Ward visitor mute decision optimization method and system based on Bayesian network

Abstract

The invention discloses a ward visitor mute decision optimization method and system based on a Bayesian network. The method comprises the following steps: S1, generating a complete ward data set; s2, forming a preprocessed ward data set; s3, constructing a patient state Bayesian sub-network based on patient condition data in the preprocessed ward data set; s4, constructing a visitor behavior Bayesian sub-network based on the visitor behavior data in the preprocessed ward data set; s5, constructing an environment noise Bayesian sub-network based on the ward environment data in the preprocessed ward data set; s6, constructing an adaptive hierarchical Bayesian network to form a comprehensive mute decision model; s7, generating mute decision output; and S8, according to the mute decision output, realizing automatic mute control through linkage with a hospital Internet of Things system. According to the invention, the problem of excessive silence or insufficient silence in the prior art is avoided, and the intelligent level of silence decision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of ward environment, and in particular to a method and system for optimizing ward visitor mute decision based on Bayesian network. Background Art

[0002] With the development of intelligent medical technology, the environmental management of hospital wards has gradually introduced intelligent means to improve the hospitalization experience of patients and optimize the allocation of medical staff resources. Ward noise management is an important part of hospital environmental control. A reasonable mute strategy can effectively reduce the interference of environmental noise on the rehabilitation of patients.

[0003] Currently, most hospitals mainly use fixed rules or manual intervention methods for ward mute management. Hospitals usually stipulate unified mute requirements throughout the hospital during specific periods, or medical staff remind visitors to lower their voices through manual inspections.

[0004] In the manual management mode, nurses or other medical staff need to frequently inspect the wards and remind them when they find that the noise level of visitors or in the wards exceeds the standard. However, this method not only increases the workload of medical staff, but also may lead to unstable implementation effects due to the inconsistency of subjective judgments.

[0005] In recent years, some hospitals have introduced intelligent monitoring devices, including environmental noise sensors or intelligent broadcasting systems, to achieve partial automated mute management. However, existing intelligent management systems usually can only provide basic monitoring of noise intensity and cannot make intelligent mute decisions based on the specific conditions of patients and the behavioral characteristics of visitors. Some systems set noise alarms based on fixed thresholds but cannot distinguish the noise tolerance of different patients, nor can they flexibly adjust the mute control strategy.

[0006] In summary, the existing technology has obvious deficiencies in the mute management of ward visitors, which are mainly reflected in the following aspects: First, the mute management method is relatively fixed and difficult to adapt to the personalized needs of different patients; Second, the manual management method relies on medical staff inspections, with low efficiency and subjectivity; Third, although some intelligent devices can monitor environmental noise, they cannot comprehensively analyze patient conditions, visitor behaviors, and environmental factors, lacking precise intelligent decision-making capabilities. Therefore, there is an urgent need for a mute optimization method that can dynamically and adaptively combine patient conditions, visitor behaviors, and ward environments to improve the intelligent level of mute management, reduce manual intervention, and improve the efficiency and accuracy of ward management. Summary of the Invention

[0007] An object of the present invention is to propose a method and system for optimizing ward visitor mute decision based on Bayesian network. The present invention avoids the problems of excessive or insufficient muting in the prior art and improves the intelligent level of mute decision-making.

[0008] A method for optimizing the silent decision of ward visitors based on a Bayesian network according to an embodiment of the present invention includes the following steps:

[0009] S1. Collect ward environment data, patient condition data, and visitor behavior data in real time, and generate a complete ward data set;

[0010] S2. Perform data cleaning, normalization, feature extraction, and abnormal data identification on the complete ward data set to form a preprocessed ward data set;

[0011] S3. Construct a patient status Bayesian sub-network based on the patient condition data in the preprocessed ward data set. The patient status Bayesian sub-network is used to infer the patient's noise sensitivity and comfort information and output patient status indicators;

[0012] S4. Construct a visitor behavior Bayesian sub-network based on the visitor behavior data in the preprocessed ward data set. The visitor behavior Bayesian sub-network is used to evaluate the behavior characteristics of visitors in the ward and their potential interference effects on the silent demand, and output visitor interference indicators;

[0013] S5. Construct an environmental noise Bayesian sub-network based on the ward environment data in the preprocessed ward data set. The environmental noise Bayesian sub-network is used to evaluate the background noise level in the ward and its time-varying trend, and output environmental noise indicators;

[0014] S6. Integrate the patient status indicators, visitor interference indicators, and environmental noise indicators to construct an adaptive hierarchical Bayesian network. The adaptive hierarchical Bayesian network realizes information fusion and interaction between multiple sub-networks according to the hierarchical structure, and forms a comprehensive silent decision model;

[0015] S7. Input the real-time data in the preprocessed ward data set into the comprehensive silent decision model, calculate the silent control trigger probability based on the comprehensive silent decision model, and compare it with the preset silent decision threshold to generate a silent decision output;

[0016] S8. According to the silent decision output, realize automatic silent control by linking with the hospital Internet of Things system, including remotely controlling the visitor's mobile device to enter the silent mode, automatically adjusting the ward background volume, and sending a silent reminder notification.

[0017] Optionally, the step S1 includes:

[0018] S11. Collect the noise level, noise spectrum characteristics, and noise time-varying trend collected by the noise sensors inside and outside the ward, and construct the ward environment data D e ;

[0019] S12. Collect the current health status, physiological monitoring data, and noise sensitivity of the patient, and construct the patient's condition data D p ;

[0020] S13. Collect the identity, stay time, conversation volume, and behavior characteristics of the visitor. The behavior characteristics include whether to stand and talk for a long time and whether to move in the ward, and construct the visitor behavior data D v ;

[0021] S14. Integrate the ward environment data, patient condition data, and visitor behavior data collected in S11 to S13 to form a complete ward data set D total :

[0022] D total =D e ∪D p ∪D v .

[0023] Optionally, the step S2 includes:

[0024] S21. Clean the data of the complete ward data set, remove incomplete, duplicate, or abnormal data points, and obtain the cleaned ward data set;

[0025] S22. Normalize the cleaned ward data set to unify the value ranges of different data types;

[0026] S23. Extract features from the normalized ward data set, extract ward environment features, patient status features, and visitor behavior features. The ward feature data set after feature extraction is represented as:

[0027] D feature ={f e ,f p ,f v};

[0028] Among them, f e represents the ward environment feature, f p represents the patient status feature, f v represents the visitor behavior feature;

[0029] S24. Identify abnormal data in the ward feature data set after feature extraction, detect abnormal patterns of the ward environment, patient status, and visitor behavior. The ward data set after abnormal data identification is represented as:

[0030] D pre ={x∣x∈D feature , satisfying the anomaly detection constraint};

[0031] Among them, the anomaly detection constraint means that the data conforms to the normal distribution range or is determined as non-anomalous data by the anomaly detection model.

[0032] Optionally, the step S3 includes:

[0033] S31. Extract the patient's condition feature data from the preprocessed ward data set and construct the patient status Bayesian sub-network B p , and the joint probability distribution of the patient status Bayesian sub-network is defined as:

[0034]

[0035] Among them, H p represents the current health status of the patient, R p represents the patient's noise sensitivity, C p represents the patient's comfort information, γ1 and γ2 are positive adjustment indices used to adjust the non-linear degree of the conditional probability influence, and Z1 and Z2 are normalization constants respectively;

[0036] S32. The conditional probability distribution of the node C p in the patient status Bayesian sub-network is defined as:

[0037]

[0038] Among them, λ1, λ2, λ3 are weight coefficients reflecting the influence of the patient's current health status and noise sensitivity on the patient's comfort information, δ1, δ2 are non-linear indices capturing the non-linear characteristics of the influence of the patient's current health status and noise sensitivity on the patient's comfort information, and C represents the discrete set of possible values of the patient's comfort;

[0039] S33. Use the patient's condition feature data f p to perform parameter learning on the patient status Bayesian sub-network B p :

[0040]

[0041] Among them, Θ = {P(H p ), P(R p |H p ), λ1, λ2, λ3, δ1, δ2, γ1, γ2} represents the set of all parameters to be learned, and H p,i , R p,i , C p,i are the health status, noise sensitivity, and comfort data of the i-th patient respectively, ρ is a positive regularization coefficient, and R(Θ) is a regularization function;

[0042] S34. Calculate the patient status index I through the Bayesian inference method in the form of the generalized power meanp :

[0043]

[0044] Among them, I p is the patient status indicator, which is used to reflect the overall response of the patient to the mute demand. α and β are weight coefficients, which reflect the relative importance of patient comfort and noise sensitivity. η1 and η2 are positive exponential parameters.

[0045] Optionally, the step S6 includes:

[0046] S61. Obtain the patient status indicator I p , the visitor interference indicator I v , and the environmental noise indicator I e respectively from the patient status Bayesian sub-network, the visitor behavior Bayesian sub-network, and the environmental noise Bayesian sub-network;

[0047] S62. Use a non-linear adaptive fusion function to fuse the patient status indicator, the visitor interference indicator, and the environmental noise indicator to construct a primary fusion indicator I fusion :

[0048]

[0049] Among them, I k represents the corresponding indicator, which is I p when k = p, I v when k = v, and I e when k = e. λ k is a positive fusion exponent, which is used to adjust the degree of non-linear influence of each indicator. α k is the adaptive fusion weight:

[0050]

[0051] Among them, represents the deviation between the current indicator I k and its historical predicted value . φ k is the sensitivity adjustment parameter, which reflects the sensitivity of each indicator to the fusion result;

[0052] S63. Map the primary fusion indicator I fusion to the mute control probability P quiet through an adaptive logic function:

[0053]

[0054] Among them, P quietrepresents the probability of mute triggering, ζ is the slope parameter used to control the response rate of the logic function, and θ is the mute decision threshold;

[0055] S64. Adopt a dynamic weight adjustment mechanism to update the adaptive fusion weight α k as follows:

[0056] w k (t) = (1 - η)·w k (t - 1) + η·α k ;

[0057] where w k (t) represents the weight value at time t, and η is the learning rate that controls the smoothness of weight adjustment;

[0058] S65. Output the mute control probability P calculated in steps S61 to S64 quiet as the final decision result of the comprehensive mute decision model.

[0059] Optionally, step S7 includes:

[0060] S71. Input the preprocessed ward data set into the comprehensive mute decision model:

[0061]

[0062] where is the patient status index at the current time t, is the visitor interference index at the current time t, is the environmental noise index at the current time t;

[0063] S72. Calculate the mute control trigger probability based on the comprehensive mute decision model, and calculate the mute control trigger probability according to the input data at the current time

[0064] S73. Make a mute decision by comparing the mute control trigger probability with the set mute trigger threshold θ s as follows:

[0065]

[0066] where S decision represents the mute decision output;

[0067] When S decision = 1, mute control is triggered; when S decision = 0, mute control is not triggered.

[0068] A ward visitor silent decision optimization system based on Bayesian network, which is used to execute a ward visitor silent decision optimization method based on Bayesian network, includes the following modules:

[0069] A data acquisition module, which is used to obtain ward environment data, patient condition data and visitor behavior data, including the noise level collected by a noise sensor, the physiological state data collected by a patient physiological monitoring device, and the visitor information collected by a visitor identity recognition and behavior monitoring device, and generate a complete ward data set;

[0070] A data preprocessing module, which is used to perform data cleaning, normalization, feature extraction and abnormal data identification on the complete ward data set, and generate a preprocessed ward data set;

[0071] A patient status Bayesian sub-network construction module, which is used to construct a patient status Bayesian sub-network based on the patient condition data in the preprocessed ward data set, infer the patient's noise sensitivity and comfort information, and output patient status indicators;

[0072] A visitor behavior Bayesian sub-network construction module, which is used to construct a visitor behavior Bayesian sub-network based on the visitor behavior data in the preprocessed ward data set, evaluate the behavior characteristics of the visitor in the ward and its potential interference impact on the silent demand, and output a visitor interference indicator;

[0073] An environmental noise Bayesian sub-network construction module, which is used to construct an environmental noise Bayesian sub-network based on the ward environment data in the preprocessed ward data set, evaluate the background noise level in the ward and its time-varying trend, and output an environmental noise indicator;

[0074] An adaptive hierarchical Bayesian network fusion module, which is used to integrate the patient status Bayesian sub-network, the visitor behavior Bayesian sub-network and the environmental noise Bayesian sub-network, construct an adaptive hierarchical Bayesian network, and realize information fusion and interaction between multiple sub-networks according to the hierarchical structure, and form a comprehensive silent decision model;

[0075] A silent control probability calculation module, which is used to input the real-time data in the preprocessed ward data set into the comprehensive silent decision model, calculate the silent control trigger probability based on the comprehensive silent decision model, and compare it with a preset silent decision threshold to generate a silent decision output;

[0076] A silent control execution module, which is used to automatically execute a silent control strategy according to the silent decision output, including remotely controlling the visitor's mobile device to enter the silent mode, automatically adjusting the background volume of the ward, and sending a silent reminder notification, and dynamically optimizing the quietness of the ward environment.

[0077] The beneficial effects of the present invention are:

[0078] (1) The present invention constructs a Bayesian sub-network for patient status, a Bayesian sub-network for visitor behavior, and a Bayesian sub-network for environmental noise, and uses a hierarchical probabilistic inference model to achieve multi-factor fusion for silent decision-making. It can dynamically analyze the physiological status of patients, the behavior patterns of visitors, and the characteristics of the ward environment, comprehensively calculate the silent trigger probability, thereby ensuring the accuracy and personalization of the silent management strategy. Through the adaptive weight adjustment mechanism, the silent strategy can be adjusted according to the patient's condition changes, visitor types, and real-time environmental noise levels, avoiding the problems of excessive silence or insufficient silence in the prior art and improving the intelligent level of silent decision-making.

[0079] (2) The present invention introduces a non-linear adaptive fusion function, constructs a silent control trigger probability calculation model by integrating patient status indicators, visitor interference indicators, and environmental noise indicators, avoiding the problem that the traditional linear weighting method cannot accurately reflect the non-linear relationship between various factors. The fusion function adopts a dynamic weight update mechanism to be able to adjust the influence weights of different indicators in silent decision-making in real time, enabling the silent strategy to be adaptively optimized over time. An adaptive logic function is introduced to map the fusion result to the silent trigger probability, enabling the system to perform accurate reasoning based on real-time data in the face of a complex ward environment, improving the robustness and dynamic adjustment ability of silent decision-making, and avoiding the problem of insufficient adaptability caused by fixed silent rules in the traditional method. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0081] Figure 1 is a flowchart of an optimization method and system for ward visitor silent decision-making based on a Bayesian network proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0082] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0083] Refer to Figure 1 , an optimization method for ward visitor silent decision-making based on a Bayesian network, includes the following steps:

[0084] S1. Real-time collect ward environment data, patient condition data, and visitor behavior data, and generate a complete ward data set;

[0085] S2. Perform data cleaning, normalization, feature extraction, and abnormal data identification on the complete ward data set to form a preprocessed ward data set;

[0086] S3. Construct a patient status Bayesian sub-network based on the patient condition data in the preprocessed ward data set. The patient status Bayesian sub-network is used to infer the patient's noise sensitivity and comfort information and output patient status indicators.

[0087] S4. Construct a visitor behavior Bayesian sub-network based on the visitor behavior data in the preprocessed ward data set. The visitor behavior Bayesian sub-network is used to evaluate the behavior characteristics of visitors in the ward and their potential interference effects on the mute demand, and output visitor interference indicators.

[0088] S5. Construct an environmental noise Bayesian sub-network based on the ward environment data in the preprocessed ward data set. The environmental noise Bayesian sub-network is used to evaluate the background noise level in the ward and its time-varying trend, and output environmental noise indicators.

[0089] S6. Integrate the patient status indicators, visitor interference indicators, and environmental noise indicators to construct an adaptive hierarchical Bayesian network. The adaptive hierarchical Bayesian network realizes information fusion and interaction among multiple sub-networks according to the hierarchical structure, forming a comprehensive mute decision-making model.

[0090] S7. Input the real-time data in the preprocessed ward data set into the comprehensive mute decision-making model, calculate the mute control trigger probability based on the comprehensive mute decision-making model, and compare it with the preset mute decision threshold to generate a mute decision output.

[0091] S8. According to the mute decision output, realize automatic mute control by linking with the hospital Internet of Things system, including remotely controlling the visitor's mobile device to enter the mute mode, automatically adjusting the background volume of the ward, and sending a mute reminder notification.

[0092] In this embodiment, step S1 includes:

[0093] S11. Collect the noise level, noise spectrum characteristics, and noise time-varying trend collected by the noise sensors inside and outside the ward, and construct the ward environment data D e ;

[0094] S12. Collect the patient's current health status, physiological monitoring data, and sensitivity to noise, and construct the patient condition data D p ;

[0095] S13. Collect the identity, stay time, conversation volume, and behavior characteristics of visitors. The behavior characteristics include whether to stand and talk for a long time and whether to move in the ward, and construct the visitor behavior data D v ;

[0096] S14. Integrate the ward environment data, patient condition data, and visitor behavior data collected in S11 to S13 to form a complete ward data set D total:

[0097] D total = D e ∪D p ∪D v 。

[0098] In this embodiment, step S2 includes:

[0099] S21. Clean the complete ward data set, remove incomplete, duplicate or abnormal data points, and obtain the cleaned ward data set;

[0100] S22. Normalize the cleaned ward data set to unify the value ranges of different data types;

[0101] S23. Extract features from the normalized ward data set, extract ward environment features, patient status features and visitor behavior features. The ward feature data set after feature extraction is represented as:

[0102] D feature = {f e , f p , f v};

[0103] Among them, f e represents the ward environment feature, f p represents the patient status feature, f v represents the visitor behavior feature;

[0104] S24. Identify abnormal data in the ward feature data set after feature extraction, detect abnormal patterns of the ward environment, patient status and visitor behavior. The ward data set after abnormal data identification is represented as:

[0105] D pre = {x | x ∈ D feature , satisfying the anomaly detection constraint};

[0106] Among them, the anomaly detection constraint means that the data conforms to the normal distribution range or is determined to be non-abnormal data by the anomaly detection model.

[0107] In this embodiment, step S3 includes:

[0108] S31. Extract the patient's condition feature data from the preprocessed ward data set and construct the patient status Bayesian sub-network B p , and the joint probability distribution of the patient status Bayesian sub-network is defined as:

[0109]

[0110] Among them, H pDenote the current health status of the patient as R p Denote the noise sensitivity of the patient as C p Denote the comfort information of the patient. γ1 and γ2 are positive adjustment indices used to adjust the non - linear degree of the influence of conditional probability, and Z1 and Z2 are normalization constants respectively;

[0111] S32. Node C in the patient status Bayesian sub - network p The conditional probability distribution is defined as:

[0112]

[0113] Where λ1, λ2, λ3 are weight coefficients reflecting the influence of the patient's current health status and noise sensitivity on the patient's comfort information, δ1, δ2 are non - linear indices capturing the non - linear characteristics of the influence of the patient's current health status and noise sensitivity on the patient's comfort information, and C represents the discrete set of possible values of the patient's comfort;

[0114] S33. Use the patient's disease characteristic data f p To perform parameter learning on the patient status Bayesian sub - network B p :

[0115]

[0116] Where Θ = {P(H p ), P(R p |H p ), λ1, λ2, λ3, δ1, δ2, γ1, γ2} represents the set of all parameters to be learned, H p,i , R p,i , C p,i Are the health status, noise sensitivity and comfort data of the i - th patient respectively, ρ is a positive regularization coefficient, and R(Θ) is a regularization function;

[0117] S34. Calculate the patient status index I p By the Bayesian inference method in the form of generalized power mean:

[0118]

[0119] Where I p Is the patient status index used to reflect the overall response of the patient to the demand for quiet, α and β are weight coefficients reflecting the relative importance of the patient's comfort and noise sensitivity, and η1 and η2 are positive exponential parameters.

[0120] In this embodiment, step S6 includes:

[0121] S61. Obtain the patient status indicator I from the patient status Bayesian sub-network, the visitor behavior Bayesian sub-network, and the environmental noise Bayesian sub-network respectively p , the visitor interference indicator I v , and the environmental noise indicator I e ;

[0122] S62. Use a non-linear adaptive fusion function to fuse the patient status indicator, the visitor interference indicator, and the environmental noise indicator to construct the primary fusion indicator I fusion :

[0123]

[0124] wherein, I k represents the corresponding indicator, which is I p when k = p, I v when k = v, and I e when k = e, λ k is a positive fusion exponent used to adjust the degree of non-linear influence of each indicator, and α k is the adaptive fusion weight:

[0125]

[0126] wherein, represents the deviation between the current indicator I k and its historical predicted value , and φ k is the sensitivity adjustment parameter, reflecting the sensitivity of each indicator to the fusion result;

[0127] S63. Map the primary fusion indicator I fusion to the mute control probability P quiet through an adaptive logic function:

[0128]

[0129] wherein, P quiet represents the probability of mute triggering, ζ is the slope parameter used to control the response rate of the logic function, and θ is the mute decision threshold;

[0130] S64. Update the adaptive fusion weight α k using a dynamic weight adjustment mechanism:

[0131] w k (t) = (1 - η)·w k (t - 1)+η·α k ;

[0132] wherein, w k(t) represents the weight value at time t, and η is the learning rate, which controls the smoothness of weight adjustment;

[0133] S65. Output the mute control probability P calculated in steps S61 to S64 quiet as the final decision result of the comprehensive mute decision model.

[0134] In this embodiment, step S7 includes:

[0135] S71. Input the preprocessed ward data set into the comprehensive mute decision model:

[0136]

[0137] where is the patient status index at the current time t, is the visitor interference index at the current time t, is the environmental noise index at the current time t;

[0138] S72. Calculate the mute control trigger probability based on the comprehensive mute decision model, and calculate the mute control trigger probability according to the input data at the current time

[0139] S73. Compare the mute control trigger probability with the set mute trigger threshold θ s for mute decision-making:

[0140]

[0141] where S decision represents the mute decision output;

[0142] When S decision = 1, mute control is triggered; when S decision = 0, mute control is not triggered.

[0143] A ward visitor mute decision optimization system based on a Bayesian network, used to execute a ward visitor mute decision optimization method based on a Bayesian network, includes the following modules:

[0144] The data acquisition module is used to obtain ward environment data, patient condition data, and visitor behavior data, including the noise level collected by a noise sensor, the physiological state data collected by a patient physiological monitoring device, and the visitor information collected by a visitor identity recognition and behavior monitoring device, and generate a complete ward data set;

[0145] The data preprocessing module is used to perform data cleaning, normalization, feature extraction, and abnormal data identification on the complete ward data set, and generate a preprocessed ward data set;

[0146] A patient status Bayesian sub-network construction module, which is used to construct a patient status Bayesian sub-network based on the patient condition data in the preprocessed ward data set, infer the patient's noise sensitivity and comfort information, and output patient status indicators;

[0147] A visitor behavior Bayesian sub-network construction module, which is used to construct a visitor behavior Bayesian sub-network based on the visitor behavior data in the preprocessed ward data set, evaluate the behavior characteristics of visitors in the ward and their potential interference effects on the mute demand, and output visitor interference indicators;

[0148] An environmental noise Bayesian sub-network construction module, which is used to construct an environmental noise Bayesian sub-network based on the ward environmental data in the preprocessed ward data set, evaluate the background noise level in the ward and its time-varying trend, and output environmental noise indicators;

[0149] An adaptive hierarchical Bayesian network fusion module, which is used to integrate the patient status Bayesian sub-network, the visitor behavior Bayesian sub-network and the environmental noise Bayesian sub-network, construct an adaptive hierarchical Bayesian network, and realize information fusion and interaction between multiple sub-networks according to the hierarchical structure to form a comprehensive mute decision-making model;

[0150] A mute control probability calculation module, which is used to input the real-time data in the preprocessed ward data set into the comprehensive mute decision-making model, calculate the mute control trigger probability based on the comprehensive mute decision-making model, and compare it with the preset mute decision threshold to generate a mute decision output;

[0151] A mute control execution module, which is used to automatically execute the mute control strategy according to the mute decision output, including remotely controlling the visitor's mobile device to enter the mute mode, automatically adjusting the background volume of the ward, and sending a mute reminder notification, and dynamically optimizing the quietness of the ward environment.

[0152] Example 1:

[0153] In the actual application test of the present invention, the experiment was carried out in the neurology ward of a third-class first-class hospital in a certain city. The test time was from March 1, 2024 to April 30, 2024. The test objective was to verify the mute optimization effect of the present invention in the actual ward environment, reduce the adverse effects caused by visitor noise on patients, and improve the overall mute management efficiency of the ward.

[0154] In this experiment, 20 wards in the neurology ward were selected. Among them, 10 wards adopted the method of the present invention (experimental group), and 10 wards adopted the traditional manual management method (control group). The area of each ward was 25 square meters, the daily visitor volume of each ward was about 4-6 people, and the average stay duration of each visitor was 25-40 minutes.

[0155] The patients in the ward are mainly stroke recovery patients and patients with postoperative nerve damage. Some patients are sensitive to external noise, and noise interference may affect the patients' sleep quality and recovery process. The experimental process includes four stages: environmental data collection, silent decision execution, noise intervention strategy, and effect evaluation.

[0156] (1) At 19:30 on March 5, 2024, in Ward 203, Patient ID: P203-A1;

[0157] The patient is a 65-year-old male in the fourth week of stroke recovery. He has a lot of daytime rehabilitation training and needs sufficient rest at night. At this time, the patient has fallen asleep at 19:15. The bedside monitoring system detects that his heart rate has dropped to 65 beats per minute, and his breathing rate is stable, entering the light sleep state.

[0158] At 19:30, the patient's family member (Visitor ID: V203-02) enters the ward to visit and talks with another visitor (Visitor ID: V203-03). The sound decibel is 63dB, higher than the patient's set comfortable noise threshold (50dB).

[0159] The system automatically analyzes:

[0160] The ward environmental noise sensor records that the continuous noise level is 63dB, higher than the set threshold;

[0161] The patient's physiological state monitoring shows that the patient has entered the light sleep stage and is sensitive to noise. The silent trigger probability is calculated as 0.87 (higher than the 0.75 silent trigger threshold);

[0162] The visitor behavior analysis system identifies that two visitors are talking at the same time with too high a volume, predicts that the conversation duration will exceed 15 minutes, and the silent strategy needs to be activated.

[0163] The system triggers the silent management measures:

[0164] At 19:32, intelligent voice reminder: The ward audio equipment automatically plays a gentle voice reminder - "Please lower your conversation volume to reduce the impact on the patient."

[0165] At 19:33, the visitors' mobile phones are muted: The system sends a mute command to Visitor IDs: V203-02, V203-03, and automatically adjusts the visitors' mobile phones to the mute mode.

[0166] At 19:35, the ward background sound is optimized: The air conditioner wind speed mode in the ward is automatically adjusted to reduce the equipment operation noise and improve the patient's comfort.

[0167] Subsequent monitoring:

[0168] At 19:40, the conversation volume of the visitor decreased to 50 dB, the noise level in the ward returned below the set threshold, and the patient's heart rate remained stable without sleep interruption due to noise.

[0169] (2) On March 12, 2024, at 14:10, in Ward 305, Patient ID: P305 - B2;

[0170] The patient is a 54 - year - old female in the second week of the recovery period after postoperative nerve injury. She needs appropriate rest during the day, and the doctor recommends reducing strong noise interference.

[0171] At 14:10, the ward noise monitoring system detected that the noise rose to 68 dB. The source of the noise was a visitor (Visitor ID: V305 - 05) in the ward using the mobile phone to play videos with the speaker on, and the duration exceeded 1 minute, which might affect the patient's rest.

[0172] The system automatically analyzed:

[0173] The ward environmental noise sensor recorded a noise of 68 dB, higher than the set threshold (55 dB);

[0174] The patient status monitoring system showed that the patient had fallen asleep at 13:55, with a stable breathing rate and relatively light sleep quality. Noise interference might cause sleep interruption;

[0175] The visitor behavior analysis system identified that the visitor was using the mobile phone with the speaker on and was in a stationary state, and might continue to play videos.

[0176] The system triggered the mute management measure:

[0177] At 14:12, intelligent voice reminder: The system automatically sent a voice reminder to the visitor to turn off the external speaker volume of the mobile phone.

[0178] At 14:13, remote mute execution: The system sent a mute command to the visitor's mobile phone through the hospital Internet of Things to forcefully reduce the external speaker volume of the mobile phone.

[0179] At 14:15, nurse patrol adjustment: The system sent a reminder to the nurse workstation, indicating that there was noise interference in Ward 305, and the nurse went to confirm whether the visitor's behavior continued to be abnormal.

[0180] Subsequent monitoring:

[0181] At 14:20, the noise level in the ward returned to 50 dB. The patient was not awakened by the noise, the visitor had manually reduced the mobile phone volume, and no more high - noise interference occurred.

[0182] Experimental period: From March 1, 2024 to April 30, 2024;

[0183] Comparative experiment: The method of the present invention was used in 10 wards (experimental group), and the traditional manual management method was used in 10 wards (control group).

[0184]

[0185] Experimental data show that the silent management system of the present invention can real-time sense the ward environment, accurately analyze the patient's state, and intelligently intervene in the behavior of visitors. It performs excellently in reducing ward noise and optimizing the patient's rest quality. Compared with the traditional manual management method, the method of the present invention can greatly reduce the nighttime noise level, the number of times patients wake up due to noise, and the workload of nurses' rounds, and significantly improve the hospitalization comfort and satisfaction of patients.

[0186] In practical applications, the method of the present invention can be applied to a variety of ward environments, including those in the neurology department, ICU, and postoperative recovery wards that require precise silent management. Compared with the traditional method, it has stronger intelligence, automation, and adaptability, providing an efficient and accurate solution for hospital ward environment management.

[0187] The present invention constructs a patient state Bayesian sub-network, a visitor behavior Bayesian sub-network, and an environmental noise Bayesian sub-network, and uses a hierarchical probability inference model to realize multi-factor fusion for silent decision-making. It can dynamically analyze the patient's physiological state, visitor behavior patterns, and ward environmental characteristics, comprehensively calculate the silent trigger probability, so as to ensure the accuracy and personalization of the silent management strategy. Through the adaptive weight adjustment mechanism, it can adjust the silent strategy according to the patient's condition changes, visitor types, and real-time environmental noise levels, avoiding the problems of excessive silence or insufficient silence in the prior art and improving the intelligence level of silent decision-making.

[0188] The present invention introduces a non-linear adaptive fusion function, constructs a silent control trigger probability calculation model by integrating patient state indicators, visitor interference indicators, and environmental noise indicators, avoiding the problem that the traditional linear weighting method cannot accurately reflect the non-linear correlation of various factors. The fusion function adopts a dynamic weight update mechanism to be able to adjust the influence weights of different indicators in silent decision-making in real time, so that the silent strategy can be adaptively optimized over time. Introducing an adaptive logic function to map the fusion result to the silent trigger probability enables the system to perform precise reasoning based on real-time data when facing a complex ward environment, improving the robustness and dynamic adjustment ability of silent decision-making and avoiding the problem of insufficient adaptability caused by fixed silent rules in the traditional method.

[0189] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A ward visitor mute decision optimization method based on Bayesian network, characterized in that: The steps include: S1. Collect ward environment data, patient condition data and visitor behavior data in real time, and generate a complete ward data set; S2. Performing data cleaning, normalization, feature extraction, and abnormal data identification on the complete ward data set to form a pre-processed ward data set; S3. Constructing a patient state Bayesian subnetwork based on the patient condition data in the preprocessed ward data set, the patient state Bayesian subnetwork is used to infer the patient's noise sensitivity and comfort information, and output the patient state index; S4. constructing a visitor behavior Bayesian sub-network based on the visitor behavior data in the pre-processed ward data set, the visitor behavior Bayesian sub-network is used to evaluate the behavioral characteristics of visitors in the ward and their potential interference effects on the quiet requirements, and output the visitor interference index; S5. Construct an environmental noise Bayesian subnetwork based on the ward environment data in the preprocessed ward data set, the environmental noise Bayesian subnetwork is used to evaluate the background noise level in the ward and its time-varying trend, and output the environmental noise index; S6. Integrate the patient status index, visitor interference index and environmental noise index to construct an adaptive hierarchical Bayesian network. The adaptive hierarchical Bayesian network realizes information fusion and interaction among multiple sub-networks according to the hierarchical structure to form a comprehensive mute decision model. S7. Input the real-time data in the pre-processed ward data set into the comprehensive mute decision model, calculate the mute control trigger probability based on the comprehensive mute decision model, and compare it with the preset mute decision threshold to generate a mute decision output; S8. Based on the mute decision output, automatic mute control is achieved by linking with the hospital's Internet of Things system, including remotely controlling visitor mobile devices to enter mute mode, automatically adjusting the background volume of the ward, and sending mute reminder notifications.

2. According to the Bayesian network-based ward visitor mute decision optimization method of claim 1, it is characterized in that: The step S1 comprises: S11. Collect the noise level, noise spectrum characteristics and noise time-varying trend collected by noise sensors inside and outside the ward to construct the ward environment data D e ; S12. Collect the patient's current health status, physiological monitoring data and sensitivity to noise, and construct the patient's condition data D p ; S13. Collect the visitor's identity, length of stay, conversation volume and behavioral characteristics, including whether they stand and talk for a long time and whether they move around in the ward, to build visitor behavior data D v ; S14. Integrate the ward environment data, patient condition data and visitor behavior data collected in S11 to S13 to form a complete ward data set D total : D total =D e ∪D p ∪D v 。 3. According to the Bayesian network-based ward visitor mute decision optimization method of claim 1, it is characterized in that: The step S2 comprises: S21. Perform data cleaning on the complete ward data set, remove incomplete, duplicate or abnormal data points, and obtain a cleaned ward data set; S22. Normalize the cleaned ward data set to unify the numerical ranges of different data types; S23. Extract features from the normalized ward data set, extract ward environment features, patient status features and visitor behavior features, and the ward feature data set after feature extraction is expressed as: D feature ={f e ,f p ,f v }; Among them, f e represents the ward environment characteristics, f p represents the patient status characteristics, f v Indicates visitor behavior characteristics; S24. Perform abnormal data identification on the ward feature data set after feature extraction to detect abnormal patterns of ward environment, patient status and visitor behavior. The ward data set after abnormal data identification is expressed as: D pre ={x|x∈D feature ,satisfy anomaly detection constraints}; Among them, anomaly detection constraints refer to data conforming to the normal distribution range, or being determined as non-abnormal data by an anomaly detection model.

4. According to the Bayesian network-based ward visitor mute decision optimization method of claim 1, it is characterized in that: The step S3 comprises: S31. Extract patient condition characteristic data from the pre-processed ward data set and construct the patient status Bayesian subnetwork B p , the joint probability distribution of the patient status Bayesian sub-network is defined as: Among them, H p Indicates the patient’s current health status, R p represents the patient's noise sensitivity, C p represents the patient comfort information, γ1 and γ2 are positive adjustment indexes used to adjust the nonlinear degree of conditional probability influence, and Z1 and Z2 are normalization constants respectively; S32. Node C in the Bayesian subnetwork of patient status p The conditional probability distribution of is defined as: Among them, λ1,λ2,λ3 are weight coefficients, which reflect the influence of the patient's current health status and noise sensitivity on the patient's comfort information; δ1,δ2 are nonlinear indices, which capture the nonlinear characteristics of the influence of the patient's current health status and noise sensitivity on the patient's comfort information; C represents the discrete set of possible values ​​of the patient's comfort; S33. Using patient condition characteristic data p Bayesian subnetwork B for patient status p Perform parameter learning: Where Θ={P(H p ),P(R p ∣H p ),λ1,λ2,λ3,δ1,δ2,γ1,γ2} represents the set of all parameters to be learned, H p,i ,R p,i ,C p,i are the health status, noise sensitivity and comfort data of the i-th patient, ρ is a positive regularization coefficient, and R(Θ) is a regularization function; S34. Calculate the patient status index I using the generalized power mean form through the Bayesian inference method p : Among them, I p is the patient status index, which is used to reflect the patient's overall response to the quiet demand. α and β are weight coefficients, which reflect the relative importance of patient comfort and noise sensitivity. η1 and η2 are positive exponential parameters.

5. According to the Bayesian network-based ward visitor mute decision optimization method of claim 1, it is characterized in that: The step S6 comprises: S61. Obtain patient status index I from the patient status Bayesian sub-network, visitor behavior Bayesian sub-network and environmental noise Bayesian sub-network respectively p , Visitor Interference Index I v And environmental noise index I e ; S62. Using a nonlinear adaptive fusion function to fuse the patient status index, visitor interference index and environmental noise index to construct a primary fusion index I fusion : Among them, I k Represents the corresponding index, when k = p, it is I p , when k=v, it is I v , when k=e, it is I e ,λ k is a positive fusion index, which is used to adjust the degree of nonlinear influence of each indicator. k is the adaptive fusion weight: in, Indicates the current indicator I k Compared with its historical forecast The deviation between k It is the sensitivity adjustment parameter, which reflects the sensitivity of each indicator to the fusion result; S63. The primary fusion index I is converted into fusion Mapped to the mute control probability P quiet : Among them, P quiet represents the probability of mute triggering, ζ is the slope parameter used to control the response rate of the logic function, and θ is the mute decision threshold; S64. Adopt dynamic weight adjustment mechanism to adjust the adaptive fusion weight α k To update: w k (t)=(1-η)·w k (t-1)+η·a k ; Among them, w k (t) represents the weight value at time t, η is the learning rate, which controls the smoothness of weight adjustment; S65. The mute control probability P calculated in steps S61 to S64 is quiet The output is the final decision result of the comprehensive mute decision model.

6. According to the Bayesian network-based ward visitor mute decision optimization method of claim 1, it is characterized in that: The step S7 comprises: S71. Input the pre-processed ward data set into the comprehensive silence decision model: in, is the patient status indicator at the current time t, is the visitor interference index at the current time t, is the environmental noise index at the current time t; S72. Calculate the mute control trigger probability based on the comprehensive mute decision model, and calculate the mute control trigger probability based on the current input data S73. The mute control trigger probability is compared with the set mute trigger threshold θ s Make a silent decision: Among them, S decision Represents the mute decision output; When S decision =1, trigger the mute control. decision When =0, the mute control is not triggered.

7. A ward visitor mute decision optimization system based on a Bayesian network, used to execute a ward visitor mute decision optimization method based on a Bayesian network according to any one of claims 1 to 6, characterized in that: Includes the following modules: The data collection module is used to obtain ward environment data, patient condition data and visitor behavior data, including noise level collected by noise sensors, physiological status data collected by patient physiological monitoring equipment, visitor information collected by visitor identification and behavior monitoring equipment, and generate a complete ward data set; The data preprocessing module is used to perform data cleaning, normalization, feature extraction and abnormal data identification on the complete ward data set, and generate a preprocessed ward data set; A patient status Bayesian sub-network construction module is used to construct a patient status Bayesian sub-network based on the patient condition data in the pre-processed ward data set, to infer the patient's noise sensitivity and comfort information, and output the patient status index; A visitor behavior Bayesian sub-network construction module is used to construct a visitor behavior Bayesian sub-network based on the visitor behavior data in the pre-processed ward data set, evaluate the behavior characteristics of visitors in the ward and their potential interference effects on the quiet requirements, and output a visitor interference index; An environmental noise Bayesian sub-network construction module is used to construct an environmental noise Bayesian sub-network based on the ward environmental data in the pre-processed ward data set, evaluate the background noise level in the ward and its time-varying trend, and output the environmental noise index; The adaptive hierarchical Bayesian network fusion module is used to integrate the patient status Bayesian sub-network, the visitor behavior Bayesian sub-network and the environmental noise Bayesian sub-network to construct an adaptive hierarchical Bayesian network, and realize the information fusion and interaction among multiple sub-networks according to the hierarchical structure to form a comprehensive mute decision model; A mute control probability calculation module is used to input the real-time data in the pre-processed ward data set into the comprehensive mute decision model, calculate the mute control trigger probability based on the comprehensive mute decision model, and compare it with the preset mute decision threshold to generate a mute decision output; The mute control execution module is used to automatically execute the mute control strategy according to the mute decision output, including remotely controlling the visitor's mobile device to enter mute mode, automatically adjusting the background volume of the ward, and sending mute reminder notifications, so as to dynamically optimize the quietness of the ward environment.

Images