Fall monitoring method and device, electronic equipment and storage medium

Abstract

The invention provides a tumble monitoring method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining time dimension parameters and space dimension parameters related to a tumble event under the condition that the tumble event of a monitored object is determined, the time dimension parameter is used for representing the continuous state of the monitored object after falling down, and the space dimension parameter is used for representing the space position relation between the monitored object and a surrounding environment object when the monitored object falls down; and determining the severity level of the tumble event based on the time dimension parameter and the space dimension parameter, thereby effectively solving the defects that only whether tumble can be judged and the severity cannot be deeply evaluated in the prior art, realizing graded alarm, improving the response efficiency of a guardian, and avoiding false alarm or missing of a high-risk situation.

Description

Technical Field

[0001] This application relates to the field of monitoring and early warning technology, and in particular to a method and apparatus, electronic device and storage medium for monitoring falls. Background Technology

[0002] With the increasing global trend of population aging, home-based elder care has become the mainstream model. Against this backdrop, ensuring the safety of elderly people living alone or in empty nests, especially responding promptly to accidental falls, has become a focus of attention for both society and the technology sector. Currently, various fall detection technologies for the elderly exist on the market. The mainstream solutions include those based on wearable devices (such as smart bracelets with built-in accelerometers or emergency button pendants) and those based on environmental sensors. Among these, solutions utilizing cameras for computer vision analysis have attracted significant attention due to their non-intrusive monitoring advantages.

[0003] However, existing technical solutions, especially those based on computer vision, still have significant limitations in practical applications. Most of these solutions rely on image recognition algorithms (such as human posture estimation) to determine whether an elderly person's body is horizontal on the ground, using this as the basis for triggering an alarm. They can answer the question of "whether a fall has occurred," but their analytical dimensions are usually extremely simplistic, relying solely on the single metric of "duration of fall," or completely failing to differentiate between different scenarios. This "one-size-fits-all" detection logic ignores the complexity of fall events, failing to distinguish between a harmless slip and a potentially dangerous fall involving the head near sharp furniture that could lead to head injuries.

[0004] Therefore, a core technical problem that urgently needs to be solved in this field is that existing technologies cannot effectively assess the severity of fall incidents. Existing solutions generally lack the ability to comprehensively analyze the temporal, spatial, and environmental hazards during a fall, resulting in an inability to distinguish the danger levels of different fall incidents. This directly leads to the failure of alarm mechanisms—either due to frequent false alarms or overreactions causing "alarm fatigue" and leading to caregivers becoming complacent; or due to the failure to identify truly high-risk fall incidents, resulting in missed opportunities for optimal rescue, thus failing to achieve precise and effective differentiated alarms and interventions that match the risk level. Summary of the Invention

[0005] This application provides a method, apparatus, electronic device, and storage medium for monitoring falls, in order to overcome the shortcomings of existing technologies that cannot effectively assess the severity of fall events, and to achieve the technical effect of classifying and assessing the severity of fall events.

[0006] This application provides a method for monitoring falls, including: When it is determined that the monitored object has fallen, time dimension parameters and spatial dimension parameters related to the fall event are obtained. The time dimension parameters are used to characterize the continuous state of the monitored object after the fall, and the spatial dimension parameters are used to characterize the spatial positional relationship between the monitored object and the surrounding environmental objects when the fall occurs. The severity level of the fall event is determined based on the time dimension parameters and the spatial dimension parameters.

[0007] The method for monitoring falls provided in this application further includes: determining whether a fall event has occurred on the monitored object based on monitoring data acquired from at least one sensor.

[0008] According to the fall monitoring method provided in this application, the sensor includes a wireless communication device and a video camera; based on monitoring data acquired from at least one sensor, determining whether a fall event has occurred on the monitored object specifically includes: The wireless communication device monitors the channel status information of the wireless signal in real time, and determines that abnormal activity has occurred when the fluctuation amplitude of the channel status information exceeds a first preset threshold. In response to determining that the abnormal activity has occurred, video data is acquired from the video camera, the video data being data from a preset time period prior to the time when the abnormal activity occurred; The video data is subjected to human posture recognition, and when the key points of the torso of the monitored object remain on the ground for more than a first preset time, it is determined that the monitored object has fallen.

[0009] According to the method for monitoring falls provided in this application, the sensor includes a depth camera; The steps for obtaining spatial dimension parameters related to the fall event specifically include: scanning the environment where the monitored object is located using the depth camera to establish a three-dimensional point cloud model of the scene; segmenting and identifying furniture objects in the three-dimensional point cloud model, and performing geometric analysis on the model of the furniture objects to determine their sharp edges or sharp corners as danger zones; calculating the three-dimensional spatial distance between the head of the monitored object and the danger zone, and using the three-dimensional spatial distance as the spatial dimension parameter.

[0010] The method for monitoring falls provided in this application includes the following steps for obtaining time dimension parameters related to the fall event: after determining that a fall event has occurred, continuously calculating the total duration for which the monitored object's body remains stationary on the ground, and using the total duration as the time dimension parameter.

[0011] According to the method for monitoring falls provided in this application, the severity level of the fall event is determined based on the time dimension parameter and the spatial dimension parameter, specifically including: comparing the spatial dimension parameter with at least one spatial threshold to determine the severity of the spatial dimension; comparing the time dimension parameter with at least one time threshold to determine the severity of the time dimension; and determining the final severity level of the fall event based on the severity of the spatial dimension and the severity of the time dimension.

[0012] According to the method for monitoring falls provided in this application, after determining the final severity level of the fall event, the method further includes: executing a corresponding graded alarm step based on the determined severity level; The tiered alarm process includes: when the severity level is low risk, sending an SMS reminder to a preset guardian; when the severity level is medium risk, sending an in-app alarm message to the guardian; and when the severity level is high risk, automatically dialing the guardian's alarm phone number.

[0013] According to the method for monitoring falls provided in this application, after acquiring video data from the video camera, the method further includes: applying a crowd category recognition algorithm to the human body image in the video data to obtain a category recognition result; If the category identification result matches the preset category of the monitoring object, the step of performing human posture recognition on the video data continues.

[0014] This application also provides a device for monitoring falls, comprising: The parameter acquisition module is used to acquire time dimension parameters and spatial dimension parameters related to the fall event when it is determined that the monitored object has fallen. The time dimension parameters are used to characterize the continuous state of the monitored object after the fall, and the spatial dimension parameters are used to characterize the spatial positional relationship between the monitored object and the surrounding environmental objects when the fall occurs. The severity determination module is used to determine the severity level of the fall event based on the time dimension parameters and the spatial dimension parameters.

[0015] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the fall monitoring method as described above.

[0016] This application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for monitoring falls as described above.

[0017] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the fall monitoring method as described above.

[0018] The method and apparatus for monitoring falls provided in this application comprehensively assess the severity of a fall event by introducing two dimensions: time and space. The time dimension parameter characterizes the continuous state of the monitored object after the fall, while the space dimension parameter characterizes the spatial positional relationship between the monitored object and surrounding objects at the time of the fall. The time and space dimension parameters are then combined to determine the severity level of the fall event. This effectively solves the shortcomings of existing technologies that can only determine whether a fall has occurred but cannot deeply assess the severity, enabling graded alarms, improving the response efficiency of caregivers, and avoiding false alarms or missing high-risk situations. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the fall monitoring method provided in this application.

[0021] Figure 2 This is a schematic diagram of an elderly person falling, as provided in this application.

[0022] Figure 3 This is a schematic diagram of the fall monitoring device provided by the present invention.

[0023] Figure 4 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] In existing technologies, fall detection technologies are mainly divided into three types: The first method is based on video monitoring, which records the elderly’s activities through cameras and allows guardians to remotely view and assess the safety situation. However, it relies on manual monitoring, cannot provide proactive alerts, and is subject to delays. The second type is based on monitoring with special sensor equipment. The elderly wear watches or pendants containing accelerometers and gyroscopes, which can capture abnormal human movements in real time. However, the elderly need to wear them continuously, which can cause problems such as discomfort, signal interference and equipment wear and tear. The third method is based on computer vision to recognize human posture. It uses cameras and algorithms to analyze human posture to determine if a fall has occurred. It provides seamless monitoring throughout the process, but it requires high computing resources and often uses "whether a person has fallen" as a single indicator, making it impossible to analyze the severity of the fall in depth.

[0026] The common limitation of these existing technologies is that they can only determine whether a fall has occurred, but it is difficult to assess the severity of the fall, such as whether the elderly person can get up on their own after a fall, or whether the fall scene is dangerous. This results in an inaccurate alarm mechanism, which may lead to false alarms or delays in rescue.

[0027] To address the aforementioned issues, this application proposes an innovative solution. By combining parameters from both temporal and spatial dimensions, the severity of fall events is comprehensively assessed. Specifically, the temporal dimension focuses on the duration of stillness after a fall, reflecting the elderly person's physical reaction and recovery status; while the spatial dimension considers the spatial relationship between the elderly person and surrounding objects at the time of the fall, such as the distance between the head and dangerous furniture, thereby accurately determining the danger level of the fall event. Based on this comprehensive analysis of these two dimensions, this application enables tiered alarms, taking different alarm measures according to the severity of the fall. This avoids the confusion caused by false alarms and ensures timely response to high-risk events, providing a more reliable technical guarantee for the home safety of elderly people living alone.

[0028] Before describing the technical solutions of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained illustratively.

[0029] Wireless communication equipment: This is a device capable of transmitting and receiving wireless signals, which propagates wireless signals into the surrounding space via an antenna during operation. In this application, the wireless communication equipment can sense changes in wireless signals in the indoor environment in real time. When an elderly person experiences abnormal activity such as a fall, the propagation path and strength of the wireless signal will change, and the equipment can quickly detect this change, thereby triggering subsequent monitoring procedures.

[0030] A video camera is a device capable of capturing continuous image signals, typically composed of a lens and image sensor. It can be installed in areas frequented by the elderly to collect video images in real time. Unlike ordinary surveillance cameras, the video camera in this application is deeply integrated with a fall detection system. When the wireless communication device detects an abnormal signal change, the system immediately instructs the video camera to acquire video clips for a specific time period, providing intuitive image evidence for subsequent human posture analysis.

[0031] Depth cameras are devices that add depth sensing capabilities to traditional video cameras. They measure the precise distance between objects and the camera using technologies such as infrared projection, structured light, or time-of-flight. They can generate depth maps of scenes, providing data support for building 3D spatial models. Depth cameras can accurately identify the 3D spatial relationship between an elderly person who has fallen and surrounding furniture and other objects, providing crucial information for assessing the severity of the fall.

[0032] Channel State Information (CSI) is a key parameter reflecting the transmission characteristics of a wireless channel during wireless communication. It includes multi-dimensional information such as signal strength, phase, and delay. CSI is extremely sensitive to environmental changes; even minor alterations such as human activity and furniture placement can affect wireless signal propagation, leading to fluctuations in CSI. This application utilizes this characteristic, using CSI as a "tactile sensor" for indoor environmental monitoring to perceive the elderly person's activity status in real time, providing an efficient and low-power initial screening method for fall detection.

[0033] Time-dimensional parameters refer to quantitative indicators used to describe changes in an elderly person's physical condition over time after a fall. For example, the duration an elderly person remains stationary on the ground after a fall directly reflects their physical condition and ability to recover independently. The time-dimensional parameters are set based on statistical analysis of numerous fall cases; different time thresholds correspond to different levels of fall risk, providing important references for subsequent severity assessments.

[0034] Spatial dimension parameters are quantitative indicators that describe the spatial relationship between an elderly person's body and surrounding objects during a fall. For example, the distance between an elderly person's head and the edge of a sharp piece of furniture when they fall. These parameters are calculated using depth information obtained from a depth camera and combined with human posture recognition technology to accurately determine the dangerous environment in which the elderly person falls, providing a direct basis for assessing the potential risk of injury from the fall.

[0035] A 3D point cloud model is a model that describes the geometry of an object or scene using a large amount of 3D point coordinate data. In this application, depth data collected by a depth camera is converted into a 3D point cloud to construct a scene model of the elderly person's activity area. This model can accurately identify furniture objects in the scene and determine their shape features and dangerous areas, such as sharp edges and corners, through geometric analysis. When the elderly person falls, the system can quickly locate their position in the 3D point cloud model and calculate their distance from surrounding dangerous objects, providing a basis for determining spatial dimensional parameters.

[0036] Human posture recognition is a technology based on computer vision that automatically analyzes and understands human posture. It typically utilizes deep learning algorithms to detect and track key parts of the human body (such as joints and torso) and extract posture features. In this application, human posture recognition technology is used to analyze image data captured by a video camera to determine whether an elderly person has fallen and their body posture after the fall. By continuously tracking key points of the human torso, the time dimension parameters of the elderly person's body remaining still after a fall can be accurately calculated, providing core support for confirming the fall event and assessing its severity.

[0037] Trunk key points are critical nodes in a human posture model used to describe the posture and position of the upper body, such as the center points of the shoulders, chest, waist, and hips. These key points play a central role in human movement and posture changes, reflecting the body's balance and postural changes. In fall detection, real-time monitoring of trunk key points can determine whether an elderly person has changed from a standing or sitting position to an abnormal fall, which is an important basis for confirming a fall. Furthermore, the time spent at trunk key points after a fall is also one of the time-dimensional parameters for assessing the severity of the fall.

[0038] In this embodiment, the hardware environment consists of multiple key devices and network components, which work together to monitor, analyze, and alert on falls in elderly people living alone. The following is a detailed description of the hardware environment: Wireless communication devices: Wireless access points (such as Wi-Fi routers) deployed in the home environment, equipped with Channel State Information (CSI) acquisition capabilities. Their function is to monitor changes in the transmission characteristics of wireless signals in real time. When abnormal fluctuations are detected (such as signal interference caused by human activity), the system is triggered to further analyze potential fall events. These wireless communication devices connect to other devices in the home network (such as smart cameras) via FTTR (Fiber to the Room) networking.

[0039] Video Camera: A high-definition smart camera installed in the main areas where the elderly person is active (such as the living room and bedroom). The camera has high-resolution image acquisition capabilities and supports real-time video streaming. When the wireless communication device detects abnormal fluctuations, the video camera captures video clips before and after the abnormal moment according to instructions, providing image data support for subsequent human posture analysis.

[0040] Depth camera: A device that works in conjunction with a video camera and is installed in the same area. Depth cameras acquire depth information of a scene through infrared projection or structured light technology, generating 3D point cloud models, identifying objects in the scene (such as furniture), and calculating their spatial relationship to the human body. In fall incident analysis, depth cameras are used to determine the distance between an elderly person and surrounding dangerous objects (such as furniture with sharp edges) when they fall.

[0041] FTTR (Fiber to the Room) devices: As a core component of home networks, FTTR devices extend high-speed, stable network connections to every room in the home via fiber optic cables. FTTR devices not only provide network support for terminal devices such as video cameras and depth cameras, but also integrate edge computing capabilities to perform preliminary analysis of CSI data and determine whether to trigger the video analysis module. Furthermore, FTTR devices support WiFi 6 or higher wireless communication protocols, ensuring real-time transmission of video and depth data.

[0042] Server or edge computing module: A computing unit deployed in a home network or in the cloud, responsible for running human posture recognition algorithms and fall severity analysis models. The server analyzes human posture using deep learning technology to identify whether the elderly person is in a fall state, and comprehensively assesses the severity of the fall by combining time-dimensional parameters (such as dwell time at key points on the torso) and spatial-dimensional parameters (such as the distance between the head and furniture). Based on the assessment results, the server triggers the corresponding alarm mechanism.

[0043] Guardian terminal device: The guardian's smartphone, tablet, or computer connected to the system via home network or the internet. Once a fall event is confirmed and assessed, the system sends corresponding alarm information (such as SMS, in-app notification, or automatic phone call) to the guardian's terminal based on the severity level, ensuring the guardian can respond promptly.

[0044] The entire hardware environment achieves high-speed and stable connection between devices through the FTTR network. Combined with the CSI screening function of wireless communication devices, the data acquisition capabilities of video and depth cameras, and the powerful computing capabilities of the server, a highly efficient and accurate fall monitoring and alarm system for elderly people living alone is constructed.

[0045] Figure 1 This is one of the flowcharts illustrating the fall monitoring method provided in this application, such as... Figure 1 As shown, the method includes the following: Step 101: If it is determined that a fall event has occurred on the monitored object, obtain the time dimension parameters and spatial dimension parameters related to the fall event.

[0046] The time dimension parameter is used to characterize the continuous state of the monitored object after it falls, and the spatial dimension parameter is used to characterize the spatial positional relationship between the monitored object and surrounding environmental objects when it falls.

[0047] In step 101, the system first triggers the video camera to capture video data before and after the fall, and obtains depth information of the scene through the depth camera. Then, the system uses human posture recognition technology to accurately locate key points of the human torso and calculates the time the person remains still on the ground as a time dimension parameter. Simultaneously, it constructs a 3D point cloud model based on the depth information, identifies furniture objects in the surrounding environment, and calculates the 3D spatial distance between the person's head and the danger zone of the furniture as a spatial dimension parameter. These two parameters together provide a quantitative basis for subsequent analysis of the severity of the fall. Specifically, the time dimension parameter reflects the elderly person's physical recovery ability after the fall; a prolonged stillness time may indicate that the elderly person is unable to get up independently. The spatial dimension parameter assesses the degree of danger in the environment where the elderly person was at the time of the fall; for example, proximity of the head to a sharp object may increase the risk of injury.

[0048] By combining these two dimensions, the system can more comprehensively assess the severity of a fall incident, providing an important basis for taking appropriate warning and rescue measures.

[0049] Step 102: Determine the severity level of the fall event based on the time dimension parameters and spatial dimension parameters.

[0050] In step 102, the severity level of the fall event is determined by comprehensively analyzing time-dimensional parameters (such as the duration of time the torso's key points remain stationary on the ground) and spatial-dimensional parameters (such as the minimum distance between the elderly person's head and a dangerous object). Specifically, different combinations of these two dimensions are mapped to pre-set severity levels, such as minor, severe, and extremely severe, and corresponding alarm actions are triggered accordingly.

[0051] For example, if the torso remains still for 5 to 15 seconds and the distance between the head and the dangerous object is greater than or equal to 20 centimeters, it may be judged as a minor fall and only a text message reminder will be sent; however, if the torso remains still for much longer than 60 seconds and the distance between the head and the dangerous object is extremely small, it may be judged as a particularly serious fall and the guardian will be automatically called and emergency services will be summoned.

[0052] The fall monitoring method provided in this application introduces two dimensions, time and space, to comprehensively assess the severity of a fall event when it is determined from monitoring data. The time dimension parameter characterizes the continuous state of the monitored object after the fall, while the space dimension parameter characterizes the spatial positional relationship between the monitored object and surrounding objects at the time of the fall. The time and space dimension parameters are then combined to determine the severity level of the fall event. This effectively solves the shortcomings of existing technologies that can only determine whether a fall has occurred but cannot deeply assess the severity, enabling graded alarms, improving the response efficiency of caregivers, and avoiding false alarms or missing high-risk situations.

[0053] Furthermore, in this embodiment of the application, the method further includes: determining whether a fall event has occurred on the monitored object based on monitoring data obtained from at least one sensor.

[0054] In this embodiment, the sensors involved include wireless communication devices and video cameras. The wireless communication devices are primarily used for real-time monitoring of channel state information (CSI) of wireless signals. Channel state information refers to the characteristics of the propagation path of a signal from the transmitter to the receiver during wireless communication, including parameters such as signal strength and phase change. When the monitored object (such as an elderly person) is active indoors, their body movements will affect the propagation of the wireless signal, thereby causing fluctuations in the channel state information. Specifically, when the fluctuation amplitude of the channel state information exceeds a pre-set first threshold, the system determines that abnormal activity may have occurred.

[0055] Once abnormal activity is detected, the system will respond immediately by acquiring video data from the video cameras. This video data refers to data from a preset time period preceding the occurrence of the abnormal activity. For example, the system might acquire video footage from 30 seconds before the abnormal activity occurs until the moment it does. Video cameras are typically installed in key locations within the monitored area, such as the living room or bedroom, to ensure that the activities of the monitored individuals are captured.

[0056] After acquiring video data, the system performs human pose recognition on this data. Human pose recognition is a computer vision technology that analyzes features such as human contours and joint positions in video images to determine the posture and movement of the human body. In this application, particular attention is paid to the torso key points of the monitored object. Torso key points refer to specific locations on the human torso, such as the shoulders and waist, which exhibit significant posture changes when a person falls. The system determines whether a fall has occurred by identifying whether these key points contact the ground and the duration of contact. Specifically, if the torso key points remain on the ground for more than a first preset duration (e.g., more than 5 seconds), a fall is determined to have occurred.

[0057] Through the above steps, this embodiment of the application effectively combines the advantages of wireless communication devices and video cameras. It utilizes fluctuations in channel state information as a preliminary basis for judging abnormal activity, and then further confirms this through human posture recognition of video data, thereby achieving accurate detection of falls by the monitored object. This method not only improves the timeliness and accuracy of fall detection but also reduces the possibility of false alarms, ensuring the reliability and practicality of the monitoring system.

[0058] Furthermore, in the process of monitoring fall events, in order to ensure that the system can accurately focus on the "elderly" group that needs attention and avoid false alarms caused by non-target events such as children playing or adults' activities, thereby improving the system's usability and user trust, this application proposes a visual feature classification method based on deep learning, and uses it as the core implementation method of the system, which is cleverly deployed on the edge computing unit of the FTTR device.

[0059] In terms of data collection and annotation, a large-scale visual database covering three groups of people: the elderly, adults, and children was constructed. This dataset was carefully designed to include family activity videos taken under different lighting conditions, shooting angles, and clothing conditions, and each frame of the people in the video was accurately labeled with their respective categories, providing a high-quality data foundation for subsequent model training.

[0060] The model training process is divided into two stages: feature extraction and classifier training. In the feature extraction stage, mainstream convolutional neural networks, such as ResNet-50, or lightweight networks like MobileNetV3, are used as the backbone network to extract rich deep visual features from images. In the classifier training stage, the model focuses on learning to distinguish key physiological and behavioral characteristics of different age groups. Static physiological characteristics include height, head-to-body ratio, and body contours, such as the slight hunchback commonly seen in the elderly. Dynamic behavioral characteristics are captured by analyzing consecutive frames, such as typical movement patterns of different groups—the gait of the elderly is usually relatively slow and with small strides, while children's movements are characterized by high speed, irregularity, and large amplitude, such as running, jumping, and rolling on the ground.

[0061] The lightweight model is deployed at the edge. When CSI triggers the video analysis process, the system first inputs the captured video frames into the classification model. Only when the model outputs the category with the highest confidence level as "elderly" will the system continue with the subsequent fall detection and severity analysis processes, thus ensuring the accuracy of monitoring. If the identification result is not "elderly," the process will stop to avoid unnecessary interference with caregivers.

[0062] Furthermore, this application proposes an extended implementation method: a method based on gait and skeletal keypoint temporal analysis. This approach does not rely on single-frame image classification. Instead, it analyzes a sequence of human skeletal keypoints over a period of time. After extracting continuous skeletal point data using human pose estimation algorithms (such as OpenPose), the system deeply analyzes its temporal characteristics, including gait frequency, stride length, stability of joint flexion angles, and the vertical displacement amplitude of the body's center of mass. Through these quantified biomechanical indicators, it can more robustly distinguish the relatively regular and stable gait of the elderly from that of middle-aged or young adults or children, further enhancing the system's recognition ability in complex environments, especially exhibiting stronger anti-interference capabilities when facing occlusion and changes in lighting.

[0063] In practical applications, after acquiring video data from the video camera, the system applies a crowd category recognition algorithm to the human images within the video data to obtain category recognition results. Only when this result matches the preset category of the monitored object—namely, "elderly"—will the system proceed to the step of human posture recognition on the video data, thereby initiating the subsequent fall detection process. This series of carefully designed steps ensures that the system can efficiently and accurately serve the monitoring needs of the elderly population, providing solid technical support for the safety of elderly people living alone.

[0064] Furthermore, regarding the spatial dimension parameters related to fall events, the core lies in the three-dimensional spatial distance between the head of the monitored object and various danger zones, which provides crucial data support for assessing the degree of danger of fall events.

[0065] Depth cameras play a crucial role in monitoring fall incidents. By emitting infrared light or employing structured light technology, they meticulously scan the environment in which the monitored object is located, collecting light information reflected from object surfaces to create a 3D point cloud model of the scene. This model, composed of numerous points with spatial location information, accurately reconstructs the shape, size, and positional relationships of various objects within the scene.

[0066] After obtaining the 3D point cloud model, the system uses advanced object detection and segmentation algorithms (such as the PointGroup algorithm) to accurately segment and identify the furniture objects in the model. The PointGroup algorithm combines the geometric features and semantic information of the point cloud to effectively distinguish different categories of furniture, assigning each piece of furniture a unique identifier for easy subsequent analysis. This is similar to clearly outlining and labeling each piece of furniture in a complex scene image.

[0067] Subsequently, in the crucial step of identifying hazardous areas, the system performs geometric feature calculations on the point cloud clusters of objects identified as "tables," "cabinets," and other hard furniture. By analyzing the curvature and normal vectors of each point on the point cloud surface, it automatically detects and marks areas with drastic changes in curvature or abrupt changes in normal vectors. These areas correspond to the sharp edges and corners of the furniture, constituting "hazardous areas" that require special attention.

[0068] In practical applications, when an elderly person falls, the system combines information about identified furniture and its danger zones to accurately calculate the three-dimensional spatial distance between the elderly person's body (especially the head) and these danger zones. The smaller the distance, the higher the potential danger of the fall. The system further combines time-dimensional parameters (such as the duration of stillness after the fall) and spatial-dimensional parameters (such as the distance between the head and the danger zones) to comprehensively assess the severity of the fall.

[0069] Through this series of steps, the system can more comprehensively and accurately assess the risk level of a fall, thereby providing guardians with timely and effective warning information to ensure that the elderly receive necessary attention and assistance immediately after a fall.

[0070] For the time-related parameters associated with fall events, the core lies in continuously calculating the total duration for which the monitored subject's body remains stationary on the ground. The system analyzes changes in human posture within video data to determine if the body is stationary and begins timing. Once the time the monitored subject's body remains stationary on the ground exceeds a preset threshold, the system uses this duration as a time-related parameter for subsequent fall severity assessment. Obtaining this parameter is crucial for timely and accurate assessment of the urgency of the fall event, enabling the system to react quickly and take appropriate alert measures.

[0071] Determining the severity level of a fall is a crucial step in monitoring fall events. This step comprehensively considers both temporal and spatial parameters, quantifying the severity of the fall by comparing it with preset thresholds.

[0072] Specifically, the time dimension parameter focuses on the total duration the elderly person's body remains stationary on the ground after a fall. The system compares this duration to preset time thresholds. For example, if the elderly person remains stationary on the ground for more than 15 seconds but less than 1 minute, this might be considered a moderate fall event; if the stationary time exceeds 1 minute, it might be considered a serious fall event. The spatial dimension parameter focuses on the distance between the elderly person's head and surrounding dangerous objects (such as sharp furniture edges). The system compares this distance to preset spatial thresholds to assess the environmental hazard level at the time of the fall. For example, if the distance between the elderly person's head and a dangerous object is less than 5 centimeters, this might indicate a high risk of injury.

[0073] After obtaining the severity in both the time and spatial dimensions, the system combines the results from these two dimensions to determine the final severity level of the fall event. For example, if the time dimension shows that the elderly person remained still on the ground for 30 seconds, while the spatial dimension shows that the elderly person's head was 10 centimeters away from the dangerous object, then the system will comprehensively classify this fall event as a medium-risk level.

[0074] Based on the determined severity level, the system will execute corresponding tiered alert procedures. For low-risk falls, the system will send a text message to the designated guardian, informing them that a minor fall may have occurred and advising them to monitor the elderly person's condition. For medium-risk falls, the system will not only send a text message but also push an alert to the guardian's mobile app to draw their attention and encourage them to contact the elderly person as soon as possible to confirm their safety. For high-risk falls, the system will automatically call the guardian to ensure they are quickly informed and can take emergency rescue measures.

[0075] This tiered alarm mechanism is designed to ensure that guardians can obtain relevant information about elderly people's falls in the shortest possible time and take appropriate measures according to the severity of the incident, thereby maximizing the safety of the elderly.

[0076] Furthermore, in this embodiment, to achieve a more refined monitoring approach from general whole-house monitoring to specific physical rooms, while simultaneously ensuring accurate triggering and privacy protection, the system employs a range-of-identity implementation scheme based on FTTR network topology and physical space mapping. Specifically, during the FTTR network initialization configuration phase, the accompanying app guides the user to name each sub-router device and specify its physical location, for example, naming the living room sub-router "Living Room Router". In this way, the system can establish a detailed "physical room-network device ID" mapping table, laying the foundation for subsequent accurate location and monitoring.

[0077] During monitoring, the system utilizes CSI (Channel State Information)-based area positioning technology. Since different sub-routes cover different physical spaces, human activity within a certain area primarily causes significant fluctuations in the CSI data reported by the sub-routes within that area. By analyzing the source of the CSI data packets, i.e., the sub-route ID, the main gateway can quickly pinpoint the physical room where the abnormal event occurred. This process not only improves monitoring accuracy but also reduces the system's computational burden.

[0078] To achieve precise triggering and privacy protection, after locating the room where the incident occurred, such as the "bedroom," the system only activates the cameras associated with that area for analysis, while cameras in other rooms remain inactive. This design ensures that only necessary devices are activated, effectively protecting the privacy of areas outside the incident zone.

[0079] In addition to CSI-based area positioning, the system also provides two extended implementation methods to enhance the recognition range: The first method is based on sound event localization, which integrates microphone arrays on each sub-route of the FTTR. When a fall occurs, it is usually accompanied by a heavy "thud". The system analyzes the time difference of arrival (TDOA) or intensity difference (IDOA) of the sound signal to the microphones in different rooms, and uses multi-point localization algorithms such as triangulation to calculate the approximate location of the sound source, thereby determining the room where the fall occurred and activating the corresponding camera.

[0080] The second expansion method is RSSI positioning based on Bluetooth beacons. The monitored individual wears a low-power Bluetooth beacon wristband or pendant. FTTR sub-routers distributed throughout the rooms simultaneously act as Bluetooth receivers, continuously monitoring the received signal strength indicator (RSSI) of the beacon. Since the sub-router closest to the beacon receives the strongest RSSI value, the system can estimate the elderly person's room in real time by comparing the RSSI values ​​of each sub-router, achieving more direct area identification.

[0081] Through the aforementioned solutions and their extensions, the system can effectively refine the monitoring scope to specific physical rooms, achieving zoned monitoring, precise triggering, and strict privacy protection. This multi-dimensional identification range solution not only improves monitoring efficiency but also enhances the system's intelligence and user trust.

[0082] Furthermore, in this embodiment, the system implements a time-of-occurrence identification scheme, endowing the monitoring system with time context awareness capabilities. This enables the system to automatically adjust monitoring strategies and alarm logic based on the risk differences at different times. Its core objective is to improve the accuracy and intelligence of monitoring, while ensuring that anomalies are detected earlier and appropriate alarm measures are taken, especially at night.

[0083] The system has at least two preset working modes to adapt to the different activity characteristics of daytime and nighttime: Daytime Activity Mode: Typically set between 7 AM and 10 PM. During this period, human activity is more frequent, and the system uses standard CSI detection thresholds, focusing on running a crowd classification algorithm. This algorithm effectively filters out large-scale activities by non-elderly individuals, such as children playing or adults exercising, thereby reducing false alarms.

[0084] Nighttime Quiet Mode: From 10 PM to 7 AM the following day. In this mode, the environment is typically quiet, and the background CSI signal is relatively stable. To adapt to the characteristics of nighttime, the system will activate a series of special strategies: Furthermore, the system can automatically lower the threshold for detecting abnormal fluctuations in CSI, allowing even minor nighttime activities by the elderly, such as turning over or getting up at night, to be detected. This adjustment helps to identify potential abnormalities earlier.

[0085] In addition, if the system detects an elderly person falling in a high-risk area such as the "bathroom" in night mode, it will automatically raise the initial risk rating of the event.

[0086] In night mode, once a fall is classified as "serious" or "extremely serious," the system will skip weaker alerts such as SMS messages and directly execute the highest priority action of "automatically dialing the guardian's phone." This measure ensures that even if the guardian is asleep, they can be woken up in time and rescue action can be taken.

[0087] To achieve more personalized and accurate monitoring, the system employs an adaptive learning scheme based on users' daily routines. By analyzing the activity and stability cycles of CSI signals throughout the day, the system automatically constructs a personal daily routine model for each user. This model is dynamically updated, reflecting typical sleep onset times, wake-up times, and nap habits. Based on this model, the system can more intelligently switch between "active" and "resting" modes, without relying on fixed time windows. This adaptive capability allows the system to better adapt to changes in users' lifestyles, providing more personalized monitoring services.

[0088] This system also supports integration with other smart devices in the home, such as smart lighting, smart curtains, and smart speakers. This integration allows the system's mode switching to be triggered by scene conditions, rather than solely by time. For example, when a user triggers "sleep mode" via voice command or the app (such as turning off the lights or drawing the curtains), the smart home hub simultaneously sends a command to the system, causing it to immediately switch to "nighttime rest mode." Conversely, when a user triggers "wake-up mode," the system automatically switches back to "daytime active mode." This design ensures that the system's state transitions are completely synchronized with the user's actual intentions and the environmental conditions, further enhancing the system's usability and user satisfaction.

[0089] Through the above solutions and their extensions, the system can effectively adjust monitoring and alarm strategies according to the characteristics of different time periods, while taking into account privacy protection and users' living habits, providing the elderly with comprehensive and intelligent security.

[0090] To facilitate understanding of the technical solutions of the embodiments of this application, the method for monitoring falls in the embodiments of this application will be illustrated below through a specific example.

[0091] Figure 2 The diagram illustrates a scenario of an elderly person falling (the red circle has a radius of 5cm, the yellow circle has a radius of 20cm, and the green area represents the safety zone). The method described in this embodiment utilizes all-optical networking (FTTR) technology and computer vision technology to achieve accurate monitoring and real-time alerts for elderly person falls. The specific steps are as follows: (1) Environment modeling and equipment initialization.

[0092] A depth camera is used to scan the indoor environment, constructing a 3D point cloud map containing (X, Y, Z) coordinate information. Using a 3D instance segmentation algorithm (such as PointGroup), individual furniture objects (such as tables, sofas, cabinets, etc.) are identified, and their sharp edges and corners are marked as danger zones, providing a foundation for subsequent spatial dimension analysis.

[0093] Simultaneously, during FTTR network initialization, each sub-router device is named and its physical location is specified via the accompanying app, establishing a "physical room - network device ID" mapping table. Smart surveillance cameras are connected to the FTTR device via WiFi network to obtain channel status data between the FTTR device and the smart camera, enabling stable and high-speed real-time reporting and alarms.

[0094] (2) Monitoring and preliminary screening of human activity.

[0095] The system monitors Channel Status Information (CSI) in real time using wireless communication devices. When CSI fluctuations exceed a preset threshold, abnormal activity is identified. The system locates the physical room where the abnormal event occurred based on the source of the CSI data packets (sub-router ID) and retrieves video data from the cameras associated with that area.

[0096] (3) Human posture recognition and fall confirmation.

[0097] Human posture recognition is performed on the acquired video data to extract key points of the torso. If the key points of the torso remain on the ground for more than a first preset time (e.g., 5 seconds), it is confirmed as a fall event.

[0098] (4) Time dimension analysis.

[0099] The total duration of the elderly person's body remaining stationary on the ground is continuously calculated. Based on preset time thresholds, fall events are classified into different levels: when 5 seconds < time between the center points of the hips < 15 seconds, the severity level is determined to be mild; when 15 seconds < time between the center points of the hips < 1 minute, the severity level is determined to be severe; and when the time between the center points of the hips > 1 minute, the severity level is determined to be extremely severe.

[0100] (5) Spatial dimension analysis.

[0101] The scenario of the elderly person's fall is further refined. First, it's determined whether the video footage of the fall originated from the living room, bedroom, kitchen, or bathroom. If the video footage is from the living room, furniture prone to sharp impacts (such as tables, sofas, and cabinets) is extracted from the background to obtain depth map data for the tables. Finally, the distance between the person's fall location and the edge of the table is calculated based on the depth map data.

[0102] The specific calculation steps are as follows: First, in the Kinect's depth camera coordinate system, the 3D coordinates of one of the table corners are obtained (a x ,a y ,a z The three-dimensional coordinates of the center point of the human head are (b x ,b y ,b z By querying the functions and methods of the Kinect SDK, the ground equation can be obtained as follows: Ax + By + Cz + D = 0 The constant term represents the distance from the center point of the Kinect's depth camera to the ground.

[0103] (a) x ,a y ,a z ) and (b x ,b y ,b z The projected coordinates of ) on the ground equation are respectively (a x1 ,a y1 ,a z1 ) and (b x1 ,b y1 ,b z1 (a x1 ,a y1 ,a z1 ) and (b x1 ,b y1 ,b z1The length between the corner of a table and the center point of a person's head is the same as the distance between the corner of a table and the center point of a person's head. The distance between the position where a person falls and the sofa, cabinet, etc. in the living room can be calculated in the same way.

[0104] As shown in Figure 2, when the distance between the human head and the furniture is less than 5 cm, the severity level is judged as particularly severe in the spatial dimension; when the distance between the human head and the furniture is less than 20 cm, the severity level is judged as severe in the spatial dimension; and when the distance between the human head and the furniture is greater than 20 cm, the severity level is judged as mild in the spatial dimension.

[0105] (6) Comprehensive assessment and graded alarm.

[0106] By combining time and spatial parameters, a pre-defined mapping table is used to determine the final severity level of the fall event. A tiered alert strategy is then executed based on this level. Minor: Sending text messages and surveillance videos to the guardian; Serious: Send an alarm message and surveillance video to the guardian; In particularly serious cases: Call the guardian to report the incident and send the surveillance video.

[0107] The specific details are categorized and presented in Table 1 below.

[0108] Table 1

[0109] The technical effects achievable by the embodiments of this application include: 1) This solution combines two dimensions, time and space, to comprehensively analyze the severity of falls among the elderly. By calculating in detail the duration of the elderly person's body remaining still after a fall (time dimension) and the precise distance between the elderly person's head and the danger zone of surrounding furniture (spatial dimension), the assessment of fall events becomes more comprehensive and detailed, the data support becomes more robust and reliable, and a scientific basis is provided for taking appropriate rescue measures.

[0110] 2) This solution utilizes computer vision technology to deeply mine background information from video images and accurately extract the position and shape features of furniture within the scene. With the help of distance calculation algorithms, the system can quickly determine the relative positional relationship between the elderly person and nearby sharp or hard objects when they fall. This in-depth analysis based on spatial scenes greatly enhances the ability to perceive potential risks of falls in the elderly, making safety monitoring more intelligent and human-centered.

[0111] 3) The network infrastructure built using all-optical networking technology lays the foundation for efficient and stable data transmission and processing. The integrated CSI acquisition plugin can capture subtle changes in channel status in real time. Utilizing the edge computing capabilities of the FTTR device, the system can instantly filter out potential anomalies and accurately locate the area where a fall event occurred, greatly reducing unnecessary data processing and computing resource consumption, and improving the system's response speed and operating efficiency.

[0112] 4) The smart camera, as the core sensing device, seamlessly connects to the FTTR network via WiFi, ensuring stable transmission and real-time sharing of monitoring data. With the help of accompanying intelligent analysis software, the camera can not only accurately capture the moment an elderly person falls, but also trigger corresponding alarm mechanisms based on preset severity levels. This process effectively reduces the false alarm rate, ensures the accuracy and timeliness of alarm information, and enables caregivers to obtain timely and accurate information about the elderly person's condition, buying valuable time for emergency rescue. It also avoids caregiver fatigue caused by frequent false alarms.

[0113] The fall monitoring device provided in the embodiments of this application will be described below. The fall monitoring device described below can be referred to in correspondence with the fall monitoring method described above.

[0114] This application provides a device for monitoring falls, see [link to relevant documentation]. Figure 3 ,include: The parameter acquisition module 310 is used to acquire time dimension parameters and spatial dimension parameters related to the fall event when it is determined that the monitored object has fallen. The time dimension parameters are used to characterize the continuous state of the monitored object after the fall, and the spatial dimension parameters are used to characterize the spatial positional relationship between the monitored object and the surrounding environmental objects when the fall occurs. The severity determination module 320 is used to determine the severity level of the fall event based on the time dimension parameter and the space dimension parameter.

[0115] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4As shown, the electronic device may include a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, communications interface 420, and memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a method for monitoring falls. This method includes: upon determining that a fall event has occurred on the monitored object, acquiring time-dimensional parameters and spatial-dimensional parameters related to the fall event, wherein the time-dimensional parameters characterize the continuous state of the monitored object after the fall, and the spatial-dimensional parameters characterize the spatial positional relationship between the monitored object and surrounding environmental objects at the time of the fall; and determining the severity level of the fall event based on the time-dimensional parameters and spatial-dimensional parameters.

[0116] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0117] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the fall monitoring method provided by the above methods. The method includes: when it is determined that a fall event has occurred on the monitored object, acquiring time dimension parameters and spatial dimension parameters related to the fall event, wherein the time dimension parameters are used to characterize the continuous state of the monitored object after the fall, and the spatial dimension parameters are used to characterize the spatial positional relationship between the monitored object and surrounding environmental objects when the fall occurs; and determining the severity level of the fall event based on the time dimension parameters and spatial dimension parameters.

[0118] In another aspect, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for monitoring falls provided by the methods described above. This method includes: upon determining that a fall event has occurred on the monitored object, acquiring time-dimensional parameters and spatial-dimensional parameters related to the fall event, wherein the time-dimensional parameters characterize the continuous state of the monitored object after the fall, and the spatial-dimensional parameters characterize the spatial positional relationship between the monitored object and surrounding environmental objects at the time of the fall; and determining the severity level of the fall event based on the time-dimensional parameters and spatial-dimensional parameters.

[0119] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0120] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for monitoring falls, characterized in that, include: When it is determined that the monitored object has fallen, time dimension parameters and spatial dimension parameters related to the fall event are obtained. The time dimension parameters are used to characterize the continuous state of the monitored object after the fall, and the spatial dimension parameters are used to characterize the spatial positional relationship between the monitored object and the surrounding environmental objects when the fall occurs. The severity level of the fall event is determined based on the time dimension parameters and the spatial dimension parameters.

2. The method for monitoring falls according to claim 1, characterized in that, Also includes: Based on monitoring data acquired from at least one sensor, it is determined whether a fall has occurred on the monitored object.

3. The method for monitoring falls according to claim 2, characterized in that, The sensor includes a wireless communication device and a video camera; Based on monitoring data acquired from at least one sensor, determine whether a fall has occurred on the monitored object, specifically including: The wireless communication device monitors the channel status information of the wireless signal in real time, and determines that abnormal activity has occurred when the fluctuation amplitude of the channel status information exceeds a first preset threshold. In response to determining that the abnormal activity has occurred, video data is acquired from the video camera, the video data being data from a preset time period prior to the time when the abnormal activity occurred; The video data is subjected to human posture recognition, and when the key points of the torso of the monitored object remain on the ground for more than a first preset time, it is determined that the monitored object has fallen.

4. The method for monitoring falls according to claim 1, characterized in that, The sensor includes a depth camera; The steps for obtaining the spatial dimension parameters related to the fall event specifically include: The environment in which the monitored object is located is scanned by the depth camera to establish a three-dimensional point cloud model of the scene; Furniture objects are segmented and identified in the three-dimensional point cloud model, and geometric analysis is performed on the model of the furniture objects to determine their sharp edges or sharp corners as danger zones. Calculate the three-dimensional spatial distance between the head of the monitored object and the danger zone, and use the three-dimensional spatial distance as the spatial dimension parameter.

5. The method for monitoring falls according to claim 1, characterized in that, The steps for obtaining the time dimension parameters related to the fall event specifically include: After a fall event is confirmed, the total duration for which the monitored object's body remains stationary on the ground is continuously calculated, and this total duration is used as the time dimension parameter.

6. The method for monitoring falls according to claim 1, characterized in that, Based on the aforementioned time and space parameters, the severity level of the fall event is determined, specifically including: The spatial dimension parameter is compared with at least one spatial threshold to determine the severity of the spatial dimension; The time dimension parameter is compared with at least one time threshold to determine the severity of the time dimension; and The final severity level of the fall event is determined based on the severity of the spatial dimension and the severity of the temporal dimension.

7. The method for monitoring falls according to claim 1 or 6, characterized in that, After determining the final severity level of the fall event, the method further includes: executing a corresponding graded alarm step based on the determined severity level; The hierarchical alarm step includes: When the severity level is low risk, a text message reminder is sent to the designated guardian. When the severity level is medium risk, send an in-app alert to the guardian. When the severity level is high-risk, the alarm call to the guardian will be automatically dialed.

8. The method for monitoring falls according to claim 3, characterized in that, After acquiring video data from the video camera, the method further includes: applying a crowd category recognition algorithm to the human images in the video data to obtain category recognition results; If the category identification result matches the preset category of the monitoring object, the step of performing human posture recognition on the video data continues.

9. A device for monitoring falls, characterized in that, include: The parameter acquisition module is used to acquire time dimension parameters and spatial dimension parameters related to the fall event when it is determined that the monitored object has fallen. The time dimension parameters are used to characterize the continuous state of the monitored object after the fall, and the spatial dimension parameters are used to characterize the spatial positional relationship between the monitored object and the surrounding environmental objects when the fall occurs. The severity determination module is used to determine the severity level of the fall event based on the time dimension parameters and the spatial dimension parameters.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the method for monitoring falls as described in any one of claims 1 to 8.

11. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method for monitoring falls as described in any one of claims 1 to 8.

Images