Villa elevator tumble detection system based on millimeter wave radar

Abstract

The invention relates to the technical field of elevator tumble detection systems, and discloses a villa elevator tumble detection system based on millimeter-wave radar, and the system comprises a millimeter-wave radar module which is fixedly installed at the center of the top of an elevator car; the signal processing module is electrically connected to the millimeter wave radar module and is used for processing the original radar data; and the tumble judgment module is electrically connected to the signal processing module. By adopting millimeter-wave radar non-contact detection and combining multi-level threshold judgment logic of'primary triggering-static confirmation-track assistance ', the system can effectively distinguish real falling and similar actions such as normal squatting and stooping, and the problem that a traditional single sensor or threshold method is high in false alarm rate is solved. And a Kalman filtering algorithm and a density-based clustering algorithm are adopted in the signal processing module, so that the precision of target tracking and point cloud segmentation is further ensured, and the analysis accuracy is ensured from a data source.

Description

Technical Field

[0001] This invention relates to the field of elevator fall detection system technology, specifically a villa elevator fall detection system based on millimeter-wave radar. Background Technology

[0002] With the increasing aging of society and rising demands for residential safety, the prevalence of elevators in villas is growing. However, the elderly face a higher risk of falls in the enclosed, private space of an elevator car due to declining physical function and the elevator's start-stop, acceleration, and deceleration. If a fall occurs and is not detected and rescued promptly, it can lead to serious consequences.

[0003] In implementing existing technologies, the applicant discovered that villa elevator cars are extremely small and enclosed spaces. However, existing solutions are mostly designed for open indoor environments, focusing on radar installation positions and angles without considering the complex multipath effects caused by the limited height of the elevator car top and the metal reflective surfaces of the four walls. Directly applying these solutions would result in detection blind spots and severe signal interference, making it impossible to achieve stable and accurate perception of the entire area inside the car. Based on this, the present invention designs a villa elevator fall detection system based on millimeter-wave radar to solve the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide a fall detection system for villa elevators based on millimeter-wave radar, which solves the problems of detection blind spots and severe signal interference in the prior art.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] A fall detection system for villa elevators based on millimeter-wave radar, the system comprising:

[0007] The millimeter-wave radar module is fixedly installed at the center of the top of the elevator car. It adopts frequency modulation continuous wave system and multiple input multiple output antenna technology to acquire raw radar data of targets inside the car in real time. The beam coverage angle of the millimeter-wave radar module completely covers the bottom area of ​​the elevator car.

[0008] The signal processing module, electrically connected to the millimeter-wave radar module, is used to process the raw radar data, including separating static point clouds from dynamic point clouds and extracting the target's three-dimensional coordinates and motion trajectory.

[0009] The fall detection module is electrically connected to the signal processing module and is used to determine whether a fall event has occurred based on the height change, motion speed characteristics and horizontal expansion of the point cloud through multi-level threshold judgment logic. The multi-level threshold judgment logic includes a primary triggering stage, a static confirmation stage and a trajectory-assisted judgment stage.

[0010] The control and execution module is electrically connected to the fall detection module and the elevator main control system, and is used to send a control signal to the elevator main control system to perform safety operations after confirming a fall event.

[0011] The system uses millimeter-wave radar for non-contact detection and multi-level decision logic to address the accuracy and real-time performance issues of fall detection inside elevator cars.

[0012] Preferably, the millimeter-wave radar module is installed at the center of the top of the elevator car, and its beam coverage angle can completely cover the bottom area of ​​the elevator car. Furthermore, an electromagnetic wave absorbing material layer is added between its mounting base and the metal wall of the top of the car to suppress multipath interference caused by the metal reflective surface.

[0013] Preferably, the signal processing module includes a target tracking unit, which uses a Kalman filter algorithm to perform state prediction and data association on the dynamic point cloud target. Its state prediction equation is expressed as:

[0014]

[0015] X k Let u be the state vector at time k, F be the state transition matrix, B be the control input matrix, and u be the state vector at time k. k Let w be the control input vector at time k. k Let k be the process noise vector at time k.

[0016] Preferably, the fall detection module executes multi-level threshold detection logic, including:

[0017] In the initial triggering stage, when the average height of the target dynamic point cloud continuously and monotonically decreases and falls below the first height threshold, a suspected fall state is triggered.

[0018] In the static confirmation phase, if the average height of the target static point cloud is lower than the second height threshold and continues to exceed the preset time threshold in the suspected fall state, it is initially determined to be a fall event.

[0019] In the trajectory-assisted judgment stage, the horizontal displacement of the target during the fall is calculated. When the horizontal displacement is greater than the horizontal displacement threshold, the fall event is confirmed.

[0020] Preferably, the fall determination module further includes a vital signs detection unit, which, based on micro-Doppler characteristics, extracts the breathing and heartbeat information of personnel by performing time-frequency analysis on radar echo signals, so as to monitor vital signs after a fall event.

[0021] Preferably, the signal processing module further includes an anomaly detection unit, which, based on a variational autoencoder, assists in determining fall anomalies by calculating the reconstruction loss of the input point cloud features. The loss function is expressed as:

[0022]

[0023] E represents the expected value of the probability distribution; X represents the point cloud feature vector input to the variational autoencoder, which is composed of features extracted from radar point cloud data; z represents the encoded variable of the point cloud feature vector X in the latent space of the variational autoencoder; q(z) represents the variational posterior probability distribution of the latent variable z given the input data X, which is learned by the encoder network; p(z) represents the prior probability distribution of the latent variable z; p(X|z) represents the likelihood probability distribution of the reconstructed input data X given the latent variable z, which is learned by the decoder network.

[0024] Preferably, when separating static and dynamic point clouds, the signal processing module uses a density-based clustering algorithm to segment the point cloud, and its core object discriminant expression is expressed as:

[0025]

[0026] Core(P) is a Boolean condition. If true, point P is marked as the core point. P represents any data point in the radar point cloud. N_ε(P) represents the set of all points in the neighborhood of point P with radius ε. MinPts is the preset minimum number of points required to form the core object.

[0027] Preferably, the fall detection module also integrates a vital signs detection unit, which, after a fall is detected, extracts the breathing and heart rate of the person by performing time-frequency analysis on the radar signal corresponding to the static point cloud; the vital signs detection unit estimates the frequency by calculating the peak value of the signal power spectrum, and its core calculation formula is expressed as:

[0028]

[0029] s(t) represents the preprocessed radar baseband signal containing vital signs information; PSD represents the power spectral density function, used to calculate the power distribution of the signal at different frequencies; f represents the frequency; argmax_f represents the operation of finding the frequency parameter f that maximizes the power spectral density function; f_v represents the estimated vital signs frequency.

[0030] Preferably, the fall detection system is communicatively connected to the elevator main controller, and sends a control signal to the elevator main controller when a fall event is detected in the elevator car;

[0031] The elevator main controller performs at least one of the following safety operations based on the control signal: immediately stop the elevator, control the elevator to slow down and level with the nearest floor, automatically open the car door and keep it open, and trigger an audible and visual alarm and a remote assistance notification.

[0032] Preferably, the method further includes a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the following steps:

[0033] The raw echo signal inside the elevator car is obtained using millimeter-wave radar.

[0034] The original echo signal is processed to separate the static point cloud and the dynamic point cloud, and the target is tracked to generate a motion trajectory.

[0035] Based on the point cloud and motion trajectory, a multi-level threshold judgment logic is used to determine whether a fall event has occurred.

[0036] If a fall event is detected, control commands are generated and output.

[0037] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0038] 1. In this invention, by employing millimeter-wave radar non-contact detection and combining it with a multi-level threshold judgment logic of "primary triggering - static confirmation - trajectory assistance", the system can effectively distinguish between a real fall and similar actions such as normal squatting or bending over, solving the problem of high false alarm rate of traditional single sensor or threshold methods. The Kalman filtering algorithm and density-based clustering algorithm used in the signal processing module further ensure the accuracy of target tracking and point cloud segmentation, ensuring the accuracy of analysis from the data source.

[0039] 2. In this invention, the millimeter-wave radar module is installed at the center of the top of the car. Its beam coverage angle has been optimized to completely cover the bottom area of ​​the car, eliminating blind spots. Compared with the camera-based method, this invention is not affected by changes in the light conditions inside the car and can work normally in a completely dark environment, achieving all-weather monitoring. At the same time, the radar signal processing and multi-level judgment process are optimized for real-time performance, enabling rapid response to fall events.

[0040] 3. In this invention, the millimeter-wave radar does not collect optical images, but only senses targets by processing point clouds and micro-Doppler signals, which fundamentally protects the privacy of the occupants. In addition, by adding a layer of absorbing material between the radar mounting base and the metal wall, the multipath interference caused by the metal environment inside the car is effectively suppressed, improving signal quality and system stability. Attached Figure Description

[0041] Figure 1 This is a block diagram of the overall system structure of the present invention;

[0042] Figure 2 This is a flowchart of the signal processing module of the present invention;

[0043] Figure 3This is a flowchart of the control and execution module of the present invention;

[0044] Figure 4 This is a flowchart of the vital signs detection unit of the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] Example 1;

[0047] Please see Figures 1-4 A fall detection system for villa elevators based on millimeter-wave radar is disclosed. The system includes: a millimeter-wave radar module, fixedly installed at the center of the elevator car's top, employing frequency-modulated continuous wave technology and multiple-input multiple-output antenna technology to acquire raw radar data of targets inside the car in real time; the millimeter-wave radar module's beam coverage angle completely covers the bottom area of ​​the elevator car; a signal processing module, electrically connected to the millimeter-wave radar module, for processing the raw radar data, including separating static and dynamic point clouds and extracting the target's three-dimensional coordinates and motion trajectory; a fall detection module, electrically connected to the signal processing module, for determining whether a fall event has occurred based on the point cloud's height changes, motion speed characteristics, and horizontal expansion, using multi-level threshold judgment logic, which includes a primary trigger stage, a static confirmation stage, and a trajectory-assisted judgment stage; and a control and execution module, electrically connected to the fall detection module and the elevator main control system, for sending control signals to the elevator main control system to execute safety operations after confirming a fall event.

[0048] The system utilizes millimeter-wave radar for non-contact detection and multi-level decision logic to address the accuracy and real-time performance issues of fall detection inside elevator cars.

[0049] The millimeter-wave radar module is installed at the center of the top of the elevator car. Its beam coverage angle can completely cover the bottom area of ​​the elevator car. Furthermore, an electromagnetic wave absorbing material layer is added between its mounting base and the metal wall at the top of the car to suppress multipath interference caused by the metal reflective surface.

[0050] The signal processing module includes a target tracking unit, which uses the Kalman filter algorithm to predict the state of dynamic point cloud targets and perform data association. Its state prediction equation is expressed as:

[0051]

[0052] Xk Let u be the state vector at time k, F be the state transition matrix, B be the control input matrix, and u be the state vector at time k. k Let w be the control input vector at time k. k Let k be the process noise vector at time k.

[0053] The working principle of this invention is as follows: The system transmits frequency-modulated continuous waves through a millimeter-wave radar module and receives reflected signals from targets inside the car using multiple-input multiple-output (MIMO) antenna technology, acquiring raw radar data containing distance, speed, and angle information. The millimeter-wave radar module is installed at the center of the car's top, and its beam coverage angle is precisely designed to completely cover the bottom area of ​​the car, ensuring no detection blind spots. To further improve signal quality, an electromagnetic wave absorbing material layer is added between the radar mounting base and the metal wall of the car top, effectively suppressing multipath interference caused by the metal reflective surface.

[0054] After receiving the raw radar data, the signal processing module first separates the static point cloud from the dynamic point cloud using a density-based clustering algorithm to eliminate environmental clutter interference. Subsequently, the target tracking unit uses a Kalman filter algorithm to predict the state of the target in the dynamic point cloud and associate the data. The state prediction equation is used to achieve accurate tracking of the target's trajectory, where the state vector Xk contains motion parameters such as the target's position and velocity.

[0055] The fall detection module receives processed point cloud data and motion trajectory information, and executes multi-level threshold judgment logic. In the initial triggering stage, it monitors changes in the average height of the target's dynamic point cloud. When a continuous monotonically decreasing height below a preset first height threshold is detected, a suspected fall state is triggered. In the static confirmation stage, the height characteristics of the target's static point cloud are further analyzed. If its average height is below a second height threshold and continuously exceeds a preset time threshold, a fall event is initially determined. Finally, in the trajectory-assisted judgment stage, the horizontal displacement of the target during the fall process is calculated. When the horizontal displacement exceeds a set threshold, the fall event is ultimately confirmed.

[0056] Upon receiving a fall confirmation signal, the control and execution module immediately sends a control command to the elevator main control system, triggering the corresponding safety protection mechanism.

[0057] Example 2;

[0058] Please see Figures 1-4 In this embodiment of the invention, the fall detection module executes multi-level threshold detection logic, including:

[0059] In the initial triggering phase, a suspected fall state is triggered when the average height of the target's dynamic point cloud continuously and monotonically decreases and falls below the first height threshold. In the static confirmation phase, if the average height of the target's static point cloud falls below the second height threshold and continues to exceed a preset time threshold while in the suspected fall state, a fall event is initially determined. In the trajectory-assisted judgment phase, the horizontal displacement of the target during the fall process is calculated, and the fall event is confirmed when the horizontal displacement exceeds the horizontal displacement threshold. The fall determination module also includes a vital signs detection unit, which, based on micro-Doppler characteristics, extracts breathing and heartbeat information of personnel through time-frequency analysis of radar echo signals to monitor vital signs after a fall event.

[0060] The signal processing module also includes an anomaly detection unit, which, based on a variational autoencoder, assists in determining fall anomalies by calculating the reconstruction loss of the input point cloud features. Its loss function is expressed as:

[0061]

[0062] E represents the expected value of the probability distribution; X represents the point cloud feature vector input to the variational autoencoder, which is composed of features extracted from radar point cloud data; z represents the encoded variable of the point cloud feature vector X in the latent space of the variational autoencoder; q(z) represents the variational posterior probability distribution of the latent variable z given the input data X, which is learned by the encoder network; p(z) represents the prior probability distribution of the latent variable z; p(X|z) represents the likelihood probability distribution of the reconstructed input data X given the latent variable z, which is learned by the decoder network.

[0063] When separating static and dynamic point clouds, the signal processing module uses a density-based clustering algorithm to segment the point clouds. Its core object discriminant expression is as follows:

[0064]

[0065] Core(P) is a Boolean condition. If true, point P is marked as the core point. P represents any data point in the radar point cloud. N_ε(P) represents the set of all points in the neighborhood of point P with radius ε. MinPts is the preset minimum number of points required to form the core object.

[0066] The working principle of this invention is as follows: the system further optimizes the fall detection logic and enhances the anomaly detection capability. The multi-level threshold judgment logic executed by the fall detection module adopts a progressive judgment strategy. In the initial triggering stage, the height change characteristics of the target's dynamic point cloud are monitored in real time. When the average height is detected to be continuously and monotonically decreasing and below the first height threshold, the system enters a height alert state. In the static confirmation stage, the static state of the target is judged. By analyzing whether the average height of the static point cloud is below the second height threshold and continues for more than a preset time, the system effectively distinguishes between a real fall and similar actions such as normal squatting.

[0067] The trajectory-assisted judgment stage introduces horizontal displacement parameters as an auxiliary judgment basis. By calculating the horizontal widening characteristics of the target during the fall, the accuracy of the judgment is further improved. When the target falls rapidly in the vertical direction while accompanied by obvious horizontal displacement, the system recognizes it as a typical fall feature.

[0068] The vital signs detection unit, based on the micro-Doppler effect, performs time-frequency analysis on radar echo signals to extract micro-motion features containing information about breathing and heartbeat. This unit achieves non-contact monitoring of vital signs by analyzing the frequency modulation generated in the radar echo by the minute displacements caused by chest rise and fall and heartbeat, providing crucial physiological parameters for emergency rescue after falls.

[0069] The anomaly detection unit adopts a variational autoencoder architecture, which learns the probability distribution of point cloud features in the latent space through the encoder network, and the decoder network realizes feature reconstruction.

[0070] Example 3;

[0071] Please see Figures 1-4 In this embodiment of the invention, the fall detection module also integrates a vital signs detection unit, which, after a fall is detected, extracts the breathing and heart rate of the person by performing time-frequency analysis on the radar signal corresponding to the static point cloud; the vital signs detection unit estimates the frequency by calculating the peak value of the signal power spectrum, and its core calculation formula is expressed as:

[0072]

[0073] s(t) represents the preprocessed radar baseband signal containing vital signs information; PSD represents the power spectral density function, used to calculate the power distribution of the signal at different frequencies; f represents the frequency; argmax_f represents the operation of finding the frequency parameter f that maximizes the power spectral density function; f_v represents the estimated vital signs frequency.

[0074] The fall detection system communicates with the elevator main controller. When a fall is detected inside the elevator car, the system sends a control signal to the main controller. The main controller then performs at least one of the following safety operations based on the control signal: immediately stops the elevator, slows the elevator to the nearest floor, automatically opens the car door and keeps it open, and triggers an audible and visual alarm and a remote emergency call. The system also includes a non-transitory computer-readable storage medium storing a computer program. When executed by a processor, this program performs the following steps: acquiring the raw echo signal inside the elevator car using millimeter-wave radar; processing the raw echo signal to separate static and dynamic point clouds and track the target to generate a motion trajectory; based on the point cloud and motion trajectory, determining whether a fall event has occurred according to multi-level threshold logic; and if a fall event is determined to have occurred, generating and outputting control commands.

[0075] The working principle of this invention is as follows: The vital signs detection unit automatically activates after a fall event is detected. It accurately extracts the breathing and heart rate of the person by performing time-frequency transformation and power spectral density analysis on the radar baseband signal s(t) corresponding to the static point cloud. This unit employs a peak detection algorithm to estimate the vital signs frequencies through calculation. The power spectral density function PSD is used to analyze the energy distribution of the signal at different frequency components, and the argmax_f operation determines the frequency corresponding to the peak power spectrum, thereby obtaining the breathing rate and heart rate parameters.

[0076] The system establishes a real-time communication connection with the elevator main controller. When a fall is confirmed, it immediately sends a control signal containing the event type and location information to the main controller. Based on the received signal, the elevator main controller activates a multi-level safety response mechanism. First, it immediately stops the elevator to avoid secondary injury. Then, it controls the elevator to run at a slow speed to level with the nearest floor, automatically opens the car door and keeps it open to facilitate rescue. At the same time, it triggers the audible and visual alarm device and sends a remote distress call.

[0077] Working principle: The system transmits and receives frequency-modulated continuous waves via a millimeter-wave radar module on the top of the car to acquire raw data covering the entire car. The signal processing module uses a density-based clustering algorithm to separate static and dynamic point clouds, and employs a Kalman filter algorithm to accurately track the target's three-dimensional motion trajectory.

[0078] The fall detection module executes multi-level threshold logic. In the initial stage, it monitors continuously decreasing height to trigger an early warning; in the static stage, it confirms a sustained low-height state; and in the trajectory stage, it analyzes horizontal displacement characteristics to comprehensively determine a fall event. Upon detecting a fall, the system immediately sends a signal to the elevator's main control unit, executing safety operations such as stopping operation, slowing down leveling, opening the car door, and triggering an alarm. Some embodiments also integrate a vital signs detection unit, which analyzes the micro-Doppler characteristics and power spectrum peaks of radar echo signals to non-contactly monitor breathing and heart rate, providing crucial physiological parameters for rescue.

[0079] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A fall detection system for villa elevators based on millimeter-wave radar, characterized in that, The system includes: The millimeter-wave radar module is fixedly installed at the center of the top of the elevator car. It adopts frequency modulation continuous wave system and multiple input multiple output antenna technology to acquire raw radar data of targets inside the car in real time. The beam coverage angle of the millimeter-wave radar module completely covers the bottom area of ​​the elevator car. The signal processing module is electrically connected to the millimeter-wave radar module and is used to process the raw radar data, including separating static point clouds from dynamic point clouds and extracting the target's three-dimensional coordinates and motion trajectory. The fall detection module is electrically connected to the signal processing module and is used to determine whether a fall event has occurred based on the height change, motion speed characteristics and horizontal expansion of the point cloud through multi-level threshold judgment logic. The multi-level threshold judgment logic includes a primary triggering stage, a static confirmation stage and a trajectory-assisted judgment stage. The control and execution module is electrically connected to the fall detection module and the elevator main control system, and is used to send a control signal to the elevator main control system to perform a safety operation after confirming a fall event; The system uses millimeter-wave radar for non-contact detection and multi-level decision logic to address the accuracy and real-time performance issues of fall detection inside elevator cars.

2. The villa elevator fall detection system based on millimeter-wave radar according to claim 1, characterized in that: The millimeter-wave radar module is installed at the center of the top of the elevator car. Its beam coverage angle can completely cover the bottom area of ​​the elevator car. Furthermore, an electromagnetic wave absorbing material layer is added between its mounting base and the metal wall of the top of the car to suppress multipath interference caused by the metal reflective surface.

3. The villa elevator fall detection system based on millimeter-wave radar according to claim 1, characterized in that, The signal processing module includes a target tracking unit, which uses a Kalman filter algorithm to predict the state of a dynamic point cloud target and perform data association. Its state prediction equation is expressed as: X k Let u be the state vector at time k, F be the state transition matrix, B be the control input matrix, and u be the state vector at time k. k Let w be the control input vector at time k. k Let k be the process noise vector at time k.

4. The villa elevator fall detection system based on millimeter-wave radar according to claim 1, characterized in that, The fall detection module executes multi-level threshold detection logic, including: In the initial triggering stage, when the average height of the target dynamic point cloud continuously and monotonically decreases and falls below the first height threshold, a suspected fall state is triggered. In the static confirmation phase, if the average height of the target static point cloud is lower than the second height threshold and continues to exceed the preset time threshold in the suspected fall state, it is initially determined to be a fall event. In the trajectory-assisted judgment stage, the horizontal displacement of the target during the fall is calculated. When the horizontal displacement is greater than the horizontal displacement threshold, the fall event is confirmed.

5. A fall detection system for villa elevators based on millimeter-wave radar according to claim 1, characterized in that: The fall detection module also includes a vital signs detection unit, which, based on micro-Doppler characteristics, extracts breathing and heartbeat information of personnel by performing time-frequency analysis on radar echo signals in order to monitor vital signs after a fall event.

6. A villa elevator fall detection system based on millimeter-wave radar according to claim 1, characterized in that, The signal processing module also includes an anomaly detection unit, which, based on a variational autoencoder, assists in determining fall anomalies by calculating the reconstruction loss of the input point cloud features. The loss function is expressed as: E represents the expected value of the probability distribution; X represents the point cloud feature vector input to the variational autoencoder, which is composed of features extracted from radar point cloud data; z represents the encoded variable of the point cloud feature vector X in the latent space of the variational autoencoder; q(z) represents the variational posterior probability distribution of the latent variable z given the input data X, which is learned by the encoder network; p(z) represents the prior probability distribution of the latent variable z; p(X|z) represents the likelihood probability distribution of the reconstructed input data X given the latent variable z, which is learned by the decoder network.

7. A villa elevator fall detection system based on millimeter-wave radar according to claim 1, characterized in that, When separating static and dynamic point clouds, the signal processing module uses a density-based clustering algorithm to segment the point clouds. Its core object discriminant expression is as follows: Core(P) is a Boolean condition. If true, point P is marked as the core point. P represents any data point in the radar point cloud. N_ε(P) represents the set of all points in the neighborhood of point P with radius ε. MinPts is the preset minimum number of points required to form the core object.

8. A fall detection system for villa elevators based on millimeter-wave radar according to claim 1, characterized in that, The fall detection module also integrates a vital signs detection unit, which, after a fall is detected, extracts the breathing and heart rate of the person by performing time-frequency analysis on the radar signal corresponding to the static point cloud. The vital signs detection unit estimates the frequency by calculating the peak value of the signal power spectrum, and its core calculation formula is expressed as: s(t) represents the preprocessed radar baseband signal containing vital signs information; PSD represents the power spectral density function, used to calculate the power distribution of the signal at different frequencies; f represents the frequency; argmax_f represents the operation of finding the frequency parameter f that maximizes the power spectral density function; f_v represents the estimated vital signs frequency.

9. A fall detection system for villa elevators based on millimeter-wave radar according to claim 1, characterized in that: The fall detection system is connected to the elevator main controller and sends a control signal to the elevator main controller when it detects a fall event in the elevator car. The elevator main controller performs at least one of the following safety operations based on the control signal: immediately stop the elevator, control the elevator to slow down and level with the nearest floor, automatically open the car door and keep it open, and trigger an audible and visual alarm and a remote assistance notification.

10. A villa elevator fall detection system based on millimeter-wave radar according to claim 1, characterized in that, It also includes a non-transitory computer-readable storage medium on which a computer program is stored, which, when executed by a processor, implements the following steps: The raw echo signal inside the elevator car is obtained using millimeter-wave radar. The original echo signal is processed to separate the static point cloud and the dynamic point cloud, and the target is tracked to generate a motion trajectory. Based on the point cloud and motion trajectory, a multi-level threshold judgment logic is used to determine whether a fall event has occurred. If a fall event is detected, control commands are generated and output.

Images