially containing human targets through a self-adaptive energy-entropy dual gating mechanism; the OS-CFAR detection module is used for detecting the OS-CFAR by adopting ordered statistics of constant false alarm rate on the screened window, detecting effective human targets and bed targets, and further screening by combining morphological filtering and self-adaptive area threshold values; The space-time consistency evaluation module is used for generating radar point clouds for the screened targets, extracting the radar point clouds only comprising human targets, and accurately distinguishing real human targets from dynamic interference through joint evaluation of point cloud clustering and space-time consistency.

Description

Non-contact type bed-leaving detection method based on millimeter wave radar Technical Field The invention belongs to the technical field of millimeter wave radar non-contact vital sign monitoring, and particularly relates to a non-contact type off-bed detection method based on a millimeter wave radar. Background The accurate and reliable off-bed state monitoring can provide real-time safety guarantee for solitary old people and disabled groups, can fight time for emergency rescue by timely early warning of abnormal off-bed (such as long-time non-return and off-bed risks), can provide quantitative data support for medical care, can evaluate sleep quality, rehabilitation progress and chronic disease risks by analyzing off-bed frequency, duration and other data, and can assist accurate care driven by data. In addition, the non-contact type off-bed monitoring technology can relieve the problem of shortage of pension resources, reduce manual inspection pressure, simultaneously give consideration to privacy protection, promote the living autonomy of the old, and has remarkable social value and application significance. Current off-bed monitoring technology presents a pattern of coexistence of traditional scheme limitations and emerging technological breakthroughs. The traditional contact scheme relies on pressure sensors, piezoelectric sensors and the like, is simple to operate, is easily interfered by the environment and has poor stability, and the traditional non-contact scheme such as infrared sensors and RGB cameras has high resolution but is sensitive to light conditions and has privacy leakage risk. In recent years, millimeter wave radars become research hotspots by virtue of privacy protection advantages of non-contact detection, independence of light, penetrating shielding objects and no image acquisition, and the related prior art has verified the technical feasibility. However, the conventional millimeter wave scheme still faces the key challenges that firstly, the static human body vital signals are difficult to distinguish from the bed body interference, and secondly, the vital signals are easy to be blocked when the human body lies on the side, so that detection is missed. Disclosure of Invention In view of the above, the invention provides a non-contact bed-leaving detection method based on millimeter wave radar, aiming at the problem that a stationary human body vital signal is weak and a stationary bed body is difficult to distinguish. The technical scheme for realizing the invention is as follows: in a first aspect, the invention relates to a non-contact type bed leaving detection method based on millimeter wave radar, which comprises the following specific processes: Firstly, screening radar data windows potentially containing human targets through a self-adaptive energy-entropy double gating mechanism; Secondly, detecting an OS-CFAR by adopting ordered statistics of constant false alarm rate aiming at the screened window, detecting effective human targets and bed targets, and further screening by combining morphological filtering and a self-adaptive area threshold method; Finally, radar point cloud generation is carried out on the screened targets, the radar point cloud only comprising human targets is extracted, and the real human targets and dynamic interference are accurately distinguished through joint evaluation of point cloud clustering and space-time consistency. Optionally, the method screens out the radar data window potentially containing the human target through the self-adaptive energy-entropy dual gating mechanism, which comprises the following specific processes: processing millimeter wave radar echo data in each sliding window, and calculating total energy of a distance-Doppler diagram in the region of interest; Calculating the total energy mean value and standard deviation of noise of the distance-Doppler graph, setting a strong signal energy threshold according to the total energy mean value and standard deviation, and screening out a high energy window which is obviously higher than the background by utilizing the strong signal energy threshold; for echo windows with weaker energy, renyi entropy is introduced to quantify spectral ordering Setting Reiyi entropy threshold values to screen out low entropy windows; the high-energy window and the low-entropy window are radar data windows potentially containing human targets. Optionally, the method for calculating the total energy of the distance-doppler plot in the region of interest comprises the following steps: the first 60 seconds of setting an empty scene to acquire radar echo data, acquiring millimeter wave radar bed body part echo original data, performing distance dimension FFT and Doppler dimension FFT on each sliding window, and calculating total energy of a distance-Doppler diagram in a region of interest (ROI); The strong signal energy threshold ,For the first 60s of each window background noise mean,To set a constant.