Bed Sensor System

Abstract

The present invention may include a bed sensor system that checks the physical condition of a facility user and prevents fall accidents of said facility user.

Description

Technology Field

[0001] The present invention relates to a bed sensor system. Background Technology

[0002] In modern hospitals and nursing facilities, real-time monitoring of the safety and health of users and the prevention of falls are recognized as important tasks.

[0003] Previously, reliance on periodic visual observation or simple sensor devices to check the condition of facility users was common; however, this method had limitations, such as difficulty in real-time response and inability to immediately detect dangerous situations like changes in individual user conditions, violent movements, or falling out of bed.

[0004] Meanwhile, the aforementioned background technology is technical information that the inventor possessed for the derivation of the present invention or acquired during the process of deriving the present invention, and it cannot necessarily be considered publicly known technology disclosed to the general public prior to the filing of the present invention. Prior art literature

[0005] Korean Registered Patent No. 101990118 The problem to be solved

[0006] The objective of the present invention is to provide a bed sensor system that can accurately check the physical condition of a facility user and prevent accidents such as falls.

[0007] The technical problems of the present invention are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0008] A bed sensor system according to one embodiment of the present invention may include a management server that checks the physical condition of a facility user and prevents fall accidents of said facility user.

[0009] According to one embodiment, the management server may include: an escape alarm unit that measures and collects in real time whether the bed frame is vibrating and the vibration intensity through a vibration sensor integrated into the bed frame, and generates an alarm when a facility user leaves or a specific movement occurs; a weight measurement unit that measures and collects in real time the accurate weight of a facility user through a weight sensor integrated into the bed frame, and classifies, accumulates, and records the collected weight data for each facility user to manage; a device linkage unit that links with at least one device that is provided in a facility including a hospital and is compatible with the facility, and collects data input for each facility user from the linked device to perform life monitoring for each facility user; a report unit that collects and manages data collected through the vibration sensor, the weight sensor, and the life monitoring, as well as data for at least one facility user, and generates a report in a pre-set format at a predetermined period; and a monitoring unit that outputs the health status of all facility users and the situation of facility users, including emergency response, through data collected through the sensors to a display screen linked by the device linkage unit.

[0010] According to one embodiment, the escape alarm unit may generate an alarm when the vibration intensity collected through the vibration sensor exceeds a preset threshold, by determining that the user of the facility has escaped from the bed or has made vigorous movements.

[0011] According to one embodiment, the preset threshold value may be a value entered as an initial setting by an administrator or operator, and may mean a value smaller than the intensity collected through the vibration sensor when the bed is disengaged or vigorous movement occurs.

[0012] According to one embodiment, the device linkage unit can collect updated data of a facility user in the case of a facility user for whom updated data exists from linked equipment.

[0013] According to one embodiment, the report unit may derive current status analysis information of a facility user by utilizing an AI algorithm that learns and analyzes the updated data, and provide the current status analysis information of a facility user derived by utilizing the AI ​​algorithm to the monitoring unit.

[0014] According to one embodiment, the updated data may refer to vital data including weight collected through a facility user's weight sensor and pulse collected through regular checkups.

[0015] According to one embodiment, the AI ​​algorithm can analyze the condition of a facility user by applying at least one of an anomaly detection algorithm, a time series prediction algorithm, and a pattern classification algorithm based on vital data and environmental data collected through the vibration sensor, the weight sensor, and the life monitoring.

[0016] According to one embodiment, the AI ​​algorithm may be characterized by being configured to apply pre-trained data and real-time data for each facility user in parallel to analyze a state prediction customized for the facility user, automatically adjust a threshold value customized for each facility user based on the analysis result of the state prediction customized for the facility user, and transmit to the monitoring unit and the departure alarm unit when a risk sign is predicted for the facility user based on the analysis result of the state prediction customized for the facility user.

[0017] According to one embodiment, the weight measuring unit operates as an internal processing module and utilizes an alarm judgment algorithm that refines the signal received from the weight sensor in real time and calculates the final weight value to be transmitted to the report unit and the monitoring unit, thereby increasing the reliability of the real-time measurement value obtained from the bed-integrated weight sensor and reducing false positives (errors) caused by vibration, temperature, and temporary noise, so as to accurately estimate the actual weight of the facility user.

[0018] According to one embodiment, the alarm determination algorithm may include a first item that converts the measurement value of the weight sensor into a kg unit and is designed to serve as the basis for sensor calibration by utilizing a zero offset (C0) to convert the raw sensor value into physical weight.

[0019] According to one embodiment, the alarm determination algorithm may include a second item designed to correct distortions caused by dynamic movement and ambient temperature together, thereby continuously correcting errors caused by vibration and temperature.

[0020] According to one embodiment, the alarm determination algorithm is designed to reduce short-term noise and stably handle rapid numerical changes by weighted averaging the value measured from the actual sensor and the current correction value, and may include a third item that mixes the past value and the current correction value and limits it to a physical limit. Effects of the invention

[0021] According to one aspect of the present invention described above, the bed sensor system proposed by the present invention measures the presence and intensity of vibration of the bed frame in real time through a vibration sensor integrally mounted on the bed frame, and if the vibration exceeds a preset threshold, it determines that the user of the facility has left the bed or made a violent movement, and can immediately generate an alarm.

[0022] In addition, the weight of facility users can be accurately measured through weight sensors installed on the bed frame, and long-term health can be monitored by accumulating and recording weight data for each user.

[0023] In addition, it connects to compatible equipment within hospitals or facilities to collect vital data and lifestyle monitoring data input from the connected equipment, and based on this, it can analyze the living status of each facility user.

[0024] In addition, collected vibration, weight, vital signs, and environmental data are applied to AI algorithms to perform analyses such as anomaly detection, time series forecasting, and pattern classification, and customized condition predictions for facility users can be provided by applying pre-trained data and real-time data in parallel.

[0025] The effects of the present invention are not limited to those mentioned above, and various effects may be included within the scope obvious to a person skilled in the art from the contents described below. Brief explanation of the drawing

[0026] FIG. 1 is a conceptual diagram of a bed sensor system according to one embodiment of the present invention. FIG. 2 is a conceptual diagram of a management server according to one embodiment of the present invention. Specific details for implementing the invention

[0027] The following detailed description of the invention refers to the accompanying drawings, which illustrate specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that various embodiments of the invention are different but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in relation to one embodiment.

[0028] When it is stated that one component is "connected" or "contracted" to another component, it should be understood that while it may be directly connected or contracted to that other component, there may also be other components in between. Conversely, when it is stated that one component is "directly connected" or "directly contracted" to another component, it should be understood that there are no other components in between.

[0029] Furthermore, it should be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not intended to be taken in a limiting sense, and the scope of the invention is limited only by the appended claims, including all equivalents thereof, provided appropriately described. Similar reference numerals in the drawings refer to the same or similar functions across various aspects.

[0030] Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the drawings.

[0032] FIG. 1 is a conceptual diagram of a bed sensor system according to one embodiment of the present invention.

[0033] Referring to FIG. 1, a bed sensor system according to one embodiment of the present invention may include a management server (100), a vibration sensor (300), and a weight sensor (500).

[0034] The management server (100) can measure the vibration of the bed frame and the vibration intensity in real time through the vibration sensor (300) according to the present invention, and if the vibration exceeds a preset threshold, it can determine that the facility user has stepped off the bed or made violent movements and immediately generate an alarm.

[0035] In addition, the management server (100) can accurately measure the weight of the facility user through the weight sensor (500) according to the present invention, and manage the health of the facility user in the long term by accumulating and recording weight data for each facility user.

[0036] In addition, the management server may be characterized by additionally collecting respiration and heart rate data including a non-contact radar sensor or an optical PPG sensor, analyzing sleep stages based on body weight, vibration, and vital data, and accordingly detecting abnormal signs early.

[0037] The management server (100) may be a self-contained server or a cloud server for providing the service according to the present invention, or it may be a p2p (peer-to-peer) set of distributed nodes.

[0038] The management server (100) can perform one or more of the operations, storage, reference, input / output, and control functions of a general computer, and may include an artificial neural network described later based on input data.

[0039] The management server (100) may include a processor and memory. The processor may include devices capable of preventing safety accidents in real time and managing the health status of facility users by immediately transmitting a notification through a monitoring unit and an escape alarm unit when a risk sign according to the present invention is predicted. The processor may execute a program or control the management server (100). Program code executed by the processor may be stored in memory. The memory may store relevant information for performing a service according to the present invention or a program for implementing a method. The memory may be volatile memory or non-volatile memory.

[0040] The management server (100) can send data to an external device or receive data from an external device using a network.

[0041] The management server (100) can train an artificial neural network and can also use an artificial neural network that has been trained. The processor can train or execute an artificial neural network stored in memory, and the memory can store an artificial neural network that has been trained. The electronic device that trains the artificial neural network and the electronic device that uses it may be the same, but they may also be separate.

[0042] Artificial intelligence is a computer system that partially implements the functions of the human brain and is capable of learning, speculating, and making judgments on its own. As learning progresses, the probability of extracting the correct answer can increase. Artificial intelligence can be composed of learning and component technologies that utilize it. The learning aspect of AI is an algorithmic technology that classifies and learns features based on input data, while the component technologies may be techniques that utilize these learning algorithms to partially implement the functions of the human brain.

[0043] Artificial intelligence is a technology that facilitates the approach to problems where multiple probabilistic answers are possible, enabling it to logically and probabilistically infer optimal cycles, methods, and plans based on input data. AI inference techniques can include evaluating input data, optimization prediction, knowledge and probability-based reasoning, and preference-based planning.

[0044] Artificial neural networks are learning algorithms in the field of machine learning that programmatically implement the connections between neurons and synapses in the brain. By creating a neural network structure through programming and then training it, artificial neural networks can acquire desired functions. Although errors may exist, they can learn from massive datasets to produce appropriate output data from input data. They have the advantage of being able to obtain output data that has yielded statistically good results and are similar to human reasoning.

[0045] The management server (100) can infer individual characteristics and interests by analyzing consumers' online behavior data, social media activities, search history, etc., using an artificial intelligence algorithm built based on big data, and may include a number of pre-trained artificial neural networks for this purpose.

[0046] The network is a high-speed backbone network of a large-scale communication network capable of high-capacity, long-distance voice and data services, and may be a next-generation wired and wireless network for providing the Internet or high-speed multimedia services.

[0047] If the network is a mobile communication network, it may be a synchronous mobile communication network or an asynchronous mobile communication network. As an example of an asynchronous mobile communication network, a WCDMA (Wideband Code Division Multiple Access) network may be cited. In this case, although not shown in the drawing, the network may include an RNC (Radio Network Controller). Meanwhile, although a WCDMA network was given as an example, it may be a 3G LTE network, a 4G network, a next-generation communication network such as 5G, or other IP-based IP networks.

[0048] The management server (100) can include any terminal capable of exchanging data over a network, such as a desktop computer, laptop, tablet, or smartphone.

[0049] The management server (100) may include one or more of the computation function, storage function, reference function, input / output function, and control function that a computer has in order to perform the service according to the present invention.

[0050] The management server (100) may access a website or install an application to receive the service according to the present invention. The management server (100) may exchange data through the website or the application.

[0051] The network is a high-speed backbone network of a large-scale communication network capable of high-capacity, long-distance voice and data services, and may be a next-generation wired and wireless network for providing the Internet or high-speed multimedia services.

[0052] If the network is a mobile communication network, it may be a synchronous mobile communication network or an asynchronous mobile communication network. As an example of an asynchronous mobile communication network, a WCDMA (Wideband Code Division Multiple Access) network may be cited. In this case, although not shown in the drawing, the network (300) may include an RNC (Radio Network Controller). Meanwhile, although a WCDMA network was given as an example, it may be a 3G LTE network, a 4G network, a 5G network, or other next-generation communication networks, or other IP-based IP networks.

[0053] A system (1) according to one embodiment of the present invention can overcome the limitations of simple observation or existing sensor devices and provide technical effects that enable customized monitoring and safety management for each facility user.

[0055] FIG. 2 is a conceptual diagram of a management server according to one embodiment of the present invention.

[0056] Referring to FIG. 2, a management server (100) according to one embodiment of the present invention may include an escape alarm unit (110), a weight measurement unit (130), a device linkage unit (150), a report unit (170), and a monitoring unit (190).

[0057] The escape alarm unit (110) measures and collects the vibration status and vibration intensity of the bed frame in real time through a vibration sensor (300) that is integrated into the bed frame, and can generate an alarm when a facility user escapes or a specific movement occurs.

[0058] Additionally, the escape alarm unit may generate an alarm if the vibration intensity collected through the vibration sensor (300) exceeds a preset threshold, determining that the user of the facility has escaped from the bed or has made vigorous movements.

[0059] Here, the preset threshold value is a value entered as an initial setting by an administrator or operator, and may mean a value that is smaller than the intensity collected through the vibration sensor (300) when the bed is deviated from or vigorous movement occurs.

[0060] The weight measuring unit (130) can measure and collect the accurate weight of a facility user in real time through a weight sensor (500) that is integrated into the bed frame, and can manage the collected weight data by classifying, accumulating, and recording it for each facility user.

[0061] Meanwhile, the weight measurement unit (130) can utilize an alarm judgment algorithm to accurately estimate the actual weight of the facility user by increasing the reliability of the real-time measurement value obtained from the bed-integrated weight sensor and reducing false positives (errors) caused by vibration, temperature, and temporary noise.

[0062] The alarm judgment algorithm can operate as an internal processing module of the weight measurement unit (130) to refine the signal coming from the weight sensor in real time and calculate the final weight value to be transmitted to the report unit (170) and the monitoring unit (190).

[0063] The alarm judgment algorithm may be used not only for outputting values ​​to a simple display, but also as a judgment and control means that provides input values ​​used in the judgment logic of the departure alarm unit (110) (e.g., judgment of sudden weight loss).

[0064] In terms of technical effects, compared to the existing simple W = C·L method, instantaneous errors caused by environmental and operational noise can be reduced, which lowers the false alarm rate, and if properly tuned through field calibration and test data, a substantial reduction in measurement RMSE (error) can be expected.

[0065] The alarm judgment algorithm may consist of a first item that converts raw sensor values ​​into physical weight, a second item that continuously corrects errors caused by vibration and temperature, and a third item that mixes past values ​​and current corrected values ​​and limits them to physical limits.

[0066] Specifically, the first item may be characterized by being designed to serve as the basis for sensor calibration by converting the measurement value of the weight sensor into kg units and correcting it using a zero offset (C0).

[0067] In other words, the alarm judgment algorithm uses the first item because weight sensors usually produce values ​​proportionally like a straight line, so a linear scaling factor (C) to the raw value aj By multiplying by ) and adding the zero offset (C0), you can convert it into actual body weight units (kg).

[0068] In this way, the alarm judgment algorithm can fundamentally convert the value measured by the weight sensor and accurately predict even if the value measured by the weight sensor deviates slightly.

[0069] The second item may be characterized by being designed to correct distortion caused by dynamic movement and ambient temperature together for vibration and temperature correction.

[0070] In other words, the alarm judgment algorithm uses the second item to prepare for situations where the weight value may momentarily spike if a person moves or vibrations occur in the surroundings, by controlling the vibration magnitude (V rms ) threshold(V thr It can be normalized to a value between 0 and 1 by dividing by ).

[0071] The alarm judgment algorithm can use the trigonometric function sine to smoothly adjust the normalized value, so that it makes almost no correction when the vibration is small and gradually reduces it as the vibration increases, thereby preventing unnecessary errors.

[0072] Here, the vibration correction weight (α) is used as a value to adjust the magnitude of the correction strength. aj You can use ).

[0073] In addition, the alarm judgment algorithm can adjust for changes based on temperature through the second item.

[0074] Weight sensors may have slight errors depending on the temperature. Since there is no need to react strongly to small temperature changes and not to react excessively to large temperature changes, a logarithmic function can be used so that the correction increases only slightly as the temperature difference increases.

[0075] At this time, the temperature correction weight (γ) is the value that determines the sensitivity of the correction. aj You can use ).

[0076] The third item may be characterized by being designed to reduce short-term noise and stably handle rapid numerical changes by weighted averaging the value measured from the actual sensor and the current correction value.

[0077] Since measured values ​​often spike momentarily, the current estimated value (W corr ) and the historically stable value, the historically estimated weight (W hist Mix ) to get the final weight (W est ) can be derived, at which point the current estimated value (W corr ) and past estimated weight (W hist The value that determines the ratio of ) is the weight-adjusted weight (β aj ) am.

[0078] Final weight (W est ) is the weight-adjusted weight (β ajIf ) is large, it relies more on past values, making it stable but insensitive to changes, and weight-adjusted weights (β aj If ) is small, it can reflect the current value more and react sensitively.

[0079] Finally, the alarm judgment algorithm can output the calculated weight in a range of 0 to 300 kg so that the calculated weight does not come out as an absurd value (e.g., -10 kg, 500 kg).

[0080] The alarm judgment algorithm provides a raw output value (L) to provide the basis for safety and alarm judgment. aj ), linear scaling coefficient (C aj ), zero offset (C0), vibration magnitude (V rms ), vibration threshold (V thr ), vibration correction weight (α aj ), sensor ambient temperature (T aj ), temperature correction weight (γ aj ), past estimated weight (W hist ), weight-adjusted weight (β aj Based on ), the final value (W est ) can be extracted.

[0081] The alarm judgment algorithm is an essential element that serves as the starting point for all values ​​finally calculated by the weight measurement unit (130) as source data for weight estimation, and is a raw output value (L aj You can utilize ).

[0082] Raw output value (L aj ) may mean the raw output value of the weight sensor already equipped in the bed, which is the value obtained by converting the analog voltage into an ADC or the real-time measurement value read from the system as an ADC count.

[0083] Raw output value (L aj The unit of ) can be expressed as ADC count or voltage (V) depending on the implementation, and the operating documentation must specify which unit is used.

[0084] Raw output value (Laj This can be done by sampling the analog signal of the weight sensor with an ADC, reading the sample value with a system API, and storing it in a buffer along with a timestamp.

[0085] The alarm determination algorithm uses a linear scaling coefficient (C) to accurately map the unit and scale of the weight sensor output to physical weight. aj You can utilize ).

[0086] Linear scaling factor (C aj ) is the raw output value (L aj It can have units of kg / ADC_count or kg / V as a linear scale factor for converting ) into physical mass units (kg).

[0087] Linear scaling factor (C aj ) sequentially places standard weights (e.g., 0 kg, 10 kg, 30 kg, 50 kg, etc.) and the raw output value (L) corresponding to each weight aj It can be derived by recording the value and calculating it using linear regression or two-point correction.

[0088] The alarm judgment algorithm can utilize a zero offset (C0) to improve long-term reliability by correcting zero drift that occurs during long-term operation.

[0089] That is, the zero offset (C0) is the zero offset of the weight sensor and can be expressed in kg units, and may mean correcting the deviation of the measurement value under no load (0 kg).

[0090] The zero offset (C0) can be derived from the initial setting value measured in a no-load state when the weight sensor is installed.

[0091] Since weight sensor signals can be temporarily distorted due to user movement or external impact, the alarm judgment algorithm applies dynamic correction using vibration indicators to reduce false positives, specifically by applying vibration magnitude (V rms You can utilize ).

[0092] Vibration magnitude (V rms ) may refer to the RMS value of vibration calculated as the square root of the mean square of samples acquired from a vibration sensor during a certain window (e.g., 0.5 to 1 second), and m / s, the measurement unit of the vibration sensor. 2 It can be used as a normalized dimensionless value.

[0093] Vibration magnitude (V rms ) can be derived from a value calculated by averaging multiple data from the vibration sensor, and the window length can be derived by considering sensor noise and responsiveness requirements.

[0094] The alarm judgment algorithm uses a vibration threshold (V), which is a reference value to prevent unnecessary corrections by separating normal noise levels from actual meaningful movement. thr You can utilize ).

[0095] Vibration threshold (V thr ) is a threshold value for meaningfully determining vibration, and can be derived using the mean and standard deviation of vibration samples measured in the static state of the installation environment.

[0096] Vibration threshold (V thr The unit of ) is vibration magnitude (V rms m / s, the same as ) 2 It can be used as a normalized dimensionless value.

[0097] The alarm judgment algorithm flexibly adjusts the correction strength according to the intensity of vibration, providing stronger correction when vibration is high and minimizing correction when vibration is low to prevent distortion caused by excessive correction, using vibration correction weights (α aj You can utilize ).

[0098] Vibration correction weight (α aj ) is a dimensionless weight representing the sensitivity of vibration correction and can take a value between 0 and 1, and the vibration correction weight (α aj) can be derived based on data extracted through a past alarm judgment algorithm.

[0099] Since the weight sensor and electronic components are sensitive to temperature and abnormal measurement phenomena may occur due to temperature changes, the alarm determination algorithm ensures long-term stability through temperature compensation based on the sensor ambient temperature (T aj You can utilize ).

[0100] Sensor ambient temperature (T aj ) is the temperature (°C) around the weight sensor, which can be derived from values ​​collected in real time through a temperature sensor or facility environment sensor connected to the system.

[0101] Sensor ambient temperature (T aj ) can be derived by sampling and recording the analog or digital output of the temperature sensor at regular intervals (e.g., every minute).

[0102] Sensor ambient temperature (T aj ) is based on a standard indoor temperature (e.g., 20°C), but the initial setting can be changed according to the long-term average temperature of the installation environment, and the standard indoor temperature can be recalibrated using operational data.

[0103] The alarm judgment algorithm uses a temperature correction weight (γ) to enable stable correction through a log-based scale, so that the correction effect is appropriately amplified when temperature changes are large and has almost no effect on small changes. aj You can utilize ).

[0104] Temperature correction weight (γ aj ) is a dimensionless coefficient representing the sensitivity of temperature correction, and can mean a value for determining the magnitude of the effect of temperature change on the measurement.

[0105] Temperature correction weight (γ aj) may refer to a value derived by performing regression analysis between the sensor error measured in various temperature ranges in a temperature chamber or field temperature change experiment and the ln-based temperature index.

[0106] The alarm judgment algorithm uses past estimated weight (W) to mitigate errors caused by instantaneous noise or transient vibrations and to reflect temporal trends in order to increase the reliability of the alarm trigger. hist You can utilize ).

[0107] Past estimated weight (W hist ) is a value representing a past stabilized weight value and may be derived by the EWMA (Exponential Moving Average) method or the median of the most recent n samples, and may be in kg units.

[0108] The alarm judgment algorithm uses weight correction weights (β) to prevent the generation of immediate alarms due to sporadic errors and to enable state reflection at a reasonable temporal resolution. aj You can utilize ).

[0109] Weight-adjusted weight (β aj ) is past estimated weight (W hist ) and current estimated value (W corr The weights used to mix ) have values ​​between 0 and 1, and the derivation method of β can be experimentally determined based on the trade-off between responsiveness (sensitivity to recent changes) and stability (noise mitigation).

[0110] Weight-adjusted weight (β aj Typically, a range of 0.2 to 0.4 is recommended. Since setting β large makes the results more reliant on past values ​​and thus more resistant to noise, and setting β small reflects current measurements more and improves responsiveness, the choice can be made according to the operating policy.

[0111] Vibration correction weight (α aj ) and weight-adjusted weights (β aj ) is the final value (W estIt can be set to a value directly entered by the server administrator or operator to adjust the impact on ).

[0112] For example, the server administrator or operator [is] the final value (W est If you want to increase the influence of the variable derived by the weight sensor when extracting ), weight correction weight (β aj ) vibration correction weight (α aj You can enter a value greater than ).

[0113] Here, the alarm judgment algorithm determines the final value (W est By setting an upper limit (300 kg) and a lower limit (0 kg) of ), physical limits can be set to prevent the output of unrealistic values.

[0114] These constants can be set with recommended initial values ​​but finalized through field calibration based on the actual operating environment and equipment characteristics, and can be transparently recorded through equipment-specific documentation and configuration history management.

[0115] First, the alarm judgment algorithm is the physical conversion logic of the sensor measurement system; since weight sensors generally have linear characteristics, a linear model enables practical correction with a minimum number of variables, so the weight sensor's raw output value (L aj It can be derived through the process of linearly converting ) to kg and subtracting the zero deviation.

[0116] Secondly, the alarm judgment algorithm determines the vibration based on the vibration magnitude (V rms The reason for using the trigonometric function sine to map to a smooth monotonic function to prevent excessive correction, after expressing it as ) and normalizing it to the range of 0 to 1 to provide a form that is gentle and monotonically increasing, is to keep the correction effect small when the vibration is small and to reflect that the correction gradually increases in the direction of saturation as the vibration becomes stronger.

[0117] Thirdly, the alarm judgment algorithm may be characterized by utilizing the compression effect of the logarithm to gradually increase the correction effect as the temperature difference increases, while having almost no effect on small temperature changes.

[0118] Fourth, the reason the alarm judgment algorithm ultimately uses a weighted average can be characterized as being configured to reduce distortion caused by outliers in time series data and maintain a stable trend while reflecting the current state.

[0119] Example of actual numerical input value of alarm judgment algorithm raw output value (L aj )= 12000 (ADC count), linear scaling factor(C aj ) = 0.0125 (kg / count), Zero offset(C0) = -0.5 (kg), Vibration magnitude(V rms ) = 0.8 (m / s²), vibration threshold (V thr ) = 2.0 (m / s²), vibration correction weight (α aj ) = 0.4, temperature correction weight (γ aj ) = 0.01, sensor ambient temperature(T aj ) = 25 (°C), past estimated weight (W hist ) = 150.0 (kg), weight-adjusted weight (β aj If ) = 0.3 is set, the final value (W est ) can be confirmed to be approximately 126.48 kg.

[0120] This example clearly demonstrates how each correction term changes the raw value—that is, how vibration and temperature corrections can decrease or increase the raw weight—and the actual magnitude and correction intensity are the vibration threshold (V thr ), vibration correction weight (α aj ), temperature correction weight (γ aj It may vary by site due to variables such as ).

[0121] Vibration threshold (V thr) can be derived by collecting sufficient samples in a idle state during installation and calculating the mean and standard deviation, and can be adjusted by the server administrator or operator according to the environmental noise level and requirement sensitivity.

[0122] The standard indoor temperature is generally set to 20°C as the default, but if the long-term average temperature of the installation environment varies significantly, it can be changed to the site average, and the standard indoor temperature can be recalibrated by the server administrator or operator through periodic analysis of operational data.

[0123] The upper and lower limits can be set based on physical limits, for example, 0 kg and 300 kg, and the system can be configured to automatically clamp and record a diagnostic log when extreme values ​​occur.

[0124] When making judgments by combining multiple conditions, not only AND / OR logic but also a weighted total score method can be applied; for example, "weight loss rate of 30% or more during the previous 1 second" AND "vibration magnitude (V rms When ) > 0.6", it can be determined as a possibility of deviation or fall, and this threshold can be readjusted by the server administrator or operator through periodic analysis of operational data.

[0125] The alarm determination algorithm provides a raw output value (L) for each sample time t. aj ), vibration magnitude (V rms ) and sensor ambient temperature (T aj ) can be collected and preprocessing and correction calculations can be performed in a time-synchronized state.

[0126] Vibration magnitude (V rms ) can continuously update the window-based RMS from high-frequency samples to reflect recent conditions, and the window length can be set in the range of 0.5 to 1 second depending on noise characteristics and responsiveness requirements.

[0127] Conditions requiring a temporal pattern, such as determining deviation or fall, can be implemented by comparing consecutive intervals, and for example, time-based logic can be set such that an alarm is triggered when the cumulative decrease rate exceeds X% as weight decreases continuously over a continuous 1-second interval.

[0128] The administrator can use the system interface to access the linear scaling factor (C aj ), zero offset (C0), vibration threshold (V thr ), vibration correction weight (α aj ), weight-adjusted weight (β aj ), temperature correction weight (γ aj Key variables such as standard indoor temperature, lower limit, and upper limit can be manually set or updated remotely, and such user input can be applied immediately or reflected after an approval process.

[0129] The server administrator or operator [uses] weight correction weights (β aj Setting the value of ) to a large value may cause the system to rely more on historical values, making it more resilient to noise and reducing sensitivity to sudden changes; conversely, the weight-adjusted weight (β aj Setting the value to a small size increases responsiveness by reflecting the current measurement more, but it may become vulnerable to transient errors.

[0130] Vibration correction weight (α aj Changes to ) directly adjust the correction strength for vibration, so appropriate ranges can be suggested and recommended values ​​distributed through field testing, and a function can be provided to roll back to the previous settings if the administrator input is incorrect.

[0131] If you enable the AI-based auto-adjustment option, the vibration correction weight (α aj ) and weight-adjusted weights (β aj) can periodically collect a labeled calibration dataset (correct values ​​measured by a manual scale and sensor logs) and learn the calibration data of the variable using a regression model (e.g., linear regression, ridge regression).

[0132] The AI ​​training process can be performed based on data splitting (training / validation / testing), hyperparameter exploration (grid search or random search), cross-validation, and performance metrics (e.g., RMSE, ROC-AUC), and the trained model applies weights online (oscillation-corrected weights (α aj ) and weight-adjusted weights (β aj ), temperature correction weight (γ aj You can periodically recommend or directly update ) etc., but before updating, you must always perform verification with a verification dataset and deploy in a rollback-capable manner.

[0133] The technical description and examples of the above alarm judgment algorithm are sufficiently provided so that a person skilled in the art can easily implement the present invention according to mathematical formulas and variable definitions, and it can be confirmed that the implementation of the present invention is described at a level that is obvious to a person skilled in the art.

[0134] The device linkage unit (150) is equipped in a facility including a hospital and is linked with at least one device that can be linked, and can perform life monitoring for each facility user by collecting data entered for each facility user from the linked device (e.g., heart rate measuring device, blood pressure measuring device, body temperature measuring device, exercise amount measuring device, bed sensor, caregiver terminal, medical person terminal, etc.).

[0135] In addition, the device linkage unit (150) can collect updated data from facility users in the case of facility users for whom updated data exists from linked equipment.

[0136] Meanwhile, the device linkage unit (150) automatically determines which of the devices among the devices to process data from first based on the elapsed time (ΔTuis ), vibration intensity (V uis ), weight (W uis ), health indicators (H uis ), User health indicators(S uis ), importance score (P uis ), rank threshold (T uis ), mixed weights (α uis A data processing algorithm that extracts a ranking score (UIS) from ) can be utilized.

[0137] To this end, the data processing algorithm may be characterized by being designed to quantify and comprehensively reflect the data's recency, abnormal signs detected by sensors, and importance set by the operator.

[0138] If the device linkage unit (150) uses a data processing algorithm, when strong vibration occurs in a patient's bed and a change in weight is detected, and the patient's condition is critical, the priority can be increased so that the patient's equipment data is immediately transmitted to the server and displayed on the monitoring screen.

[0139] While existing methods often simply process data in chronological order or treat all equipment equally, the device linkage unit (150) can be characterized by being designed to naturally lower the importance of old data through a data processing algorithm, reduce unnecessary data transmission by simultaneously considering sensor-based risk and external priority, and respond more quickly to emergency situations.

[0140] The data processing algorithm includes a first item indicating how recently the data of a specific piece of equipment was updated, a second item indicating whether there are issues with the facility user's condition and whether there is an equipment malfunction, and user health indicators (S uis ) and importance score (P uis It can be characterized by being composed of a third item utilizing ).

[0141] More specifically, the first item may be characterized by being designed so that the value is higher the shorter the data collection period, and the value is lower the older the data.

[0142] The data processing algorithm may be characterized by the use of a logarithmic function to reduce the importance of old data.

[0143] In other words, since older data may be unreliable, the data processing algorithm can maintain stability in calculations by using a logarithmic function that reduces the score as ΔT increases, thereby allowing the most recent data that just arrived to be processed with higher priority.

[0144] The logarithmic function is elapsed time (ΔT uis When ) is small, a large value is given, but elapsed time (ΔT uis As ) increases, the rate of decrease becomes gentler, allowing the calculation to remain stable even with extreme time values.

[0145] The second item may be characterized by being designed to numerically express whether there is a problem with the facility user's condition or an abnormality in the equipment by combining vibration sensor values, weight sensor values, and equipment status scores.

[0146] In other words, the data processing algorithm is the vibration intensity (V uis ), weight (W uis ) and equipment status indicators (H uis Normalize each of the ) and calculate the average to obtain user health indicators (S uis It can be characterized by being designed to extract ).

[0147] Here, equipment status indicator (H uis ) can be characterized by being designed so that the indicator increases as the condition worsens, using a value subtracted from 1.

[0148] The data processing algorithm can be characterized by normalizing vibration, weight, and equipment condition to values ​​between 0 and 1, so that when the average is calculated, it can become an indicator that summarizes the risk level at a glance.

[0149] In addition, the data processing algorithm is an equipment status indicator (H) that represents the state of the equipment. uis ) 1- H uis If changed to, the value increases as the condition worsens, so the risk level can be intuitively expressed, and it can be characterized as being designed to be used as a value subtracted from 1.

[0150] The third item is the importance score (P uis User health indicators extracted from ) and the second item (S uis Mixed weights (α) in ) uis It can be characterized by being designed to reflect ).

[0151] Importance score (P) assigned by the operator or system uis ) and user health indicators (S uis ) mixed weights (α uis By utilizing and mixing ), field sensor changes and external priorities can be adjusted according to the situation, so the importance score (P uis ) and user health indicators (S uis ) mixed weights (α uis It can be characterized by being designed to mix using ).

[0152] If, mixed weight (α uis If ) is increased, the importance score (P) in extracting the ranking score (UIS) uis You can place more weight on ), and mixed weight (α uis If ) is reduced, sensor-based judgment may be given higher priority in extracting the ranking score (UIS).

[0153] Since the reliability of data processing algorithms decreases as data ages, elapsed time (ΔT) is used to lower the integration priority. uisYou can utilize ).

[0154] Elapsed time (ΔT uis ) can use the unit of 'seconds' and may refer to the time elapsed from the time the data of the equipment was last updated until now.

[0155] Elapsed time (ΔT uis ) may mean a value derived by comparing the timestamp of the last data received from the equipment by the management server (100) with the current server time.

[0156] The data processing algorithm incorporates vibration intensity (V) to reflect important factors for detecting the possibility of falls, such as bed displacement and violent movements. uis You can utilize ).

[0157] Vibration intensity (V uis ) is m / s 2 Units can be used, and it can refer to the vibration intensity value measured from the vibration sensor.

[0158] Vibration intensity (V uis ) can be a value derived from values ​​measured and collected in real time by an acceleration-based vibration sensor attached to the bed frame.

[0159] The data processing algorithm uses weight (W) to indirectly determine health status, whether the patient has left the bed, etc. uis You can utilize ).

[0160] Weight (W uis ) can use the unit 'kg' and may refer to the weight value of the facility user measured by a weight sensor already equipped on the bed, and may be derived by a measurement value collected in real time from a load sensor attached to the bottom of the bed or a specific location.

[0161] Since the reliability of collected data decreases if the equipment itself is unstable, the data processing algorithm uses equipment status indicators (H) to reflect this. uis You can utilize ).

[0162] Equipment Status Indicator (H uis ) represents a health status indicator of the connected equipment and can be used as a dimensionless value, and can refer to a value derived by converting the error rate, response delay, battery level, etc. of the equipment itself into a value between 0 and 1.

[0163] The data processing algorithm uses the user health indicator (S) to more intuitively express the overall likelihood of abnormalities in facility users by combining the average value of vibration, weight, and equipment condition, rather than viewing each item separately. uis You can utilize ).

[0164] User health indicators (S uis As mentioned above, ) is a comprehensive indicator representing the possibility of abnormal conditions of facility users, and vibration intensity (V uis ), weight (W uis ) and equipment status indicators (H uis It can mean the average value after normalizing each of the ).

[0165] Data processing algorithms use importance scores (P) to prioritize the processing of more important patient data, even if they have the same risk level. uis You can utilize ).

[0166] Importance score (P uis ) may represent the importance score of the equipment and patient assigned by the operator or system, and may have a dimensionless value between 0 and 1.

[0167] The server operator or administrator has an importance score (P uis You can directly set the value between 0 and 1 based on the patient's condition or medical records and directly input a value between 0 and 1 into the server or device linkage unit (150).

[0168] Data processing algorithms use a rank threshold (T) to reduce unnecessary throughput. uis Using ), the rank score (UIS) is the rank threshold (T uisIf it exceeds ), it can be synchronized immediately.

[0169] The server operator or administrator must meet the rank threshold (T uis You can directly set the value between 0 and 1 based on past data records and directly input it to the server or device linkage unit (150).

[0170] Rank score (UIS) rank threshold (T uis If it exceeds ), it can be processed immediately through integration and notification.

[0171] Rank threshold (T uis Lowering ) makes it more sensitive and allows for the detection of more risk situations, but since throughput may increase, the rank threshold (T uis The standards of ) may be changed in accordance with the operating policy.

[0172] The data processing algorithm uses mixed weights (α) to appropriately combine external settings and sensor information depending on the situation. uis You can utilize ).

[0173] Mixed weights (α uis ) is a user health indicator (S uis ) and importance score (P uis It can mean a ratio for mixing ) and can have a dimensionless value between 0 and 1.

[0174] The server operator or administrator has mixed weights (α uis You can directly set the value between 0 and 1 based on past data records and directly input it to the server or device linkage unit (150).

[0175] The actual numerical examples of input and output values ​​of a data processing algorithm can be explained as follows.

[0176] Vibration intensity (V), which is the measured value of the vibration sensor attached to the bed uis ) is 20, and the weight (W), which is the measured value of the weight sensor. uis ) is 70, equipment status indicator (H uis) is 0.9 and the elapsed time (ΔT) since the last update of the equipment uis ) is 300 seconds and importance score (P uis ) is 0.8, mixed weight (α uis ) is 0.6, a fixed rank threshold (T uis When ) is 0.5, user health indicators (S uis ) can be extracted as 0.3333, and the final result value, the ranking score (UIS), can be extracted as 0.0915.

[0177] At this time, the threshold value (T uis Since the ranking score (UIS) is smaller compared to 0.5, it may not immediately trigger synchronization or emergency notifications.

[0178] Conversely, from the same equipment just (elapsed time (ΔT) uis ) = 0 sec) Vibration intensity (V uis ) = 90, weight(W uis ) = 85, Equipment Status Indicator(H uis ) = 0.6, Importance score (P uis ) = 0.9, mixed weights(α uis If ) = 0.7 is collected and set, the value of the ranking score (UIS) can be extracted as 0.845.

[0179] At this time, the rank score (UIS) is the threshold (T uis Since the value is greater than ), the data of the corresponding equipment can be immediately linked, and screen display and administrator notifications can be generated.

[0180] The management server (100) can perform the following procedure in real time by running a repeating loop for each connected device.

[0181] First, the latest timestamp and vibration intensity (V) from the equipment (e.g., weight sensor, vibration sensor). uis ), weight (W uis ) and equipment status indicators (H uis Receive ) and elapsed time (ΔT uis ) can be calculated.

[0182] Next, vibration intensity (V uis ) and weight (W uis ), Equipment status indicator (H uis User health indicators (S) from ) uis Calculate ) and importance score (P uis Mixed weights (α) in ) uis Reflecting ) and the elapsed time processed by the logarithmic function (ΔT uis Rank scores (UIS) can be extracted using ).

[0183] The device linkage unit (150) has a ranking score (UIS) and a ranking threshold (T uis If it exceeds ), you can perform immediate synchronization and notification as an event trigger.

[0184] If missing values ​​occur at each step, missing value handling rules (e.g., interpolating with the EWMA of the most recent N samples or attaching a low confidence flag) can be applied.

[0185] Since the data processing algorithm requires only constant-time operations (normalization, arithmetic operations, comparison) per sensor, it can be scaled to a level that satisfies real-time requirements through parallelization and event-based processing even when large-scale equipment is connected.

[0186] In addition, for continuous samples, noise can be suppressed and the stability of threshold comparisons ensured by applying EWMA (Exponentially Weighted Moving Average) or a sliding window (e.g., the average of the most recent N samples).

[0187] Here, EWMA coefficients or window lengths can be adjusted according to operating policies or data characteristics, and since specific values ​​may vary depending on the installation environment, the initial values ​​may be indicated as "estimated."

[0188] The device linkage unit (150) can normalize all input values ​​to a range of 0 to 1 to perform comparison and combination on the same scale, and may also be characterized by reflecting a weighted average according to α.

[0189] At this point, the final decision condition is basically the rank score (UIS) being the rank threshold (T uis If it exceeds ), immediate integration and notification are executed, but multiple condition combinations may be provided as supplementary rules.

[0190] Elapsed time (ΔT) while no data is received from the vibration sensor or weight sensor uis Since ) continuously increases, the log-based elapsed time (ΔT uis ) has the characteristic of gradually decreasing over time, so the ranking score (UIS) decreases over time, which allows old data to be automatically excluded from priority processing.

[0191] Conversely, when the equipment transmits the latest data, the elapsed time (ΔT uis As ) approaches 0, the logarithmic function value approaches 1, so the same sensor risk level can be assigned a higher priority.

[0192] The server administrator or operator directly mixes the weights (α uis Increasing the ) value increases the importance score (P uis As the influence of ) increases, the priority assigned by the administrator may take on a larger weight in the system's judgment.

[0193] Conversely, mixed weights (α uis Lowering the ) value of user health indicators (S uis As the influence of ) becomes relatively large, automatic priority judgment based on field sensor fluctuations may be prioritized.

[0194] For example, same user health indicators (S uis ) and importance score (P uis In the situation, mixed weights (α uis It can be numerically confirmed that changing ) from 0.5 to 0.8 can cause the ranking score (UIS) to rise or fall.

[0195] Rank threshold (T uisIn the case of ), if the operator lowers the value, sensitivity increases, allowing more equipment to be immediately linked, and if the operator raises the value, sensitivity decreases, allowing system throughput to be saved, so the operator can select an appropriate value that reflects the balance between the safety and processing costs required.

[0196] The technical description and examples of the above data processing algorithms are sufficiently provided so that a person skilled in the art can easily implement the present invention according to mathematical formulas and variable definitions, and it can be confirmed that the implementation of the present invention is described at a level that is obvious to a person skilled in the art.

[0197] The report unit (170) can collect and manage data collected through the vibration sensor (300), weight sensor (500) and life monitoring, as well as data on at least one facility user, and can generate a report in a pre-set format at a predetermined period.

[0198] Additionally, the report unit (170) can derive information on the current status of facility users by utilizing an AI algorithm that learns and analyzes updated data, and provide the information on the current status of facility users derived by utilizing the AI ​​algorithm to the monitoring unit (190).

[0199] Here, the updated data may refer to weight collected through the facility user's weight sensor and vital data, including pulse collected through regular checkups.

[0200] The above-described AI algorithm can analyze the condition of a facility user by applying at least one of an anomaly detection algorithm, a time series prediction algorithm, and a pattern classification algorithm based on vital data and environmental data collected through a vibration sensor (300), a weight sensor (500), and life monitoring.

[0201] In addition, the AI ​​algorithm may be characterized by being configured to apply pre-trained data and real-time data for each facility user in parallel to analyze a customized state prediction for the facility user, automatically adjust a customized threshold for each facility user based on the results of the customized state prediction analysis, and transmit to a monitoring unit and an exit warning unit if a risk indicator is predicted for the facility user based on the results of the customized state prediction analysis.

[0202] In addition, AI algorithms can be characterized by applying federated learning techniques to learn customized prediction models while protecting personal information.

[0203] In addition, the report section may be characterized by applying a blockchain-based data recording method to store collected health data in a form that cannot be tampered with or falsified.

[0204] Meanwhile, the report unit (170) can utilize an anomaly detection algorithm to quantitatively evaluate the real-time health status of facility users and to detect abnormal signs early by extracting an integrated result (S(t)) from real-time weight (W(t)), vibration intensity (V(t)), vital data (B(t)), environment temperature (T), weight weight (w1), and vibration weight (w2).

[0205] The anomaly detection algorithm can improve prediction accuracy compared to single-sensor-based evaluations by integrating and calculating weight, vibration, vital signs, and environmental data, and can perform judgment and control functions by directly linking with the monitoring unit and deviation alarm unit that convert abnormal conditions into early warnings.

[0206] In addition, it can reduce the false positive rate that occurred with existing single sensors or simple reference value comparison methods, thereby improving the average anomaly detection rate and enhancing facility operational efficiency.

[0207] The anomaly detection algorithm is designed to reflect different measurement data for evaluating the condition of facility users and can be characterized by being designed as the sum of a first item that sensitively reflects signs of health abnormalities while ensuring that sudden changes in weight do not have an excessively large impact on the overall score, a second item that evaluates the risk of bed displacement or falls in real time, and a third item that incorporates vital data into the score through logarithm and multiplication after normalization and is designed to comprehensively consider the impact of environmental factors on the health status evaluation.

[0208] Since the reliability of the condition assessment may decrease due to extreme values ​​or deviations when simply summing weight, vibration, and vital data, the anomaly detection algorithm may be characterized by being configured to normalize outliers in weight change by applying a logarithmic function and to reflect the periodic characteristics of vibration through a sine function.

[0209] In other words, weight values ​​can suddenly spike or plummet due to brief movements or placing objects on them; for this reason, using the weight sensor measurements directly can result in inaccurate readings.

[0210] To prevent this, the anomaly detection algorithm uses a logarithmic function to mitigate large input values, allowing for stable reflection when extracting the integrated result (S(t)).

[0211] In addition, vibration is caused by periodic movements (tossing and turning, regular shaking, etc.), and the trigonometric function sine can express periodicity like a wave, so it can be characterized by being designed using it to naturally express vibration patterns.

[0212] As mentioned above, the anomaly detection algorithm can distinguish between regular shaking and sudden impacts thanks to being designed using the trigonometric function sine.

[0213] In addition, the anomaly detection algorithm is characterized by being designed to enable an integrated health status assessment rather than a single-factor-based judgment by combining vital data and environmental temperature data, and thereby allowing for the automatic adjustment of customized thresholds in conjunction with an AI-based state prediction algorithm.

[0214] Since heart rate and blood pressure differ in both unit and magnitude, applying them to anomaly detection algorithms without normalization can lead to data bias; therefore, they can be reflected after first aligning their units through a normalization process.

[0215] Vital data values ​​aligned with units through the normalization process can detect even slight anomalies as significant changes, allowing for sensitive detection of minor abnormalities.

[0216] In addition, since the environment can affect health data—such as the heart rate naturally increasing as the environmental temperature rises—reflecting the environmental temperature is a characteristic feature, as it allows for obtaining results closer to the actual health condition.

[0217] Anomaly detection algorithms can utilize real-time body weight (W(t)) as an important indicator of signs of health abnormalities for quantitative evaluation.

[0218] Real-time weight (W(t)) can use the unit 'kg' and may refer to the real-time weight of the facility user; it can collect values ​​measured in real-time through a weight sensor already installed in the bed frame and can be derived by converting the electrical signal of the weight sensor into a digital value.

[0219] Anomaly detection algorithms can utilize vibration intensity (V(t)) to detect bed deviation or violent movement, and can use the trigonometric function sine to analyze user patterns.

[0220] Vibration intensity (V(t)) may use a sensor output unit that is an 'arbitrary unit' and may refer to the vibration intensity collected from a vibration sensor already equipped in the bed frame.

[0221] Vibration intensity (V(t)) can refer to a value derived by measuring the vibration of the bed frame in real time using a vibration sensor and converting the voltage change into a digital signal.

[0222] The anomaly detection algorithm can utilize vital data (B(t)) for a comprehensive assessment of the user's health status and can derive and utilize the user's risk score by combining it with environmental variables.

[0223] Vital data (B(t)) can utilize vital units (bpm, mmHg), can refer to the vital data (heart rate) of a facility user, and can be derived from data collected through medical equipment and wearable sensors connected to the user.

[0224] Since temperature changes affect the health status score, the anomaly detection algorithm can utilize ambient temperature (T) as a correction factor affecting the integrated result.

[0225] The ambient temperature (T) may be expressed in °C units and may refer to the current temperature value received from the facility user's internal temperature sensor or external environmental data.

[0226] The anomaly detection algorithm can derive weight weights (w1) and vibration weights (w2) based on past data through statistical optimization or AI model learning, and the setting criteria and initial values ​​for each weight can be automatically determined by the system or adjusted by the user.

[0227] As described above, the weight weight (w1) and vibration weight (w2) can be determined from the analysis of past data or the results of AI learning. For example, if the analysis of past fall accidents revealed that changes in vibration were a more important signal than weight, the AI ​​model can derive a large value for the vibration weight (w2) to give greater importance to vibration.

[0228] Conversely, since the combination of weight and vital changes is important for health deterioration, the AI ​​model can derive a large weight weight (w1) so that the weight weight (w1) plays a significant role in giving greater importance to weight.

[0229] Anomaly detection algorithms can clarify the computation process by inputting actual data.

[0230] For example, if a heart rate of 85 bpm and an activity level of 1200 steps are input, the integrated result (S(t)) can be extracted as 0.78 depending on the design configuration of the algorithm.

[0231] This is classified as a 'caution required' level, which may mean that the calculation results of the formula can be directly connected to real-time monitoring and warning systems. This process is handled through a flow of input data collection, formula computation, and result classification, and numerical changes at each stage can be reflected in the technical effects.

[0232] The anomaly detection algorithm can be implemented as a software algorithm and can be configured with a flow that calculates the average of data collected from the sensor every 3 seconds, compares it with a reference value to determine whether the condition is met, and outputs a warning when the condition is met.

[0233] Anomaly detection algorithms involve real-time variable updates and condition evaluation processes, which may imply that the same logical functions can be implemented in software code.

[0234] If conditions exist within the anomaly detection algorithm, the criteria for judging the conditions may be defined according to specific numerical values ​​or setting methods.

[0235] For example, if conditions A and B are each evaluated with a weight of 1.0 and condition C is evaluated with a weight of 1.5, and the total score is 70 points or more, it can be determined that the conditions are satisfied.

[0236] If the reference value is dynamic, it implies that the system can automatically adjust the reference value based on user feedback or environmental change data, and can combine conditions through AND, OR operations and weighted summation methods.

[0237] Anomaly detection algorithms can be designed to process data that changes over time.

[0238] For example, abnormal patterns can be identified by comparing the change in average heart rate measured at 5-minute intervals, and an alert can be triggered if it increases by more than 20% compared to the previous interval. This logic for comparison and judgment at different time intervals enables continuous data monitoring and real-time response.

[0239] The target value set by the user is directly reflected in the anomaly detection algorithm to adjust its operation.

[0240] For example, if the user enters a target heart rate of 120 bpm, automatic correction within a range of ±15 bpm is performed according to the exercise intensity, and the notification threshold can be adjusted accordingly.

[0241] User input values ​​interact with other variables to influence the calculation of results and the determination of conditions, which may mean that the system can reflect this to perform customized warnings or actions.

[0242] The technical description and examples of the above anomaly detection algorithm are sufficiently provided so that a person skilled in the art can easily implement the present invention according to mathematical formulas and variable definitions, and it can be confirmed that the implementation of the present invention is described at a level that is obvious to a person skilled in the art.

[0243] The monitoring unit (190) can output the health status of all facility users, including emergency response, and the situation of facility users, through data collected via the sensor, to a display screen linked via the device linkage unit (150).

[0245] The embodiments described above are for illustrative purposes only, and those skilled in the art will understand that the embodiments described above can be easily modified into other specific forms without altering the technical concept or essential features of the embodiments described above. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0247] The scope of protection sought through this specification is defined by the claims set forth below rather than by the detailed description, and should be interpreted to include all modifications or variations derived from the meaning and scope of the claims and the concept of equivalents. Explanation of the symbols

[0248] 100: Management Server

Claims

Claim 1 A bed sensor system comprising: a management server for checking the physical condition of a facility user and preventing fall accidents of said facility user; wherein the management server comprises: an escape alarm unit that measures and collects in real time whether the bed frame is vibrating and the vibration intensity through a vibration sensor already provided integrally on the bed frame, and generates an alarm when the facility user leaves or a specific movement occurs; a weight measurement unit that measures and collects the accurate weight of the facility user in real time through a weight sensor already provided integrally on the bed frame, and classifies, accumulates, records, and manages the collected weight data by each facility user; a device linkage unit that links with at least one device provided in a facility including a hospital and capable of linkage, and collects data input for each facility user from the linked device to perform lifestyle monitoring for each facility user; and a report unit that collects and manages data collected through the vibration sensor, the weight sensor, and lifestyle monitoring, as well as data for at least one facility user, and generates a report in a pre-set format at predetermined intervals. A monitoring unit that outputs the health status of all facility users and the situation of facility users, including emergency response, through data collected via the sensor to a display screen linked via the device linkage unit; wherein the departure alarm unit generates an alarm by determining that the facility user has left the bed or has made vigorous movements when the vibration intensity collected via the vibration sensor exceeds a preset threshold value, and the preset threshold value is a value entered as an initial setting by a manager or operator, which is a value smaller than the intensity collected via the vibration sensor when the user leaves the bed or makes vigorous movements; wherein the device linkage unit collects updated data of the facility user in the case of a facility user for whom updated data exists from the linked equipment, and the report unitThe present state analysis information of a facility user is derived by utilizing an AI algorithm that learns and analyzes the aforementioned updated data, and the present state analysis information of a facility user derived by utilizing the aforementioned AI algorithm is provided to the aforementioned monitoring unit; the aforementioned updated data refers to vital data including weight collected through the facility user's weight sensor and pulse collected through regular checkups; the aforementioned AI algorithm analyzes the state of a facility user by applying at least one of an anomaly detection algorithm, a time series prediction algorithm, and a pattern classification algorithm based on vital data and environmental data collected through the aforementioned vibration sensor, the aforementioned weight sensor, and the aforementioned life monitoring; the aforementioned AI algorithm analyzes a state prediction customized for the facility user by applying pre-learned data and real-time data for each facility user in parallel; the facility user's state prediction analysis result is configured to automatically adjust a customized threshold value for each facility user according to the facility user's state prediction analysis result; and if a risk sign is predicted for the facility user according to the facility user's state prediction analysis result, the AI ​​algorithm is configured to transmit to the aforementioned monitoring unit and the aforementioned departure alarm unit; and the weight measurement unit operates as an internal processing module to refine the signal coming from the weight sensor in real time and to transmit the final weight value to the aforementioned reporting unit and the aforementioned monitoring unit. By utilizing an alarm judgment algorithm to calculate, the reliability of real-time measurements obtained from a bed-integrated weight sensor is increased, and false positives (errors) caused by vibration, temperature, and temporary noise are reduced to accurately estimate the actual weight of the facility user. The alarm judgment algorithm is designed to convert the measurement value of the weight sensor into kg units and correct it using a zero offset (C0) to serve as the basis for sensor calibration, and includes a first item that converts the raw sensor value into physical weight. The alarm judgment algorithm is designed to correct distortions caused by dynamic movement and ambient temperature together, and includes a second item that continuously corrects errors caused by vibration and temperature.The above alarm judgment algorithm is designed to reduce short-term noise and stably handle rapid numerical changes by weighted averaging the value measured from the actual sensor and the current correction value, and includes a third item that mixes past values ​​and current correction values ​​and limits them by physical limits, and the above device linkage unit has an elapsed time (ΔT, to automatically determine which of the devices among several devices to process the data of first uis ), vibration intensity (V uis ), weight (W uis ), health indicators (H uis ), User health indicators(S uis ), importance score (P uis ), rank threshold (T uis ), mixed weights (α uis A data processing algorithm is utilized to extract a ranking score (UIS) from ), and the data processing algorithm is designed to quantify and comprehensively reflect the data recency, abnormal signs detected by sensors, and importance set by the operator, and the device linkage unit uses the data processing algorithm to increase the priority so that when strong vibration occurs in a patient's bed, weight change is detected, and the patient's condition is critical, the patient's equipment data is transmitted to the server and displayed on the monitoring screen, and the data processing algorithm includes a first item indicating how recently the data of a specific piece of equipment was updated, a second item indicating whether there is a problem with the facility user's condition and whether there is an equipment malfunction, and user health indicators (S uis ) and importance score (P uis It is characterized by being composed of a third item utilizing ), and the first item of the data processing algorithm is characterized by being designed so that the value is higher the shorter the data collection period and the value is lower the older the data, and the data processing algorithm is characterized by elapsed time (ΔT) to lower the importance of old data uis When ) is small, a large value is given, but elapsed time (ΔT uis It is characterized by using a logarithmic function that ensures the calculation remains stable even with extreme time values ​​by making the rate of decrease gentler as ) increases, and the second item of the data processing algorithm is characterized by being designed to numerically express whether there is a problem with the facility user's condition or an abnormality in the equipment by combining the vibration sensor value, the weight sensor value, and the equipment status score, and the data processing algorithm is characterized by vibration intensity (V uis ), weight (W uis ) and equipment status indicators (H uis Normalize each of the ) and calculate the average to obtain user health indicators (S uis Characterized by being designed to extract ), the data processing algorithm normalizes vibration, weight, and equipment status to 0 to 1, calculates the average, and provides a user health indicator (S) that summarizes the risk level at a glance. uis Characterized by outputting ), and the data processing algorithm comprises the equipment status indicator (H) representing the state of the equipment. uis ) to 1- H uis It is characterized by being designed to intuitively express the level of risk by using a replacement so that the value increases as the condition worsens, and the third item of the above data processing algorithm is the importance score (P uis User health indicators extracted from ) and the second item (S uis Mixed weights (α) in ) uis Characterized by being designed to reflect ), and the data processing algorithm uses 'seconds' units and represents the elapsed time (ΔT) which signifies the time elapsed from the point when the data of the equipment was last updated until the present. uis It is characterized by being designed to lower the integration priority as the reliability decreases with age by utilizing the elapsed time (ΔT uis ) refers to a value derived by comparing the timestamp of the last data received by the management server from the equipment with the current server time, and the data processing algorithm is m / s 2 The above vibration intensity (V), which uses units and refers to the vibration intensity value measured by the vibration sensor. uis It is characterized by being designed to reflect important factors in detecting the possibility of falls, including bed displacement and violent movements, by utilizing the vibration intensity (V uis ) is derived from values ​​collected and measured in real-time by an acceleration-based vibration sensor attached to the bed frame, and the data processing algorithm uses the unit 'kg' and refers to the weight (W) of the facility user measured by a weight sensor already equipped on the bed. uis Indirectly determining health status, whether the patient has left the bed, etc. by utilizing ), and the above weight (W uis ) is derived by measurement values ​​collected in real time from a load sensor attached to the bottom of the bed or a specific location, and the data processing algorithm refers to an equipment status indicator (H) that represents the health status indicator of the linked equipment and is derived by comprehensively analyzing the equipment's own error rate, response delay, and remaining battery level and converting them to a value between 0 and 1. uis It is characterized by being designed to reflect that if the equipment itself is unstable, the reliability of the collected data also decreases by utilizing ), and the data processing algorithm refers to a comprehensive indicator representing the possibility of abnormal condition of a facility user and vibration intensity (V uis ), weight (W uis ) and equipment status indicators (H uis User health indicator (S), which is the average value obtained after normalizing each of ) uis It is characterized by being designed to express the overall possibility of abnormalities in facility users based on vibration, weight, and equipment condition using ), and the data processing algorithm refers to the importance score of the corresponding equipment and patient assigned in advance, and the importance score (P uis It is characterized by being designed to prioritize the processing of more important patient data even with the same risk level by utilizing a ), and the data processing algorithm comprises a rank threshold (T uis By utilizing ), unnecessary throughput is reduced, and the rank score (UIS) is the rank threshold (T uis If it exceeds ), it synchronizes immediately, and the above data processing algorithm, user health indicator (S uis ) and importance score (P uis Mixing weight (α) representing the ratio for mixing ) uis Utilizing ), it combines external settings and sensor information depending on the situation, and mixes weights (α uis ) is a user health indicator (S uis ) and importance score (P uis It refers to the ratio for mixing ), can have a dimensionless value between 0 and 1, and the server operator or administrator uses the mixing weight (α uis ) is directly set based on past data records and a value between 0 and 1 is directly input to the server or device linkage unit, and the device linkage unit, wherein the ranking score (UIS) is a threshold value (T uis If the value is greater than ), the data of the relevant equipment is immediately linked, and a screen display and administrator notification are generated; and the management server, for each connected equipment, receives the latest timestamp and vibration intensity (V) from the equipment including the weight sensor and vibration sensor. uis ), weight (W uis ) and equipment status indicators (H uis Receive ) and elapsed time (ΔT uis Calculate ) and vibration intensity (V uis ) and weight (W uis ), Equipment status indicator (H uis User health indicators (S) from ) uis Calculate ) and importance score (P uis Mixed weights (α) in ) uis Reflecting ) and the elapsed time processed by the logarithmic function (ΔT uis The device interface unit extracts the ranking score (UIS) using ), and the ranking score (UIS) is the ranking threshold (T uis A bed sensor system that performs immediate synchronization and notification via an event trigger when ) is exceeded, and executes a procedure to apply missing value processing rules in real time by running a repeating loop when a missing value occurs at each step, and the data processing algorithm requires only constant-time operations (normalization, arithmetic operations, comparison) per sensor, and is scalable to a level that satisfies real-time requirements through parallelization and event-based processing when large-scale equipment is connected. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete Claim 6 delete Claim 7 A bed sensor system according to claim 1, characterized by additionally collecting respiration and heart rate data including the management server, a non-contact radar sensor, or an optical PPG sensor. Claim 8 A bed sensor system according to claim 1, wherein the AI ​​algorithm is characterized by applying a federated learning technique to learn a customized prediction model while protecting personal information. Claim 9 A bed sensor system according to claim 1, wherein the report unit stores collected health data in an unalterable form by applying a blockchain-based data recording method. Claim 10 A bed sensor system according to claim 1, wherein the management server analyzes sleep stages based on body weight, vibration, and vital data, and accordingly detects abnormal signs early.

Images