Self-service medicine taking device management system for refined anesthesia and poison medicine

Abstract

The invention relates to the technical field of medical management and control, in particular to a fine anesthesia medicine self-service medicine taking device management system which comprises an acquisition unit, a judgment unit, a classification unit, an adjustment unit and a correction unit. According to the method, real-time credible judgment and graded response processing of the operation process are achieved through dynamic collection, structural analysis and multi-dimensional cross verification of the self-service medicine taking behavior of the refined anesthesia medicine, the system decomposes an operation behavior sequence into sub-behavior units with a sequential dependency relationship, a standard sequence model serves as a reference, and the operation behavior sequence is analyzed to be the sub-behavior units with the sequential dependency relationship. The integrity of the operation process is measured by using a matching degree feature formed by time sequence consistency and duration deviation, and the relative deviation between an inventory variation and an expected change value obtained by scanning and recording is used as a matching feature; the problems of low drug management accuracy and slow response speed caused by abnormal operation detection lag and insufficient operation behavior granularity are effectively solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical management, and particularly relates to a self-service medicine taking device management system for precision toxic drugs. BACKGROUND

[0002] With the increasingly strict management requirements of hospital pharmacy departments and closed management wards for precision toxic drugs, the scene of self-service medicine taking is gradually increasing. However, in actual operation, risks such as non-standard operation sequence, abnormal quantity of taking and placing, and personnel violation are still widespread. The traditional method of relying on manual monitoring or single sensor determination is difficult to achieve real-time and fine behavior supervision, especially in a multi-user, multi-time period, and high-frequency operation environment. How to accurately identify potential abnormalities, dynamically assess risks, and ensure drug safety is still a major challenge faced by hospital management.

[0003] Chinese patent application publication No. CN117198460A discloses a narcotic drug management system and method. The system comprises: a drug cabinet for placing narcotic drugs and registering the narcotic drugs; a verification module for verifying the identity information and medicine taking information of personnel; a medicine taking detection module for detecting whether the narcotic drugs in the drug cabinet are taken out; a drug cabinet lock control module for opening the drug cabinet; a surgical kit for transferring the narcotic drugs to an operating room; a kit tracking module for collecting path data of the surgical kit, denoted as real-time path data; a kit lock control module for opening the surgical kit; an alarm module for abnormal alarm; and a control module electrically connected with the drug cabinet, the verification module, the medicine taking detection module, the drug cabinet lock control module, the kit tracking module, and the kit lock control module. The control module controls the verification module to be always open, and controls the medicine taking detection module, the drug cabinet lock control module, the kit tracking module, the kit lock control module, and the alarm module to be always closed. When the verification module verifies that the identity information and medicine taking information of personnel are both true, the control module controls the drug cabinet lock control module and the medicine taking detection module to be opened. When the medicine taking detection module detects that the narcotic drugs in the drug cabinet do not correspond to the medicine taking information after being taken out, the control module controls the alarm module to perform medicine taking abnormal alarm. When the medicine taking detection module detects that the narcotic drugs in the drug cabinet correspond to the medicine taking information after being taken out, the control module controls the kit tracking module to be opened. When the real-time path data collected by the kit tracking module does not conform to the standard path data, the control module controls the alarm module to perform path abnormal alarm. When the real-time path data collected by the kit tracking module conforms to the standard path data, the control module controls the kit lock control module to open the surgical kit.

[0004] Therefore, the narcotic drug management system has the following problems: the system abnormal alarm triggering delay, which is easy to cause the response after the drug has been taken; the system operation behavior record is limited to event occurrence and path data, the information granularity is rough; the system lacks a grading or dynamic adjustment mechanism for abnormal events, all abnormalities are alarmed in a single way, and the severity cannot be distinguished. SUMMARY

[0005] To this end, the present application provides a precision narcotic drug self-service dispensing device management system, which acquires operation behavior sequence, inventory change and identity authentication information in real time, and overcomes the problems of low drug management accuracy and slow response speed caused by abnormal operation detection lag and insufficient operation behavior granularity in the prior art based on multi-dimensional feature analysis and dynamic threshold adjustment.

[0006] To achieve the above-mentioned purpose, the present application provides a precision narcotic drug self-service dispensing device management system, comprising: An acquisition unit is configured to acquire in real time the operation behavior sequence of a target in the process of taking and dispensing drugs in the precision narcotic drug self-service dispensing device, the identity authentication credential, and the inventory change of the drug; A determination unit is configured to determine whether an abnormal operation event occurs based on the matching degree feature of the logical integrity of the operation behavior sequence and the preset standard sequence model, and the fitting feature of the inventory change and the operation record, in combination with a preset baseline deviation threshold; A classification unit is configured to associate the identity authentication credential based on the abnormal operation event, acquire the historical behavior data of the target, calculate the confidence index of the abnormal operation event based on the deviation degree of the abnormal segment feature of the operation behavior sequence and the historical behavior data, and classify the abnormal operation event into a low-risk abnormal event or a high-risk abnormal event according to the confidence index and a preset risk determination threshold; An adjustment unit is configured to generate a prompt information based on a low-risk abnormal event, and adjust the preset baseline deviation threshold according to the confidence index of all low-risk abnormal events generated within a preset adjustment window, and generate an alarm information based on a high-risk abnormal event and adjust the preset risk determination threshold according to the historical classification feature of the high-risk abnormal event and the confidence index; A correction unit is configured to monitor the distribution feature of the abnormal operation event and the confidence index generated within a preset observation period after adjusting the preset baseline deviation threshold or the preset risk determination threshold, in order to modify the adjusted preset baseline deviation threshold again.

[0007] Further, the determination unit is configured to perform time logic analysis on the operation behavior sequence to decompose it into continuous sub-behaviors of identity authentication, cabinet door opening, medicine code scanning, quantity confirmation, and cabinet door closing, to calculate a matching degree feature of the continuous sub-behaviors and a preset standard sequence model, to calculate a coincidence feature of the inventory change amount and the operation record, to calculate a comprehensive deviation index based on the matching degree feature and the coincidence feature by using a weighted fusion algorithm, and to determine whether an abnormal operation event occurs based on a comparison result of the comprehensive deviation index and a preset baseline deviation threshold.

[0008] Further, the determination unit is configured to calculate a consistency of an occurrence order of each of the continuous sub-behaviors and a preset standard sequence in the preset standard sequence model to obtain a time sequence matching degree, to calculate a deviation degree of a duration of each of the continuous sub-behaviors from a preset standard duration in the preset standard sequence model to obtain a time length matching degree, and to comprehensively obtain the matching degree feature based on the time sequence matching degree and the time length matching degree.

[0009] Further, the determination unit is configured to obtain an expected change value based on scanned medicine information in the operation record, and to calculate a relative deviation of the inventory change amount from the expected change value to obtain the coincidence feature.

[0010] Further, the determination unit is configured to determine that the abnormal operation event occurs when the comprehensive deviation index is greater than the preset baseline deviation threshold.

[0011] Further, the classification unit is configured to extract the continuous sub-behaviors in the operation behavior sequence that are inconsistent with the preset standard sequence model to form an abnormal fragment, to retrieve historical behavior data of the target in a preset historical period based on the identity authentication credential, and to calculate a Euclidean distance between a behavior mode vector of the abnormal fragment and a behavior mode vector of an operation mode recorded in the historical behavior data to obtain the deviation degree.

[0012] Further, the classification unit is configured to fuse the deviation degree and the comprehensive deviation index of the abnormal operation event by weighted summation to calculate the confidence index, and to classify the abnormal operation event as the low-risk abnormal event when the confidence index is less than a preset risk determination threshold, and to classify the abnormal operation event as the high-risk abnormal event when the confidence index is greater than or equal to the preset risk determination threshold.

[0013] Further, the adjustment unit is configured to generate prompt information containing the abnormal segment feature and recommended operation specification based on the low-risk abnormal event, and display the prompt information through a human-computer interaction interface, and calculate a change slope and a fluctuation amplitude of a sequence formed by all the confidence indicators in the preset adjustment window, and when the change slope is positive and the fluctuation amplitude is less than a preset fluctuation threshold, adjust the preset baseline deviation threshold downward based on the change slope.

[0014] Further, the adjustment unit is configured to generate alarm information containing the identity of the target and abnormal details based on the high-risk abnormal event, and perform abnormal density analysis on a sequence of confidence indicators of all high-risk abnormal events generated in a preset historical adjustment period to calculate a frequency of high-risk abnormal events per unit time, and when the frequency is greater than a preset frequency threshold, adjust the preset risk determination threshold upward based on the frequency.

[0015] Further, the adjustment unit is configured to extract a mean value drift and a variance change rate of the confidence indicators in the preset observation period, and when the mean value drift is greater than a preset drift threshold and the variance change rate is less than a preset variance change threshold, correct the adjusted preset baseline deviation threshold based on a direction and amplitude of the mean value drift.

[0016] Compared with the prior art, the beneficial effects of the present application are that, by dynamically collecting, structuring analyzing and multi-dimensionally cross-checking the self-service drug taking behavior of precision toxic drugs, real-time reliable judgment and hierarchical response processing of the operation process are realized. The system decomposes the operation behavior sequence into sub-behavior units with sequential dependency, uses a standard sequence model as a reference, and uses the matching degree features composed of time sequence consistency and time length deviation to measure the integrity of the operation process. At the same time, the relative deviation between the inventory change and the expected change value obtained from the scanning record is used as the fitting feature, so that a verifiable closed-loop association is formed between the behavior logic and the material flow data. The comprehensive deviation index output by the judgment unit can accurately reflect the standard degree of process execution and provide a quantitative basis for anomaly detection. Further, the classification unit compares the deviation degree of the abnormal segment and the individual historical behavior pattern to generate a confidence index representing the reliability of the anomaly, so that the risk judgment considers both process deviation and individual feature difference, which helps to reduce false positives and false negatives. The adjustment unit adjusts the related threshold based on the change slope of the confidence index within the time window and the frequency of abnormal events, so that the system can maintain reasonable sensitivity and stability at different use stages. Finally, the drift and fluctuation of the confidence index distribution are corrected by the correction unit, which can further weaken the impact of environmental changes or usage fluctuations, so that the threshold remains effective for a long time, realizing a multi-dimensional fusion monitoring mechanism centered on operation behavior logic, material change and user historical features, which significantly improves the accuracy, robustness and adaptive ability of anomaly identification while ensuring the safety management of precision toxic drugs, effectively solving the problems of low drug management accuracy and slow response speed caused by lagging anomaly operation detection and insufficient operation behavior granularity.

[0017] Further, by decomposing the operation behavior sequence into continuous sub-behaviors that can reflect the real steps of the drug taking process, calculating the matching degree features of each sub-behavior with the standard sequence and the fitting features between the inventory change and the operation record, and then fusing the two types of features into a comprehensive deviation index in a weighted manner, the system can capture changes in both "whether the process is executed correctly" and "whether the quantity is consistent". Since behavior logic deviation and inventory anomaly essentially represent time sequence instability and quantity balance disruption respectively, their contributions to the comprehensive deviation index have clear quantifiable manifestations, which can form a stable abnormal judgment basis combined with the preset baseline deviation threshold, improve the comprehensive identification ability of hidden irregular drug taking, step-skipping operation, quantity inconsistency and other abnormal events, and make the anomaly monitoring more reliable, sensitive and interpretable.

[0018] Further, by decomposing the operation behavior into two dimensions of quantifiable occurrence sequence and duration, and respectively calculating the time sequence matching degree and the duration matching degree with the standard sequence, and then integrating the two indicators, the system can capture whether the process steps are executed in the correct order, and whether each step has subtle changes such as abnormal delay, skipping or acceleration. Since the sequence stability and time stability of the operation process respectively reflect the characteristics of user action logic and behavior rhythm, both often present synchronous deviation in abnormal cases, therefore, joint analysis of the two types of matching degrees can build a more sensitive and robust behavior deviation representation, thereby significantly improving the recognition accuracy of abnormal behaviors such as non-standard drug taking, abnormal staying, and quickly bypassing key steps, making the drug taking process monitoring more reliable and easy to explain.

[0019] Further, by quantifying the operation behavior logic and the time sequence matching degree and the inventory change matching degree as matching degree features and matching features respectively, and giving appropriate weights for weighted fusion, the system can simultaneously focus on the two key dimensions of process execution standardization and drug quantity consistency. The matching degree feature has a higher weight, so that the system pays more attention to the deviation of operation sequence and duration when determining abnormalities, while the matching feature weight ensures that inventory change abnormalities can be effectively captured, thereby achieving sensitive monitoring of skipping operations, too fast or too slow operations, and quantity abnormalities. The comprehensive deviation index formed by the weighted fusion of the two types of indicators uniformly maps the behavior deviation in the time dimension and the inventory deviation in the quantity dimension into quantifiable abnormality determination standards, so that the system can accurately identify hidden irregular drug taking or abnormal operation events while maintaining fault tolerance for normal operations, improving the reliability, explainability and response timeliness of monitoring.

[0020] Further, by comparing the comprehensive deviation index formed by the fusion of the matching degree features and the matching features with the preset baseline deviation threshold, the system can make a clear determination of operation behavior abnormalities. When the comprehensive deviation index exceeds the threshold, it means that the operation sequence or duration deviates significantly from the standard process, or the inventory change is inconsistent with the operation record, and the system identifies it as an abnormal operation event in time; while the index does not exceed the threshold, it is determined as normal operation, thereby effectively distinguishing between normal fluctuations and real abnormalities, using the quantitative relationship between behavior logic deviation and inventory deviation, so that the abnormality determination can reflect both time sequence stability and quantity consistency, ensuring sensitive capture of hidden irregular drug taking, skipping operations and quantity abnormalities, while avoiding misjudgment of normal operations, improving the accuracy, reliability and explainability of system monitoring.

[0021] Further, by extracting the continuous sub-behaviors in the operation behavior sequence that do not conform to the standard process to form an abnormal segment, and comparing its behavior pattern with the target historical behavior data, the system can quantify the degree of abnormality of the abnormal behavior. The deviation degree not only reflects the difference between the operation steps and the historical habits, but also comprehensively reflects the influence of operation rhythm, stay time and sequence change on the behavior pattern. In this way, the classification unit can effectively distinguish between occasional minor deviations and real rule violations, achieve sensitive identification of abnormal behavior, and at the same time tolerate individual differences in normal operation, so that the system is accurate and reliable when identifying abnormal events, and can provide quantifiable basis for subsequent risk level judgment.

[0022] Further, by weighting and fusing the abnormal segment, the historical behavior deviation degree and the comprehensive deviation index, the classification unit can generate a confidence index reflecting the severity of the abnormality, and realize quantitative evaluation of the operation abnormality. This method integrates the behavior pattern deviation and the quantity deviation into two independent dimensions, so that the system can consider the influence of operation process abnormality and inventory fluctuation when judging abnormal risk. The comparison of the confidence index with the preset risk threshold further distinguishes abnormal events into low risk or high risk, ensuring that the identification of potential rule violations is sensitive and stable, while avoiding excessive alarms for minor and occasional deviations, thereby realizing fine and interpretable risk management.

[0023] Further, by generating prompt information for low-risk abnormal events and displaying it to the operator, the system realizes immediate intervention on potential operation deviations, and dynamically adjusts the baseline deviation threshold based on the change slope and fluctuation amplitude of the confidence index sequence within the preset adjustment window, so that the threshold can be adjusted appropriately with the cumulative minor deviation of consecutive low-risk events, thereby improving the sensitivity to operation trend changes while ensuring the fault tolerance of the system to occasional abnormalities. By associating the abnormal segment features, confidence index sequence and change amplitude and direction of the threshold adjustment, the prompt information, operation behavior trend and threshold adjustment form a closed-loop feedback, ensuring that the capture and response to minor deviations are timely and stable, thereby enhancing the accuracy and interpretability of the medicine taking process monitoring.

[0024] Further, by performing abnormal density analysis on the confidence index sequence of high-risk abnormal events and calculating the occurrence frequency per unit time, the system can quantify the concentration and trend of abnormal events. When the occurrence frequency exceeds the preset threshold, adjusting the risk determination threshold can enhance the sensitivity of the system to high-risk events, while preventing false positives caused by occasional abnormalities. By combining the severity of individual abnormal events with the cumulative impact of consecutive events, the threshold adjustment can respond to real rule violations in a timely manner without excessive interference with normal operations, thereby achieving dynamic, controllable and reliable management of abnormal behavior.

[0025] Further, by performing mean shift amount and variance change rate analysis on the confidence index sequence of abnormal operation events in the preset observation period, the system can identify the continuous deviation and fluctuation characteristics of the abnormal behavior trend, thereby dynamically correcting the baseline deviation threshold. In the case of significant increase in the mean shift amount and small change in the variance, it indicates that the overall level of abnormal operation is rising and the fluctuation is stable. At this time, the threshold is adjusted according to the shift direction and amplitude, which can effectively improve the sensitivity of the system to long-term behavior changes, while avoiding false judgments caused by accidental fluctuations, realizing continuous adaptive optimization of abnormal event monitoring, so that the threshold adjustment reflects the recent operation trend, and also considers the system stability and responsiveness. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 FIG. 1 is a schematic diagram of the fine toxic medicine self-service dispensing device management system of the present embodiment; Figure 2 FIG. 3 is a determination logic diagram of the determination unit of the present embodiment for determining an abnormal operation event; Figure 3 FIG. 4 is a determination logic diagram of the classification unit of the present embodiment for classification; Figure 4 FIG. 5 is a determination logic diagram of the adjustment unit of the present embodiment for adjusting the preset baseline deviation threshold. DETAILED DESCRIPTION

[0027] In order to make the objects and advantages of the present application clearer, the present application will be further described below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0028] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present application and do not limit the protection scope of the present application.

[0029] Please refer to Figure 1 FIG. 1, which is a schematic diagram of the fine toxic medicine self-service dispensing device management system of the present embodiment. The present embodiment provides a fine toxic medicine self-service dispensing device management system, which comprises: an acquisition unit, which is used to acquire the operation behavior sequence of the target in the process of accessing the medicine in the fine toxic medicine self-service dispensing device, the identity authentication voucher, and the inventory change amount of the medicine in real time; a determination unit, which is connected with the acquisition unit, and is used to determine whether an abnormal operation event occurs based on the matching degree characteristics of the logical integrity of the operation behavior sequence and the preset standard sequence model, and the coincidence characteristics of the inventory change amount and the operation record, in combination with the preset baseline deviation threshold; a classification unit connected with the determination unit and the acquisition unit respectively, configured to associate the identity authentication credential based on the abnormal operation event, acquire the historical behavior data of the target, and calculate the confidence index of the abnormal operation event based on the deviation degree of the abnormal segment feature of the operation behavior sequence and the historical behavior data, and classify the abnormal operation event as a low-risk abnormal event or a high-risk abnormal event according to the confidence index and a preset risk determination threshold; an adjustment unit connected with the determination unit and the classification unit respectively, configured to generate prompt information based on the low-risk abnormal event, and adjust the preset baseline deviation threshold according to the confidence indexes of all low-risk abnormal events generated within a preset adjustment window, and generate alarm information based on the high-risk abnormal event and adjust the preset risk determination threshold according to the historical classification features of the high-risk abnormal event and the confidence index; a correction unit connected with the determination unit and the classification unit respectively, configured to monitor the distribution features of the abnormal operation events and the confidence indexes generated within a preset observation period after adjusting the preset baseline deviation threshold or the preset risk determination threshold, so as to correct the adjusted preset baseline deviation threshold again.

[0030] In the embodiment, the precision medicine self-service dispensing device is an intelligent medicine storage and taking device deployed in the hospital pharmacy, day ward or closed management ward, which is internally provided with special storage bins for precision medicine, addictive medicine and controlled toxic medicine, and integrates an identity recognition module (real name information reading module, face recognition camera), an intelligent weighing / stock detection unit, an electronic lock control execution mechanism and a man-machine interaction interface. The device is used to replace manual counter dispensing to realize controlled self-service medicine taking operation under strict authorization supervision.

[0031] In this embodiment, the target is authorized nursing personnel, and under the premise of limited number and limited amount electronic prescription of medical staff configuration, the medicine taking operation is carried out in front of the self-service device within the specified time period. The system first acquires identity authentication credentials (including real name information associated with the prescription, face recognition features or in-hospital finger vein information) through the acquisition unit, and synchronously records the fine-grained operation behavior sequence of the user in the medicine taking process, including but not limited to: the trigger time of identity verification action, the click path of interface operation, the initiation and confirmation process of authorized opening instruction, the residence time during medicine taking and the time node of cabinet door closing, to form a complete behavior trajectory that can be used for subsequent time sequence analysis. The inventory sensor inside the device records the inventory change of the medicine in real time, including the number of single taking, the remaining number in the warehouse and the corresponding time stamp of inventory change. These parameters together constitute the feature input data of the present application, which is used for the subsequent judgment unit to identify whether the medicine taking behavior is consistent with the prescription authorization, whether there is overage, overnumber or abnormal operation trend. Through the above structure, the acquisition unit in this embodiment can completely capture the key feature parameters of the self-service medicine storage and taking process in the real clinical scene, and provide a reliable data basis for the subsequent abnormal judgment, risk prompt and violation interception.

[0032] The preset standard sequence model is a structured template used to describe the normal operation process of the precision toxic medicine self-service dispensing device, including two core parameters: one is the preset standard sequence, i.e., the sequence of key operation steps arranged according to the standard medicine dispensing process, such as "identity verification → cabinet door opening → medicine code scanning → quantity confirmation → cabinet door closing"; the other is the preset standard time length, i.e., the expected duration range of each key operation step under normal operation conditions. The model depends on the interaction logic, operation complexity and historical typical use data of the device, and can provide the behavior sequence and operation rhythm reference standard for the judgment unit, thereby quantifying the sequence consistency and duration deviation degree of continuous sub-behaviors, and providing comparable basic data for abnormal operation detection; the preset baseline deviation threshold is the allowed range for measuring the deviation degree between the current operation behavior sequence and the standard sequence model, which depends on the interaction complexity of the device, the operation difference of the user and the historical behavior statistical results, and is usually set in the similarity deviation range of 0.05-0.20, and in this embodiment, 0.10 is taken, so as to improve the identification sensitivity of slight abnormalities while maintaining normal use tolerance; the preset adjustment window is a continuous time interval used to statistically analyze low-risk abnormal events within a period of time, which depends on the daily use frequency of the medicine dispensing device and the event distribution stability, and is usually set between 10-60 minutes, and in this embodiment, 30 minutes is set, so as to ensure sufficient sample size while avoiding excessive lag of threshold adjustment; the preset observation period is a monitoring period used to reevaluate the abnormal event distribution after threshold adjustment, which depends on the system response speed, the alarm tolerance and the periodic characteristics of event generation, and is usually set between 5-20 minutes, and in this embodiment, 10 minutes is taken, which can be used to verify whether the adjusted threshold matches the current operation state and timely complete secondary correction.

[0033] By dynamic collection, structured analysis and multi-dimensional cross-verification of self-service drug taking behavior of precision toxic drugs, real-time and reliable judgment and hierarchical response processing of operation process are realized. The system decomposes the operation behavior sequence into sub-behavior units with sequential dependency, and uses the matching degree feature composed of time sequence consistency and time length deviation to measure the integrity of the operation process by taking the standard sequence model as a reference. At the same time, the relative deviation between the inventory change and the expected change value obtained from the scanning record is taken as the fitting feature, so that the behavior logic and the material flow data form a verifiable closed loop association. The comprehensive deviation index output by the judgment unit can accurately reflect the standard degree of process execution, providing a quantitative basis for anomaly detection. Further, the classification unit compares the deviation degree of the abnormal fragment and the individual historical behavior pattern to generate a confidence index representing the reliability of the anomaly, so that the risk judgment considers both process deviation and individual feature difference, which helps to reduce false positives and false negatives; and the adjustment unit adjusts the related threshold based on the change slope of the confidence index in the time window and the frequency of abnormal events, so that the system can maintain reasonable sensitivity and stability at different use stages. Finally, the drift and fluctuation of the confidence index distribution are corrected by the correction unit, which can further weaken the impact of environmental changes or use quantity fluctuations, so that the threshold remains effective for a long time, realizing a multi-dimensional fusion monitoring mechanism based on operation behavior logic, material change and user historical characteristics, which not only ensures the safety management of precision toxic drugs, but also significantly improves the accuracy, robustness and self-adaptive ability of anomaly identification, effectively solving the problems of low drug management accuracy and slow response speed caused by lagging anomaly operation detection and insufficient operation behavior granularity.

[0034] Specifically, the judgment unit is configured to perform sequential logic analysis on the operation behavior sequence to decompose it into continuous sub-behaviors of identity verification, cabinet door opening, drug code scanning, quantity confirmation, and cabinet door closing, calculate the matching degree feature of the continuous sub-behaviors and the preset standard sequence model, calculate the fitting feature of the inventory change and the operation record, and based on the matching degree feature and the fitting feature, calculate a comprehensive deviation index using a weighted fusion algorithm, and determine whether an abnormal operation event occurs based on a comparison result of the comprehensive deviation index and a preset baseline deviation threshold.

[0035] By decomposing the operation behavior sequence into continuous sub-behaviors that can reflect the real steps of the dispensing process, calculating the matching degree characteristics of each sub-behavior with the standard sequence respectively, and calculating the consistency characteristics between the inventory change and the operation record, and then fusing the two types of characteristics into a comprehensive deviation index in a weighted manner, the system can capture the changes in the two key dimensions of "whether the process is executed correctly" and "whether the quantity is consistent" at the same time. Since the behavior logic deviation and the inventory anomaly essentially represent time sequence instability and quantity balance disruption respectively, their contributions to the comprehensive deviation index have clear quantifiable manifestations. Combined with the preset baseline deviation threshold, a stable anomaly judgment basis can be formed, which can maintain fault tolerance for normal operations while improving the comprehensive identification capability of abnormal events such as hidden irregular dispensing, step-skipping operation, and quantity inconsistency, making the anomaly monitoring more reliable, sensitive, and interpretable.

[0036] Specifically, the determination unit is configured to calculate the consistency of the occurrence order of each continuous sub-behavior with a preset standard sequence in the preset standard sequence model to obtain a time sequence matching degree, calculate the deviation degree of the duration of each continuous sub-behavior from a preset standard duration in the preset standard sequence model to obtain a duration matching degree, and integrate the time sequence matching degree and the duration matching degree to obtain the matching degree characteristics.

[0037] In this embodiment, the process of integrating the time sequence matching degree and the duration matching degree to obtain the matching degree characteristics includes weighted summation of the time sequence matching degree and the duration matching degree, wherein the preset time sequence matching weight corresponding to the time sequence matching degree is used to reflect the contribution of operation order to the matching degree characteristics, which depends on the importance of step order in the dispensing process and the operation specification requirement, and is usually set between 0.5-0.8, and is set to 0.6 in this embodiment, which can emphasize the correctness of the order while taking into account the influence of other factors; the preset duration matching weight corresponding to the duration matching degree is used to reflect the contribution of operation duration to the matching degree characteristics, which depends on the allowed operation time fluctuation range and abnormal sensitivity of each step, and is usually set between 0.2-0.5, and is set to 0.4 in this embodiment, which can capture abnormal behavior of fast or delayed operation while avoiding false judgment of normal operation.

[0038] The preset standard duration is the allowed duration range of each key operation step in the normal dispensing process, which depends on the complexity of the operation step, the user's operation habit, and the device response speed, and is usually set between 3-15 seconds, and is set to 5 seconds in this embodiment, which can capture abnormal accelerated or delayed operation while avoiding false judgment of normal operation.

[0039] By decomposing the operation behavior into two dimensions of quantifiable occurrence sequence and duration, and respectively calculating the time sequence matching degree and the duration matching degree with the standard sequence, and then synthesizing the two indicators, the system can capture whether the process steps are executed in the correct order, and whether there are subtle changes such as abnormal delay, step skipping or acceleration for each step. Since the sequence stability and time stability of the operation process respectively reflect the characteristics of user action logic and behavior rhythm, both of which often present synchronous deviation in abnormal cases, joint analysis of the two types of matching degrees can build a more sensitive and robust behavior deviation representation, thereby significantly improving the recognition accuracy of abnormal behaviors such as non-standard medicine taking, abnormal staying, and quickly bypassing key steps, making the medicine taking process monitoring more reliable and easy to interpret.

[0040] Specifically, the determination unit is configured to obtain an expected change value based on the scanned medicine information in the operation record, and calculate a relative deviation between the inventory change amount and the expected change value to obtain the fitting feature.

[0041] In this embodiment, the fitting feature = | (inventory change amount - expected change value) | / theoretical inventory change amount x 100%, and the process of obtaining the expected change value based on the scanned medicine information in the operation record includes: based on the scanned medicine information in the operation record, analyzing the unique identification code of the medicine and associating the medicine basic database, obtaining the theoretical weight per unit quantity of the product specification, and combining the access quantity in the scanning record to calculate the expected change value.

[0042] Therefore, in this embodiment, based on the matching degree feature and the fitting feature, a weighted fusion algorithm is used to calculate a comprehensive deviation index, wherein the preset matching degree feature weight corresponding to the matching degree feature is used to reflect the contribution of operation behavior logic and time sequence accuracy in the comprehensive deviation index, which depends on the importance of operation sequence and duration in abnormal judgment in the medicine taking process, and is usually set between 0.6-0.8, and is set to 0.7 in this embodiment, which can highlight the process specification while taking into account the influence of inventory change; the preset fitting feature weight corresponding to the fitting feature is used to reflect the contribution of inventory change and operation record consistency in the comprehensive deviation index, which depends on the inventory accuracy requirement and abnormal sensitivity, and is usually set between 0.2-0.4, and is set to 0.3 in this embodiment, which can capture quantity abnormalities while avoiding interference with normal operation.

[0043] By quantifying the operation behavior logic and the time sequence matching degree and the inventory change matching degree as matching degree features and matching features respectively, and giving appropriate weights for weighted fusion, the system can simultaneously focus on the process execution specification and the drug quantity consistency of two key dimensions. The matching degree feature weight is higher, so that the system pays more attention to the deviation of operation sequence and duration when judging the abnormality, and the matching feature weight ensures that the inventory change abnormality can be effectively captured, so as to realize sensitive monitoring of skip operation, too fast or too slow operation and quantity abnormality. The comprehensive deviation index formed by the weighted fusion of the two types of indexes unifies the behavior deviation in the time dimension and the inventory deviation in the quantity dimension into a quantifiable abnormality judgment standard, so that the system can accurately identify hidden irregular drug taking or abnormal operation events while maintaining fault tolerance to normal operation, and improve the reliability, explainability and response timeliness of monitoring.

[0044] Referring to Figure 2 The determination logic diagram for determining that an abnormal operation event occurs is shown in FIG. 6. In this embodiment, the determination unit is configured to determine that the abnormal operation event occurs when the comprehensive deviation index is greater than the preset baseline deviation threshold.

[0045] In this embodiment, the determination unit is further configured to determine that the abnormal operation event does not occur when the comprehensive deviation index is less than or equal to the preset baseline deviation threshold.

[0046] By comparing the comprehensive deviation index formed by the fusion of the matching degree features and the matching features with the preset baseline deviation threshold, the system can make a clear determination of the operation behavior abnormality. When the comprehensive deviation index exceeds the threshold, it means that the operation sequence or duration deviates obviously from the standard process, or the inventory change is inconsistent with the operation record, and the system identifies it as an abnormal operation event in time; and when the index does not exceed the threshold, it is determined as a normal operation, thereby effectively distinguishing between normal fluctuations and real abnormalities, using the quantitative relationship between behavior logic deviation and inventory deviation, so that the abnormality determination can reflect the time sequence stability and quantity consistency at the same time, ensure sensitive capture of hidden irregular drug taking, skip operation and quantity abnormality, and avoid misjudgment of normal operation, thereby improving the accuracy, reliability and explainability of system monitoring.

[0047] Specifically, the classification unit is configured to extract the continuous sub-behaviors in the operation behavior sequence that do not conform to the preset standard sequence model, to constitute an abnormal segment, to retrieve historical behavior data of the target in a preset historical period based on the identity authentication credential, and to calculate the Euclidean distance between the behavior pattern vector of the abnormal segment and the behavior pattern vector of the operation mode recorded in the historical behavior data to obtain the deviation degree.

[0048] It can be understood that the continuous sub-behaviors in the operation behavior sequence that do not conform to the preset standard sequence model in the embodiment refer to comparing the operation steps actually performed by the target in the medicine taking process with the standard medicine taking process preset in the system, and identifying those continuous action paragraphs that deviate from the standard process in sequence, operation type or duration. For example, in actual operation, there are continuous operations such as skipping steps, repeating code scanning, abnormal staying or closing the cabinet door in advance. These continuous deviation steps are extracted to form abnormal segments, which are used for subsequent deviation degree calculation and abnormal event determination.

[0049] The preset historical period is a time range for calling the past operation behavior data of the target, which depends on the risk sensitivity and operation frequency of medicine management, and is usually set between 7 days and 30 days. In the embodiment, it is set to 14 days, which can provide reliable behavior reference and trend reference for abnormal operation determination.

[0050] By extracting the continuous sub-behaviors in the operation behavior sequence that do not conform to the standard process to form abnormal segments, and comparing the behavior mode with the historical behavior data of the target, the deviation degree is calculated. The system can quantify the abnormal degree of abnormal behavior. The deviation degree not only reflects the difference between the operation steps and the historical habits, but also comprehensively reflects the influence of operation rhythm, staying time and sequence change on the behavior mode. In this way, the classification unit can effectively distinguish between occasional slight deviation and real violation operation, realize sensitive identification of abnormal behavior, and at the same time, tolerate individual differences in normal operation, so that the system is accurate and reliable when identifying abnormal events, and can provide quantifiable basis for subsequent risk level judgment.

[0051] Please refer to Figure 3 As shown in FIG. 13, which is a determination logic diagram for classification by the classification unit in the embodiment. In the embodiment, the classification unit is configured to fuse the deviation degree and the comprehensive deviation index of the abnormal operation event by weighted summation to calculate the confidence index, and classify the abnormal operation event as the low-risk abnormal event when the confidence index is less than the preset risk determination threshold, and classify the abnormal operation event as the high-risk abnormal event when the confidence index is greater than or equal to the preset risk determination threshold.

[0052] In this embodiment, the preset deviation weight corresponding to the degree of deviation is used to reflect the contribution of operational behavior deviation from historical habits to the confidence index. It depends on the importance of operational process stability and is usually set between 0.4 and 0.6. In this embodiment, it is set to 0.5, which can highlight the current abnormal characteristics while taking into account historical behavior references. The preset comprehensive weight corresponding to the comprehensive deviation index is used to reflect the contribution of operational process matching degree and inventory consistency to the confidence index. It depends on the importance of process standardization and quantity accuracy and is usually set between 0.4 and 0.6. In this embodiment, it is set to 0.5, which can maintain balance and sensitivity when comprehensively assessing the risk of operational anomalies.

[0053] By weighted and fused abnormal segments with historical behavioral deviations and comprehensive deviation indicators, the classification unit can generate a confidence index reflecting the severity of the anomaly, enabling a quantitative assessment of operational anomalies. This method integrates two independent dimensions—behavioral pattern deviation and quantity deviation—allowing the system to consider the impact of operational process anomalies and inventory fluctuations simultaneously when assessing anomaly risks. The comparison of the confidence index with a preset risk threshold further distinguishes between low-risk and high-risk events, ensuring both sensitive and stable identification of potential violations while avoiding over-alarms for minor, sporadic deviations, thus achieving refined and interpretable risk management.

[0054] Please see Figure 4 As shown, this is a logic diagram for adjusting the preset baseline deviation threshold by the adjustment unit in this embodiment. In this embodiment, the adjustment unit is used to generate prompt information containing the abnormal segment characteristics and suggested operation specifications based on the low-risk abnormal event, and display it through the human-computer interaction interface. It also calculates the change slope and fluctuation amplitude of the sequence formed by all the confidence indicators in the preset adjustment window, and adjusts the preset baseline deviation threshold downward based on the change slope when the change slope is positive and the fluctuation amplitude is less than the preset fluctuation threshold.

[0055] In this embodiment, the process of adjusting the preset baseline deviation threshold based on the slope of change can be expressed as follows: Let the preset baseline deviation threshold be Tb, and the sequence formed by all confidence indicators within the preset adjustment window be {C1,C2,…,Cn}, with a slope of change of S=(Cn-C1) / (n-1); when S>0 and the fluctuation amplitude (the difference between the maximum and minimum values ​​in the sequence formed by all confidence indicators within the preset adjustment window) is less than the preset fluctuation threshold, the baseline deviation threshold is updated, Tb'=Tb-α×S, where Tb' is the updated baseline deviation threshold, and α is a preset adjustment coefficient used to control the magnitude of the threshold adjustment, so that the threshold can be appropriately adjusted with the slight increase in the trend of low-risk abnormal events, ensuring the system's fault tolerance for normal operation while improving the sensitivity to trend deviation.

[0056] The preset fluctuation threshold is a parameter for determining whether the fluctuation amplitude of the confidence index sequence is sufficient to prevent threshold adjustment, which depends on the sensitivity of the system to low-risk abnormal events and the natural fluctuation range of the operation behavior, and is usually set between 0.5%-5%, and is set to 2% in the embodiment, which can capture abnormal trend changes while ignoring normal operation fluctuations; the preset down-regulation coefficient is a parameter for controlling the down-regulation amplitude of the baseline deviation threshold, which depends on the sensitivity of the system to abnormal events and the length of the adjustment window, and is usually set between 0.1-0.5, and is set to 0.3 in the embodiment, which can ensure that the threshold adjustment does not excessively affect normal operation and can timely reflect the cumulative trend of low-risk abnormal events.

[0057] By generating prompt information for low-risk abnormal events and displaying it to the operator, immediate intervention on potential operation deviation is realized, and the baseline deviation threshold is dynamically down-regulated based on the change slope and fluctuation amplitude of the confidence index sequence in the preset adjustment window, so that the threshold can be adjusted appropriately with the slight deviation of the cumulative low-risk events, thereby improving the sensitivity to operation trend changes while ensuring the fault tolerance of the system to occasional abnormalities. By associating the abnormal segment features, confidence index sequence, and change amplitude and direction of the threshold adjustment, the prompt information, operation behavior trend, and threshold adjustment form a closed-loop feedback, ensuring that the capture and response to slight deviations are timely and stable, thereby enhancing the accuracy and interpretability of the medication process monitoring.

[0058] Specifically, the adjustment unit is configured to generate alarm information containing the identity of the target and abnormal details based on the high-risk abnormal event, and perform abnormal density analysis on the sequence of confidence indexes of all high-risk abnormal events generated in a preset historical adjustment period to calculate the occurrence frequency of high-risk abnormal events per unit time, and when the occurrence frequency is greater than a preset frequency threshold, the preset risk determination threshold is up-regulated based on the occurrence frequency.

[0059] The preset risk determination threshold is up-regulated based on the occurrence frequency as follows: Tf’ = Tf x (1 + k x |V-V0| / V0), where Tf’ is the preset risk determination threshold after up-regulation, Tf is the preset risk determination threshold before up-regulation, k is a preset up-regulation coefficient, V is the occurrence frequency, and V0 is the preset frequency threshold.

[0060] The preset up-regulation coefficient is used to adjust the sensitivity of the frequency of occurrence to the up-regulation amplitude of the risk determination threshold, which depends on the sensitivity of the system to respond to high-risk abnormal events, and is usually set between 0.1-0.5, and is set to 0.3 in the embodiment, which can quickly respond to high-frequency abnormalities while avoiding excessive amplification of threshold changes; the preset frequency threshold is used to determine whether the occurrence of high-risk abnormal events exceeds the normal fluctuation range, which depends on the frequency of use of the device and the distribution of historical abnormal events, and is usually set between 0.5-2 times / hour, and is set to 1 time / hour in the embodiment, which can capture abnormal aggregation behavior while preventing false positives of occasional events.

[0061] By analyzing the abnormal density of the confidence index sequence of high-risk abnormal events and calculating the occurrence frequency per unit time, the system can quantify the concentration and trend of abnormal events. When the occurrence frequency exceeds the preset threshold, up-regulating the risk determination threshold can enhance the sensitivity of the system to high-risk events, while preventing false positives caused by occasional abnormalities. The severity of individual abnormal events is combined with the cumulative impact of consecutive events, so that threshold adjustment can respond to real violations in a timely manner, and will not cause excessive interference to normal operations, thereby achieving dynamic, controllable and reliable management of abnormal behavior.

[0062] Specifically, the correction unit is configured to extract a mean shift amount and a variance change rate of the confidence index in the preset observation period, and when the mean shift amount is greater than a preset shift threshold and the variance change rate is less than a preset variance change threshold, correct the adjusted preset baseline deviation threshold Tb” based on the direction and amplitude of the mean shift amount, Tb” = Tb’ × [1 + β × (μ - μ0)], where Tb” is the correction result of the corrected adjusted preset baseline deviation threshold, β is a preset correction coefficient, μ is the mean of the confidence index in the preset observation period, μ0 is a preset historical reference value, and (μ - μ0) is the mean shift amount.

[0063] The preset drift threshold is used to determine whether the change of the confidence index mean is significant, which depends on the sensitivity of the system to abnormal judgment and the fluctuation amplitude of historical behavior, and is usually set between 0.01-0.05, and is set to 0.03 in the embodiment, which can capture significant behavior deviation while avoiding misjudgment of normal fluctuations; the preset variance change threshold is used to determine the stability of the confidence index fluctuation, which depends on the natural fluctuation range of the operation behavior and the monitoring period length, and is usually set between 0.01-0.1, and is set to 0.05 in the embodiment, which can maintain the judgment robustness when identifying abnormal concentration or abnormal dispersion trend; the preset correction coefficient is used to control the amplitude of the preset baseline deviation threshold correction by the mean drift, which depends on the sensitivity of the system to threshold dynamic adjustment and fault tolerance demand, and is usually set between 0.1-0.5, and is set to 0.2 in the embodiment, which can achieve balanced adaptive correction, responding to behavior deviation and avoiding excessive adjustment.

[0064] By analyzing the mean drift amount and variance change rate of the confidence index sequence of abnormal operation events in the preset observation period, the system can identify the continuous deviation and fluctuation characteristics of abnormal behavior trend, so as to dynamically correct the baseline deviation threshold. In the case of significant increase in mean drift amount and small variance change, it shows that the overall level of abnormal operation rises and the fluctuation is stable, at this time, the threshold is adjusted according to the drift direction and amplitude, which can effectively improve the sensitivity of the system to long-term behavior change, while avoiding misjudgment caused by accidental fluctuations, realizing continuous adaptive optimization of abnormal event monitoring, making the threshold adjustment reflect the recent operation trend, and taking into account the system stability and responsiveness.

[0065] The above only describes the preferred embodiments of the present application and is not used to limit the present application; for those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A management system for a self-service dispensing device for psychotropic and narcotic drugs, characterized in that, include: The acquisition unit is used to acquire in real time the sequence of the target's operation behavior, identity authentication credentials, and changes in drug inventory during the process of the target retrieving drugs from the self-service drug dispensing device for psychotropic drugs. The determination unit is used to determine whether an abnormal operation event has occurred based on the logical integrity of the operation behavior sequence and the matching degree feature of the preset standard sequence model, as well as the consistency feature of the inventory change and the operation record, combined with the preset baseline deviation threshold. The classification unit is used to obtain the target's historical behavior data based on the abnormal operation event, associate the identity authentication credential, calculate the confidence index of the abnormal operation event based on the deviation of the abnormal segment features of the operation behavior sequence from the historical behavior data, and classify the abnormal operation event into a low-risk abnormal event or a high-risk abnormal event according to the confidence index and a preset risk judgment threshold. An adjustment unit is used to generate prompt information based on low-risk abnormal events, adjust the preset baseline deviation threshold according to the confidence index of all low-risk abnormal events generated within a preset adjustment window, generate alarm information based on high-risk abnormal events, and adjust the preset risk judgment threshold according to the historical classification characteristics of high-risk abnormal events and the confidence index. The correction unit is used to monitor the distribution characteristics of newly generated abnormal operation events and confidence indicators within a preset observation period after adjusting the preset baseline deviation threshold or the preset risk judgment threshold, so as to correct the adjusted preset baseline deviation threshold again.

2. The self-service drug dispensing device management system for psychotropic and narcotic drugs according to claim 1, characterized in that, The determination unit is used to perform temporal logic parsing on the operation behavior sequence to decompose it into continuous sub-behaviors such as identity verification, cabinet door opening, drug barcode scanning, quantity confirmation, and cabinet door closing. It also calculates the matching degree features between the continuous sub-behaviors and a preset standard sequence model, calculates the consistency features between the inventory change and the operation record, calculates a comprehensive deviation index using a weighted fusion algorithm based on the matching degree features and consistency features, and determines whether an abnormal operation event has occurred based on the comparison result between the comprehensive deviation index and the preset baseline deviation threshold.

3. The self-service drug dispensing device management system for psychotropic and narcotic drugs according to claim 2, characterized in that, The determination unit is used to calculate the consistency between the occurrence order of each of the continuous sub-behaviors and the preset standard sequence in the preset standard sequence model to obtain the temporal matching degree, and to calculate the deviation of the duration of each of the continuous sub-behaviors from the preset standard duration in the preset standard sequence model to obtain the duration matching degree, and to combine the temporal matching degree and the duration matching degree to obtain the matching degree feature.

4. The self-service drug dispensing device management system for psychotropic and narcotic drugs according to claim 3, characterized in that, The determination unit is used to obtain the expected change value based on the scanned drug information in the operation record, and to calculate the relative deviation between the inventory change and the expected change value to obtain the matching feature.

5. The self-service drug dispensing device management system for psychotropic and narcotic drugs according to claim 4, characterized in that, The determination unit is used to determine that the abnormal operation event has occurred when the comprehensive deviation index is greater than the preset baseline deviation threshold.

6. The self-service drug dispensing device management system for psychotropic and narcotic drugs according to claim 5, characterized in that, The classification unit is used to extract continuous sub-behaviors in the operation behavior sequence that do not conform to the preset standard sequence model, forming abnormal segments, and, based on the identity authentication credentials, to retrieve the target's historical behavior data within a preset historical period, and to calculate the Euclidean distance between the behavior pattern vector of the abnormal segment and the behavior pattern vector of the operation pattern recorded in the historical behavior data, so as to obtain the degree of deviation.

7. The self-service drug dispensing device management system for psychotropic and narcotic drugs according to claim 6, characterized in that, The classification unit is used to fuse the deviation degree and the comprehensive deviation index of the abnormal operation event by weighted summation to calculate the confidence index, and to classify the abnormal operation event as the low-risk abnormal event when the confidence index is less than the preset risk judgment threshold, and to classify the abnormal operation event as the high-risk abnormal event when the confidence index is greater than or equal to the preset risk judgment threshold.

8. The self-service drug dispensing device management system for psychotropic and narcotic drugs according to claim 7, characterized in that, The adjustment unit is used to generate a prompt message containing the abnormal segment characteristics and suggested operation specifications based on the low-risk abnormal event, and display it through a human-computer interaction interface; calculate the change slope and fluctuation amplitude of the sequence formed by all the confidence indicators within the preset adjustment window; and, when the change slope is positive and the fluctuation amplitude is less than the preset fluctuation threshold, adjust the preset baseline deviation threshold downward based on the change slope.

9. The self-service drug dispensing device management system for psychotropic and narcotic drugs according to claim 8, characterized in that, The adjustment unit is used to generate alarm information containing the identity and anomaly details of the target based on the high-risk anomaly event, and to perform anomaly density analysis on the sequence of confidence indexes of all high-risk anomalies generated within a preset historical adjustment period to calculate the occurrence frequency of high-risk anomalies per unit time, and to increase the preset risk judgment threshold based on the occurrence frequency when the occurrence frequency is greater than a preset frequency threshold.

10. The self-service drug dispensing device management system for psychotropic and narcotic drugs according to claim 9, characterized in that, The correction unit is used to extract the mean drift and variance change rate of the confidence index during the preset observation period, and, when the mean drift is greater than a preset drift threshold and the variance change rate is less than a preset variance change threshold, to correct and adjust the preset baseline deviation threshold based on the direction and magnitude of the mean drift.

Images