Learning patient habitual behavior off-bed alarm delay triggering method and system

Abstract

The application belongs to the technical field of alarm devices, and specifically provides a bed leaving alarm delay triggering method and system for learning habitual behavior of a patient. The method mainly extracts the static and extreme value load of the mattress pressure sensitive data, quantifies the turning over struggle strength of the patient, combines with the historical behavior fluctuation calculation time compensation, dynamically constructs the impedance matching delay threshold, and realizes the individual adaptive triggering of the bed leaving alarm. The application effectively overcomes the defects that the fixed delay alarm is difficult to adapt to the sudden physiological impedance of the patient in the rehabilitation period and is easy to mis-trigger the timeout alarm when the patient moves slowly, can accurately perceive and adaptively match the single sudden pain and movement difficulty of the patient, provides sufficient moving time for the slow-moving patient, reduces the false alarm, and effectively ensures the monitoring safety in the abnormal stay scene.

Description

Technical Field

[0001] This application belongs to the field of alarm device technology, and specifically relates to a method and system for delayed triggering of bed exit alarm based on learning patient habitual behaviors. Background Technology

[0002] In the fields of rehabilitation monitoring and smart mattress monitoring, the pressure sensor array embedded inside the mattress monitors the patient's bed load status in real time. Based on a fixed delay time set by the system according to the patient's historical average habits as a decision threshold, the system can trigger alarms for exceeding the time limit for leaving the bed and provide safety monitoring.

[0003] However, patients in the recovery period who experience bone and joint pain or postoperative morning stiffness generally exhibit biomechanical resistance to limb movement caused by joint inflammation, effusion, or muscle stiffness. This physiological resistance is characterized by sudden and unpredictable fluctuations in each individual action and each day. Before officially leaving the bed, patients may engage in physically demanding preparatory movements such as repeatedly turning over and supporting themselves in bed. Furthermore, their walking frequency decreases significantly after getting out of bed, inevitably causing a drastic and sudden increase in the total physical time spent on toileting and returning. As a result, the system is prone to frequently triggering alarms when patients move with difficulty or walk slowly while enduring pain, which can easily lead to false alarms. Summary of the Invention

[0004] This application provides a method and system for delayed triggering of bed exit alarms based on learning patients' habitual behaviors. This effectively overcomes the shortcomings of fixed-delay alarms, which are difficult to adapt to sudden physiological resistance in patients during the recovery period and are prone to falsely triggering timeout alarms when patients are slow to move. It can accurately sense and adaptively match a patient's single sudden pain and difficulty in moving. While providing sufficient time for patients with slow movement and reducing false alarms, it effectively ensures the safety of monitoring in abnormal stay scenarios.

[0005] To achieve the above objectives, this application adopts the following technical solution:

[0006] In a first aspect, this application provides a method for delayed triggering of an alarm based on learning a patient's habitual behavior, comprising: responding to an exit trigger signal, acquiring mattress pressure-sensitive distribution data within a preset time sequence before the occurrence of the exit trigger signal, and retrieving a historical exit event record set; extracting the statistical center value of the historical exit event record set to obtain the habitual baseline duration; performing time-domain feature extraction on the mattress pressure-sensitive distribution data to separate static background load and local extreme load; filtering out resting weight deviation from the local extreme load based on the static background load, and calculating the absolute struggle intensity and relative intensity characterizing the physical turning amplitude; mapping the fluctuation bandwidth based on the physical dwelling extreme features in the historical exit event record set, and using the fluctuation bandwidth to perform gain adjustment on the relative intensity to obtain a hysteresis time compensation parameter; performing adaptive tolerance reconstruction based on the absolute struggle intensity, hysteresis time compensation parameter, and habitual baseline duration to generate an impedance matching delay threshold; recording the current dwell time in real time, and activating an alarm command when the current dwell time exceeds the impedance matching delay threshold.

[0007] Secondly, this application provides a delayed triggering system for bed exit alarm based on learning patient habitual behaviors, comprising:

[0008] Data acquisition module: In response to the bed-leaving trigger signal, it acquires mattress pressure-sensitive distribution data within a preset time sequence before the bed-leaving trigger signal occurs, and retrieves historical bed-leaving event records.

[0009] Feature extraction module: used to extract the statistical center value of the historical bed leaving event record set to obtain the habitual baseline time; to perform time-domain feature extraction on the mattress pressure-sensitive distribution data, and to separate the static background load and local extreme value load.

[0010] Intensity calculation module: used to filter out resting weight deviation from local extreme loads based on the static background load, and calculate the absolute struggle intensity and relative intensity that characterize the physical turning amplitude.

[0011] Parameter adjustment module: used to map the fluctuation bandwidth according to the physical residence extreme value characteristics in the historical bed leaving event record set, and use the fluctuation bandwidth to perform gain adjustment on the relative intensity to obtain the hysteresis time compensation parameter.

[0012] Threshold generation module: used to perform adaptive tolerance reconstruction based on the absolute struggle intensity, hysteresis time compensation parameter and habitual reference time, and generate impedance matching delay threshold.

[0013] Alarm determination module: used to record the current time spent away from the bed in real time, and to activate an alarm command when the current time spent away from the bed exceeds the impedance matching delay threshold.

[0014] Thirdly, this application provides a readable storage medium, comprising: computer program instructions stored in the readable storage medium, wherein the computer program instructions are read and executed by a processor to perform the step of a method for delaying the triggering of an alarm for learning patient habitual behaviors.

[0015] The beneficial effects of this application are:

[0016] This application extracts historical bed-out time as a habitual benchmark, separates the static background and extreme load of pressure-sensitive data before bed-out to quantify the intensity and relative severity of struggle, and combines it with historical fluctuation bandwidth for time compensation, thereby reconstructing a dynamic impedance matching delay threshold to trigger an alarm. This effectively overcomes the shortcomings of fixed delay alarms, which are difficult to adapt to sudden physiological impedance in patients during the recovery period and are prone to falsely triggering timeout alarms when patients are slow to move. It can accurately sense and adaptively match a patient's single sudden pain and difficulty in movement, providing sufficient time for patients with slow movement and reducing false alarms, while effectively ensuring the monitoring safety in abnormal stay scenarios.

[0017] Other features and advantages of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description and the accompanying drawings. Attached Figure Description

[0018] Figure 1 A flowchart illustrating the method for delayed triggering of bed exit alarm based on learning patient habitual behaviors according to this application is shown.

[0019] Figure 2 A schematic diagram illustrating the process for calculating the absolute struggle intensity and relative intensity of this application is shown.

[0020] Figure 3 A schematic diagram of the process for generating the impedance matching delay threshold in this application is shown;

[0021] Figure 4 A schematic diagram of the module of the bed-leaving alarm delay triggering system for learning patient habitual behaviors according to this application is shown. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments 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.

[0023] Patients in the recovery period experiencing bone and joint pain or postoperative morning stiffness. These patients often cannot get out of bed as quickly as healthy individuals due to physiological discomfort, joint pain, or limb stiffness. Before officially leaving the bed, patients often need to perform a series of preparatory movements in bed, including but not limited to repeatedly turning over, moving limbs, using their hands or elbows for support before slowly getting out of bed. These movements physically manifest as tossing and turning in bed. This pre-existing mechanical fluctuation, generated by mattress pressure-sensitive distribution data, objectively reflects the patient's current physiological resistance and degree of difficulty in movement at a biomechanical level.

[0024] In this protocol, physiological resistance refers to the biomechanical resistance that patients experience to limb displacement due to joint inflammation, effusion, or muscle stiffness. This resistance directly leads to a higher physical effort required for patients to perform actions such as turning over or getting out of bed, and inevitably results in a decrease in their walking frequency after officially getting out of bed, thereby increasing the total time spent on toileting and returning.

[0025] Existing alarm methods often rely on historical averages to set a fixed delay time when setting the alarm delay time for leaving the bed. This approach easily overlooks sudden fluctuations in the patient's physiological state during a single action and ignores sudden slowness in movement. It is very easy to trigger an alarm due to exceeding the time limit when the patient is moving with difficulty, resulting in false alarms.

[0026] This solution accurately captures the transient mechanical characteristics before leaving the bed and dynamically constructs an impedance matching delay threshold. Impedance matching here refers to the electronic system's prediction delay time being consistent with the patient's current biological resistance state in terms of magnitude. This allows the solution to understand and adapt to the patient's current physiological state and achieve personalized triggering of alarm timing.

[0027] In some embodiments, such as Figure 1 As shown, this application provides a delayed triggering method for bed exit alarm based on learning patient habitual behaviors, including:

[0028] S1. In response to the bed-leaving trigger signal, obtain the mattress pressure-sensitive distribution data within a preset time sequence before the bed-leaving trigger signal occurs, and retrieve the historical bed-leaving event record set.

[0029] The pressure is monitored in real time by an array of pressure sensors embedded inside the mattress. When the total load on the bed surface disappears and the preset judgment time is reached, it is determined that the action of getting out of bed has occurred, and then the action of getting out of bed trigger signal is triggered.

[0030] Upon receiving the bed-leaving trigger signal, data indexing technology is used to retrieve a continuous time observation window, or preset time sequence, from local memory preceding the occurrence of the trigger signal. This preset time sequence is typically set to 300 to 600 seconds before the bed-leaving action, used to fully capture the patient's tossing and turning process due to pain or stiffness before getting out of bed. Within this time interval, the pressure values ​​at each sampling moment output by the sensor array are collected, forming mattress pressure-sensitive distribution data. This mattress pressure-sensitive distribution data refers to a discrete numerical matrix sequence reflecting the change in pressure intensity with spatial position, collected by multiple pressure sensing units in a two-dimensional spatial plane.

[0031] At the same time, the system retrieves the patient's long-term accumulated historical record of bed-leaving events, which stores all behavioral records generated by the patient within a pre-defined historical period (such as the most recent month).

[0032] S2. Extract the statistical center value of the historical bed-leaving event record set to obtain the habitual baseline time; extract the time domain features of the mattress pressure-sensitive distribution data to separate the static background load and local extreme value load.

[0033] Statistical processing techniques, such as mean or median calculations, are performed on the retrieved historical out-of-bed event records to extract the geometric center of the historical time distribution, thus obtaining the statistical center value. This value represents the average time a patient spends using the toilet and getting out of bed under normal physiological conditions, and is defined as the habitual baseline time. The habitual baseline time reflects the standard time taken for the patient to get out of bed and return when there is no significant acute pain interference or morning stiffness.

[0034] Subsequently, signal decomposition was performed on the acquired mattress pressure-sensitive distribution data. Envelope detection or moving average filtering techniques from signal processing were used to extract time-domain features, separating the relatively stable baseline component (static background load) and the peak component (local extreme load) reflecting momentary violent struggle from the time-varying pressure signal. The static background load reflects the basic gravity distribution of various parts of the patient's body acting on the mattress in a static state, while the local extreme load reflects the explosive support pressure generated after the patient gets up.

[0035] S3. Based on the static background load, filter out the resting weight deviation from the local extreme load, and calculate the absolute struggle intensity and relative intensity that characterize the physical turning amplitude.

[0036] The local extreme load is calibrated using the static background load. By performing differential processing, static interference caused by the body's own weight as a constant reference is eliminated, thus filtering out resting weight deviation. Based on this, through linear mapping and proportional calculation, the absolute struggle intensity and relative intensity, quantifying the degree of tossing and turning, are calculated. The absolute struggle intensity reflects the absolute physical strength of the force exerted during the movement, while the relative intensity reflects the degree of fluctuation in movement relative to the background weight.

[0037] S4. Map the fluctuation bandwidth based on the physical residence extreme value characteristics in the historical bed exit event record set, and use the fluctuation bandwidth to perform gain adjustment on the relative intensity to obtain the hysteresis time compensation parameter.

[0038] By analyzing the distribution breadth of historical dwell time in the historical bed departure event record set, and extracting the maximum and minimum historical dwell time, the extreme value characteristics of physical dwell time reflecting the degree of behavioral dispersion are obtained. The time-domain difference between the maximum and minimum values ​​is defined as the fluctuation bandwidth.

[0039] Subsequently, the fluctuation bandwidth is used as a gain adjustment coefficient to perform weight amplification on the current relative intensity, i.e., gain adjustment. The final calculated predicted value in the corresponding time dimension is the hysteresis time compensation parameter. The hysteresis time compensation parameter reflects the additional grace period required to grant the patient due to the current movement impairment.

[0040] S5. Based on the absolute struggle intensity, hysteresis time compensation parameter and habitual reference time, perform adaptive tolerance reconstruction to generate impedance matching delay threshold.

[0041] The steady-state time representing habit, the compensation time representing current physiological fluctuations, and the feature representing the intensity of turning over are substituted into the nonlinear fusion logic. By performing parameter weighting and logic alignment, a dynamic trigger boundary that matches the patient's current actual physical physiological impedance level is constructed. This process is called performing adaptive tolerance reconstruction, and the final generated value is set as the impedance matching delay threshold.

[0042] S6. Record the current time spent away from the bed in real time, and activate the alarm command when the current time spent away from the bed exceeds the impedance matching delay threshold.

[0043] After the patient leaves the mattress, the timing module starts, using clock pulse signals to accumulate values ​​and generate the current time spent away from the bed in real time. The processor then performs a looping logic comparison, determining the value between the real-time updated current time spent away from the bed and the generated impedance matching delay threshold. If the current time spent away from the bed exceeds the impedance matching delay threshold, the system determines that an unexpected risk has occurred and immediately activates an alarm command.

[0044] For example, if a patient recovering from knee osteoarthritis gets out of bed at night, the system will detect high-intensity pressure characteristics if the patient's turning motion before getting out of bed is extremely large due to pain. If the patient's out-of-bed time is highly unstable, such as historical records showing sometimes 5 minutes and sometimes 12 minutes, this large fluctuation bandwidth will further amplify the current pain tolerance, ultimately generating a longer delay threshold. Even if the patient walks unsteadily after getting out of bed, as long as they return within the predicted grace period, the system will remain silent, effectively avoiding false alarms.

[0045] In some embodiments, each data record in the historical bed exit event record set is a bed exit event record, which contains three core time-dimensional features: the first is the moment when the sensor detects that the load has completely disappeared, i.e., the physical load disappearance time stamp; the second is the moment when the sensor detects that the human body returns to the bed surface and the load is restored, i.e., the physical load recovery time stamp; and the third is a parameter reflecting the total time spent on the entire process of a single behavior, calculated by the system by subtracting the physical load disappearance time stamp from the physical load recovery time stamp, i.e., the single physical dwell time.

[0046] Extract the statistical center value of the historical bed exit event record set to obtain the customary baseline time, including:

[0047] Sa21. Get the duration of all single physical stays within a preset number of days.

[0048] Retrieve each record generated within a preset number of days (e.g., the last 14 calendar days) and extract the duration of each single physical stay.

[0049] Sa22. Sum all the data on the duration of each individual physical stay to obtain the total time the patient spent on out-of-bed activities during the observation period, i.e., the total historical stay duration.

[0050] Sa23. Divide the total historical dwell time by the total number of single physical dwell time data to obtain the baseline time reflecting the patient's basic behavioral ability habits.

[0051] For example, if a patient's toilet time has mostly been around 600 seconds in the past two weeks, but on one occasion they stayed for 3600 seconds due to extreme discomfort, the baseline time calculated using the mean logic will still be anchored within a reasonable range close to 600 seconds. This ensures that the system can maintain accurate behavioral baseline recognition even when faced with a single sudden abnormality.

[0052] In some embodiments, temporal feature extraction is performed on the mattress pressure-sensitive distribution data to separate static background load and local extreme value load, including:

[0053] Sb21. Divide the preset timing sequence into multiple consecutive time sub-windows.

[0054] Using the existing clock synchronization signal, the long-term pressure signal sequence acquired during the pre-discharge observation period is divided into several segments according to a fixed time span (e.g., one segment every 5 seconds), and each segment is defined as a time sub-window.

[0055] Sb22. Obtain the mattress pressure sensitivity distribution matrix within each of the aforementioned time sub-windows.

[0056] Sb23. In the mattress pressure-sensitive distribution matrix, extract adjacent data points whose readings are greater than the preset effective contact threshold, and merge them into multiple independent force-connected domains; wherein, the effective contact threshold is specifically the maximum background noise in the unloaded pressure-sensitive distribution matrix.

[0057] Specifically, when the bed is vacant, multiple system samples are taken to obtain the unloaded pressure-sensitive distribution matrix. The maximum fluctuation value generated by the sensor array hardware circuit is extracted by traversal and defined as the maximum background noise. Only pressure data points greater than this maximum background noise are retained. Subsequently, the spatially adjacent effective force data points are clustered using a connected component labeling algorithm to form multiple independent force connected components that reflect the range of human body force.

[0058] Sb24. Calculate the spatial coverage area of ​​each of the independent force-bearing connected domains. Specifically, the actual spatial coverage area can be calculated by counting the total number of effective sensing nodes contained in each of the independent force-bearing connected domains.

[0059] Sb25. Remove the target connected regions whose spatial coverage area is less than the preset limb support threshold to obtain the remaining connected regions; wherein, the limb support threshold is specifically the arithmetic mean of the maximum and minimum spatial coverage areas within the time sub-window.

[0060] The arithmetic mean of the maximum and minimum areas of all independent force-connected regions within the current time sub-window is dynamically calculated and used as the boundary for determining the core torso support location, i.e., set as the limb support threshold.

[0061] When patients attempt to get up at night, they typically use their hands, elbows, or other limbs to support themselves on the mattress. While these normal movements generate significant pressure, the contact area is relatively small. If included in calculations without differentiation, they would severely interfere with the assessment of the true intensity of the patient's core torso struggle. The system uses logical discrimination to precisely eliminate these small pressure interference areas created by normal hand or elbow support, retaining only the larger, remaining connected areas created by the hips, shoulders, back, and other major weight-bearing parts of the torso.

[0062] Sb26. Calculate the mean pressure reading of each remaining connected region in all remaining connected regions, and take the minimum value as the static background load characterizing the patient's resting physiological state; take the maximum reading in all remaining connected regions as the local extreme load characterizing the patient's tossing and turning state due to bone and joint pain.

[0063] Static background load reflects the average basic support force exerted by the core trunk on the bed surface when a patient is lying still and preparing to get up. Local extreme load, on the other hand, objectively reflects the explosive, instantaneous maximum pressure generated when a patient violently turns or tosses and turns in bed due to severe bone and joint pain, causing the core trunk, such as the shoulders, back, or hip bones, to violently slam or compress against the bed surface. This area threshold-based separation effectively filters out local hand support noise during normal getting up, decoupling the individual's weight background information from the dynamic force information of the core trunk struggling, allowing the extracted mechanical features to accurately reflect the physiological resistance caused by pain.

[0064] For example, if a patient with morning stiffness presses down on the mattress with their palm to gain leverage after getting up, the hand support action has extremely high pressure but a small contact area. This action is accurately eliminated as motion noise through step Sb25. However, when the patient's core trunk, such as the back or pelvis, violently rolls and struggles on the bed due to severe pain, this large area of ​​intense force is preserved and accurately extracted as a local extreme load in step Sb26, thus objectively restoring the true degree of pain and struggle of the patient's core trunk.

[0065] In some embodiments, such as Figure 2 As shown, by filtering out the resting body weight deviation from the local extreme load based on the static background load, the absolute struggle intensity and relative intensity characterizing the physical turning amplitude are calculated, including:

[0066] S31. Calculate the difference between the local extreme load and the static background load to obtain the pure-state fluctuation load. The pure-state fluctuation load eliminates the static reading background generated by the inherent weight of the human core trunk and can reflect the dynamic struggle force of the trunk that the patient actively exerts on the mattress due to bone and joint pain, exceeding their static weight.

[0067] S32. Calculate the arithmetic mean of the local extreme load and the static background load to obtain the dynamic average force. The dynamic average force reflects the comprehensive average pressure intensity generated by the contact between the core of the body and the mattress during one tossing and turning movement cycle, reflecting the sustained level of trunk force exertion.

[0068] S33. Based on the pure-state fluctuation load, the dynamic force mean and the static background load, perform signal logarithmic amplification and load normalization processing to generate the vital sign drift compensation amount.

[0069] By performing logarithmic operations and amplitude standardization, the system sensing deviation caused by the large shift of the force center of the torso is corrected, thereby outputting the vital sign drift compensation amount.

[0070] S34. Calculate the ratio of the pure-state fluctuation load to the vital sign drift compensation amount to obtain the relative intensity; calculate the sum of the pure-state fluctuation load and the vital sign drift compensation amount to obtain the absolute struggle intensity amount.

[0071] The vital sign drift compensation plays a corrective role between biomechanical and physical sensor readings. Due to the large center of gravity drift of patients in the recovery period after standing up, the readings generated by the sensors often have non-linear sensing drift. The relative intensity reflects the ratio of the burst frequency of trunk movements to the intensity of the system's compensatory capacity, while the absolute struggle intensity reflects the total mechanical work released during this trunk tossing and turning process.

[0072] For example, if an 80kg patient and a 60kg patient exert the same amount of strain on their torso, the pure-state wave load, by subtracting their respective torso background loads, enables both to have a unified physical scale in evaluating the intensity of their exertion, thus achieving precise quantification of physiological impedance across individuals.

[0073] In some embodiments, based on the pure-state fluctuating load, the dynamic average force, and the static background load, signal logarithmic amplification and load normalization are performed to generate a vital sign drift compensation amount, including:

[0074] S331. Calculate the relative fluctuation parameters based on the pure-state fluctuation load, the average dynamic force, and the static background load. , Represents a relative fluctuation parameter. Represents a pure-state wave load. Represents the average dynamic force. This represents the static background load.

[0075] The natural logarithm operator is used to simulate the nonlinear characteristics of human nerve perception of pain pressure intensity, and the ratio of load variation. Transformed into a more linear computational variable, by introducing the natural logarithm, it can maintain a very high sensitivity to capturing the slight trunk twisting movements caused by early stiffness in patients, thus objectively reflecting the degree of discomfort caused by trunk stiffness in patients.

[0076] S332. Calculate the ratio of static background load to relative fluctuation parameter to obtain the tolerance reference value. The tolerance reference value reflects the physical reference deviation limit that the system can tolerate for sensor response fluctuations under the current weight load.

[0077] S333. Perform nonlinear compensation on the tolerance reference quantity and relative fluctuation parameter to generate a vital sign drift compensation quantity. , This represents the amount of compensation for vital sign drift. Represents the tolerance baseline. This represents a relative fluctuation parameter. When the patient's torso undergoes extremely violent twisting and turning, the generated sign drift compensation will converge to the tolerance baseline, thereby effectively preventing logical errors caused by excessive force at a single moment.

[0078] For example, if a patient struggles violently due to severe joint pain, generating extremely high instantaneous pressure on the torso, without nonlinear compensation, an extremely long compensation time may be calculated. By using a saturation function, the system limits this extreme force within a safe range, ensuring the rationality of the calculation results.

[0079] In some embodiments, the fluctuation bandwidth is mapped based on the physical residence extreme value characteristics in the historical bed exit event record set, and the gain adjustment of the relative intensity is performed using the fluctuation bandwidth to obtain the hysteresis time compensation parameter, including:

[0080] S41. Calculate the temporal difference between the maximum and minimum single physical dwell time in the historical bed departure event record set to obtain the fluctuation bandwidth.

[0081] S42. Calculate the product of the fluctuation bandwidth and the relative intensity to obtain the hysteresis time compensation parameter.

[0082] Fluctuation bandwidth characterizes the temporal unpredictability of a particular patient's behavioral pattern. For patients whose morning stiffness fluctuates daily, the larger the extreme value difference, the more unstable their physiological state. The hysteresis time compensation parameter completes a cross-dimensional mapping from trunk physical force to time delay; it represents the additional time required for the patient to get out of bed due to the current motor impairment.

[0083] In some embodiments, such as Figure 3 As shown, based on the absolute struggle intensity, hysteresis time compensation parameter, and habitual reference time, adaptive tolerance reconstruction is performed to generate an impedance matching delay threshold, including:

[0084] S51. Calculate the sum of the customary reference time and the hysteresis time compensation parameter to obtain the ideal grace period; calculate the ratio of the customary reference time to the hysteresis time compensation parameter to obtain the time proportionality coefficient; normalize the absolute struggle intensity based on the static background load to obtain the standard struggle intensity. , Represents standard struggle intensity. Represents the absolute intensity of the struggle. This represents the static background load.

[0085] The ideal grace period represents the upper limit of the predicted time boundary for getting out of bed, assuming complete trust in the patient's physiological responses. The time proportion factor reflects the proportion of the pain increment time relative to the habitual baseline time, indicating the severity of the deviation of the current out-of-bed behavior from the norm. The standard struggle intensity measures the trunk struggle energy density per unit weight support, used to eliminate the influence of differences in patient body size on risk assessment.

[0086] S52. Based on the ideal grace period, time ratio coefficient and standard struggle intensity, perform risk convergence feature mapping to generate an adaptive loss period that reflects safety concerns.

[0087] S53. Calculate the difference between the ideal grace period and the adaptive loss period to obtain the impedance matching delay threshold.

[0088] For example, if it is calculated that 15 minutes should be waited based on comfort level, and then the current severity of trunk rollover is assessed as indicating a high risk of fall, the duration of the fall is calculated to be 4 minutes through mapping logic, and the final output impedance matching delay threshold is 11 minutes.

[0089] In some embodiments, risk convergence feature mapping is performed based on the ideal total grace period, time proportion coefficient, and standard struggle intensity to generate an adaptive attrition period, including:

[0090] S521. Calculate the composite risk factor based on the aforementioned standard struggle intensity and time proportion coefficient. , This represents a composite risk factor, reflecting the overall probability of an accident caused by deformities in trunk movements. Represents standard struggle intensity. Represents the time scale coefficient.

[0091] By introducing an exponential term The risk of extremely intense trunk struggle is amplified. When the intensity of trunk struggle exceeds the safety threshold, the risk of the patient losing balance explodes exponentially. A composite risk factor is calculated, enabling the system to perceive this nonlinear risk surge.

[0092] S522. Calculate the ratio of the ideal grace period to the composite risk factor to obtain the time-domain convergence benchmark. The time-domain convergence benchmark defines the off-bed correction step size corresponding to a unit risk under a specific safety risk level.

[0093] S523. Based on the aforementioned time-domain convergence benchmark and composite risk factor, perform safety loss calculation to obtain the adaptive loss duration. , Represents the adaptive depreciation time. This represents the time-domain convergence baseline. When the patient's torso exhibits extremely drastic and prolonged vital signs, the calculated loss will increase accordingly. This means that the system will subtract more time from the ideal duration, forcibly shortening the alarm delay and prioritizing alarm response.

[0094] For example, if a patient suddenly rolls over and struggles while getting up from the toilet, the resulting large-area pressure spike will instantly increase the composite risk factor, causing the delay threshold to contract rapidly. The system will immediately trigger an alarm to ensure that caregivers can intervene in a timely manner.

[0095] In some embodiments, such as Figure 4 As shown, this application provides a delayed triggering system for bed exit alarm based on learning patient habitual behaviors, comprising:

[0096] Data acquisition module: In response to the bed-leaving trigger signal, it acquires mattress pressure-sensitive distribution data within a preset time sequence before the bed-leaving trigger signal occurs, and retrieves historical bed-leaving event records.

[0097] Feature extraction module: used to extract the statistical center value of the historical bed leaving event record set to obtain the habitual baseline time; to perform time-domain feature extraction on the mattress pressure-sensitive distribution data, and to separate the static background load and local extreme value load.

[0098] Intensity calculation module: used to filter out resting weight deviation from local extreme loads based on the static background load, and calculate the absolute struggle intensity and relative intensity that characterize the physical turning amplitude.

[0099] Parameter adjustment module: used to map the fluctuation bandwidth according to the physical residence extreme value characteristics in the historical bed leaving event record set, and use the fluctuation bandwidth to perform gain adjustment on the relative intensity to obtain the hysteresis time compensation parameter.

[0100] Threshold generation module: used to perform adaptive tolerance reconstruction based on the absolute struggle intensity, hysteresis time compensation parameter and habitual reference time, and generate impedance matching delay threshold.

[0101] Alarm determination module: used to record the current time spent away from the bed in real time, and to activate an alarm command when the current time spent away from the bed exceeds the impedance matching delay threshold.

[0102] In some embodiments, this application provides a readable storage medium, comprising: computer program instructions stored in the readable storage medium, wherein the computer program instructions are read and executed by a processor to perform the step of a method for delaying the triggering of an out-of-bed alarm based on learning patient habitual behaviors.

[0103] It should be noted that, in this application, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0104] Any references to memory, storage, database, or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory.

[0105] 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; and these 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 delayed triggering of bed-leaving alarm based on learning patient habitual behaviors, characterized in that, include: In response to the bed-leaving trigger signal, the mattress pressure-sensitive distribution data within a preset time sequence before the bed-leaving trigger signal occurs is obtained, and the historical bed-leaving event record set is retrieved; The statistical center value of the historical bed-leaving event record set is extracted to obtain the habitual baseline time; the time-domain features of the mattress pressure-sensitive distribution data are extracted to separate the static background load and local extreme load. Based on the static background load, the resting weight deviation is filtered out from the local extreme load, and the absolute struggle intensity and relative intensity characterizing the physical turning amplitude are calculated. The fluctuation bandwidth is mapped based on the physical residence extreme value characteristics in the historical bed exit event record set, and the gain adjustment of the relative intensity is performed using the fluctuation bandwidth to obtain the hysteresis time compensation parameter. Based on the absolute struggle intensity, hysteresis time compensation parameter, and habitual reference time, an adaptive tolerance reconstruction is performed to generate an impedance matching delay threshold. Record the current time spent away from the bed in real time, and activate an alarm command when the current time spent away from the bed exceeds the impedance matching delay threshold; Temporal feature extraction was performed on the mattress pressure-sensitive distribution data to separate static background load and local extreme load, including: The preset time sequence is divided into multiple consecutive time sub-windows; Obtain the mattress pressure sensitivity distribution matrix within each of the aforementioned time sub-windows; In the mattress pressure-sensitive distribution matrix, adjacent data points with readings greater than a preset effective contact threshold are extracted and merged into multiple independent force-connected domains; wherein, the effective contact threshold is specifically the maximum background noise in the unloaded pressure-sensitive distribution matrix; Calculate the spatial coverage area of ​​each of the independent force-bearing connected domains; Remove the target connected regions whose spatial coverage area is less than the preset limb support threshold to obtain the remaining connected regions; wherein, the limb support threshold is specifically the arithmetic mean of the maximum and minimum spatial coverage areas within the time sub-window. Calculate the average pressure reading for each of the remaining connected regions, and use the minimum value as the static background load characterizing the patient's resting physiological state; use the maximum reading in all remaining connected regions as the local extreme load characterizing the patient's tossing and turning state due to bone and joint pain. The fluctuation bandwidth is mapped based on the physical residence extreme value characteristics in the historical bed exit event record set, and the gain is adjusted for relative intensity using the fluctuation bandwidth to obtain hysteresis time compensation parameters, including: The temporal difference between the maximum and minimum single physical dwell time in the historical bed departure event record set is calculated to obtain the fluctuation bandwidth; The product of the fluctuation bandwidth and the relative intensity is calculated to obtain the hysteresis time compensation parameter; Based on the absolute struggle intensity, hysteresis time compensation parameter, and habitual reference time, an adaptive tolerance reconstruction is performed to generate an impedance matching delay threshold, including: The sum of the customary reference time and the hysteresis compensation parameter is calculated to obtain the ideal grace period; the ratio of the customary reference time to the hysteresis compensation parameter is calculated to obtain the time proportionality coefficient; the absolute struggle intensity is normalized according to the static background load to obtain the standard struggle intensity. , Represents standard struggle intensity. Represents the absolute intensity of the struggle. Represents static background load; Based on the ideal total grace period, time ratio coefficient and standard struggle intensity, risk convergence feature mapping is performed to generate an adaptive loss period; The impedance matching delay threshold is obtained by calculating the difference between the ideal grace period and the adaptive loss period. Based on the ideal grace period, time ratio coefficient, and standard struggle intensity, a risk convergence feature mapping is performed to generate an adaptive attrition period, including: The composite risk factor is calculated based on the aforementioned standard struggle intensity and time ratio coefficient. , Represents a composite risk factor. Represents standard struggle intensity. Represents the time scale factor; Calculate the ratio of the ideal grace period to the composite risk factor to obtain the time-domain convergence benchmark; Based on the aforementioned time-domain convergence benchmark and composite risk factor, a safety loss calculation is performed to obtain the adaptive loss duration. , Represents the adaptive depreciation time. This represents the convergence benchmark in the time domain.

2. A delayed triggering method for bed-leaving alarms based on learning patient habitual behaviors, as described in claim 1, characterized in that... The bed exit event record includes: a physical load disappearance time stamp, a physical load recovery time stamp, and a single physical dwell time calculated by subtracting the physical load disappearance time stamp from the physical load recovery time stamp.

3. A delayed triggering method for bed-leaving alarm based on learning patient habitual behaviors as described in claim 1, characterized in that, Extract the statistical center value of the historical bed exit event record set to obtain the customary baseline time, including: Get the duration of all single physical stays within a preset number of days; The total historical dwell time is obtained by summing up all the data on the duration of each single physical stay. Divide the total historical dwell time by the total number of single physical dwell time data to obtain the customary baseline time.

4. A delayed triggering method for bed-leaving alarm based on learning patient habitual behaviors as described in claim 1, characterized in that, Based on the static background load, resting weight deviation is filtered out from local extreme loads, and the absolute struggle intensity and relative intensity characterizing the physical turning amplitude are calculated, including: The difference between the local extreme load and the static background load is calculated to obtain the pure-state wave load; the arithmetic mean of the local extreme load and the static background load is calculated to obtain the dynamic average force. Based on the pure-state fluctuation load, the dynamic force mean and the static background load, logarithmic signal amplification and load normalization are performed to generate the vital sign drift compensation amount. The ratio of the pure-state fluctuation load to the vital sign drift compensation is calculated to obtain the relative intensity; the sum of the pure-state fluctuation load and the vital sign drift compensation is calculated to obtain the absolute struggle intensity.

5. A delayed triggering method for bed-leaving alarm based on learning patient habitual behaviors as described in claim 4, characterized in that, Based on the pure-state fluctuating load, the dynamic average force, and the static background load, logarithmic signal amplification and load normalization are performed to generate a vital sign drift compensation amount, including: The relative fluctuation parameters are calculated based on the pure-state fluctuation load, the average dynamic force, and the static background load. , Represents a relative fluctuation parameter. Represents a pure-state wave load. Represents the average dynamic force. Represents static background load; Calculate the ratio of static background load to relative fluctuation parameter to obtain the tolerance reference value; Nonlinear compensation is applied to the tolerance reference quantity and the relative fluctuation parameter to generate the vital sign drift compensation quantity. , This represents the amount of compensation for vital sign drift. Represents the tolerance baseline. This represents a relative fluctuation parameter.

6. A delayed-triggered alarm system for learning patient habitual behaviors, characterized in that, include: Data acquisition module: In response to the bed exit trigger signal, it acquires mattress pressure-sensitive distribution data within a preset time sequence before the bed exit trigger signal occurs, and retrieves the historical bed exit event record set; Feature extraction module: used to extract the statistical center value of the historical bed exit event record set to obtain the habitual baseline time; to perform time-domain feature extraction on the mattress pressure-sensitive distribution data, separating the static background load and local extreme value load; Intensity calculation module: used to filter out resting weight deviation from local extreme loads based on the static background load, and calculate the absolute struggle intensity and relative intensity characterizing the physical turning amplitude; Parameter adjustment module: used to map the fluctuation bandwidth according to the physical residence extreme value characteristics in the historical bed leaving event record set, and use the fluctuation bandwidth to perform gain adjustment on the relative intensity to obtain the hysteresis time compensation parameter; Threshold generation module: used to perform adaptive tolerance reconstruction based on the absolute struggle intensity, hysteresis time compensation parameter and habitual reference time, and generate impedance matching delay threshold; Alarm determination module: used to record the current time spent away from the bed in real time, and to activate an alarm command when the current time spent away from the bed exceeds the impedance matching delay threshold; Temporal feature extraction was performed on the mattress pressure-sensitive distribution data to separate static background load and local extreme load, including: The preset time sequence is divided into multiple consecutive time sub-windows; Obtain the mattress pressure sensitivity distribution matrix within each of the aforementioned time sub-windows; In the mattress pressure-sensitive distribution matrix, adjacent data points with readings greater than a preset effective contact threshold are extracted and merged into multiple independent force-connected domains; wherein, the effective contact threshold is specifically the maximum background noise in the unloaded pressure-sensitive distribution matrix; Calculate the spatial coverage area of ​​each of the independent force-bearing connected domains; Remove the target connected regions whose spatial coverage area is less than the preset limb support threshold to obtain the remaining connected regions; wherein, the limb support threshold is specifically the arithmetic mean of the maximum and minimum spatial coverage areas within the time sub-window. Calculate the average pressure reading for each of the remaining connected regions, and use the minimum value as the static background load characterizing the patient's resting physiological state; use the maximum reading in all remaining connected regions as the local extreme load characterizing the patient's tossing and turning state due to bone and joint pain. The fluctuation bandwidth is mapped based on the physical residence extreme value characteristics in the historical bed exit event record set, and the gain is adjusted for relative intensity using the fluctuation bandwidth to obtain hysteresis time compensation parameters, including: The temporal difference between the maximum and minimum single physical dwell time in the historical bed departure event record set is calculated to obtain the fluctuation bandwidth; The product of the fluctuation bandwidth and the relative intensity is calculated to obtain the hysteresis time compensation parameter; Based on the absolute struggle intensity, hysteresis time compensation parameter, and habitual reference time, an adaptive tolerance reconstruction is performed to generate an impedance matching delay threshold, including: The sum of the customary reference time and the hysteresis compensation parameter is calculated to obtain the ideal grace period; the ratio of the customary reference time to the hysteresis compensation parameter is calculated to obtain the time proportionality coefficient; the absolute struggle intensity is normalized according to the static background load to obtain the standard struggle intensity. , Represents standard struggle intensity. Represents the absolute intensity of the struggle. Represents static background load; Based on the ideal total grace period, time ratio coefficient and standard struggle intensity, risk convergence feature mapping is performed to generate an adaptive loss period; The impedance matching delay threshold is obtained by calculating the difference between the ideal grace period and the adaptive loss period. Based on the ideal grace period, time ratio coefficient, and standard struggle intensity, a risk convergence feature mapping is performed to generate an adaptive attrition period, including: The composite risk factor is calculated based on the aforementioned standard struggle intensity and time ratio coefficient. , Represents a composite risk factor. Represents standard struggle intensity. Represents the time scale factor; Calculate the ratio of the ideal grace period to the composite risk factor to obtain the time-domain convergence benchmark; Based on the aforementioned time-domain convergence benchmark and composite risk factor, a safety loss calculation is performed to obtain the adaptive loss duration. , Represents the adaptive depreciation time. This represents the convergence benchmark in the time domain.

Images