Event-based intelligent alarm and voice reminder method for stores

Abstract

This invention relates to the field of intelligent alarms and discloses a store intelligent alarm and voice reminder method based on event classification, comprising the following steps: collecting and preprocessing multi-source data from stores through a full-domain sensing terminal to construct a core data feature library; constructing an event classification scoring model based on the feature library, completing dimension weight initialization and risk quantification classification; dynamically scoring real-time events and determining whether to trigger proactive reminders, generating standardized alarm information; classifying reminder objects according to event levels and collaboratively pushing text and voice reminders; simultaneously implementing a duplicate alarm suppression and timeout escalation mechanism to avoid alarm proliferation and response delays; and finally iteratively optimizing the classification model based on the handling data. This invention achieves accurate event classification, differentiated proactive reminder push, and intelligent linkage between voice and text, effectively solving the problems of frequent false alarms, chaotic reminders, untimely responses, and lack of closed-loop optimization in traditional store alarms, thereby improving alarm effectiveness and handling efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent alarms, and in particular to a store intelligent alarm and voice reminder method based on event classification. Background Technology

[0002] Currently, chain stores and independent stores generally use various terminals such as sensors, video surveillance, and POS systems for security and operational monitoring. Alarm methods are mostly based on fixed threshold triggers, simple audio-visual prompts, or uniform voice broadcasts, without fine-tuning based on the severity of the event, its scope of impact, or its business importance. Most systems only collect abnormal signals and passively alert, relying on single device values ​​for event judgment, lacking multi-source information fusion and comprehensive risk assessment. Voice alert content, volume, broadcast area, and frequency are all fixed, unable to adapt to the on-site environment and the urgency of the event.

[0003] Existing technologies have significant shortcomings: First, alarm events lack scientific classification, and high-risk emergencies are easily overwhelmed by a large number of low-priority alerts, leading to delayed responses. Second, the false alarm rate is high, and the alarm logic is rigid, with repeated broadcasts of the same event causing staff to become desensitized and ignore the alarm. Third, voice alerts lack a priority scheduling mechanism, resulting in broadcast conflicts and information confusion when multiple events occur concurrently, failing to ensure the priority delivery of critical information. These problems directly lead to untimely handling of high-risk events such as fires, intrusions, and major equipment malfunctions in stores, while low-value alarms interfere with normal operations, reducing overall management efficiency and security reliability.

[0004] Therefore, a patent needs to be proposed: a store intelligent alarm and voice reminder method based on event classification, which can accurately quantify and differentiate event risks, ensure that high-risk events are given priority alarms and reach the target audience quickly; reduce false alarms and avoid broadcast confusion through dynamic correction and conflict resolution; and adaptively adjust the voice content, area and volume according to the level to improve the effectiveness of alarms and the efficiency of personnel response, thereby ensuring store safety and operational stability. Summary of the Invention

[0005] This invention overcomes the shortcomings of existing technologies and provides a method for intelligent store alarms and voice reminders based on event classification.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides a method for intelligent store alarms and voice reminders based on event hierarchy, comprising the following steps: By using the store's omni-channel sensing terminals, multi-source sensing data is collected from the store, and the collected data is preprocessed to build a core data feature library. By combining the core data feature library, weight initialization processing and store event classification and scoring model construction are carried out in the terminal control system. Based on the store event classification and scoring model, dynamic classification calculation and proactive reminder trigger determination are performed on real-time events, and standardized alarm basic information is generated. Based on standardized alarm information, the terminal control system allocates reminder permissions and classifies target stores for reminders, and coordinates the push of text and voice reminder information. Continuously push text and voice reminders, and manage and handle repeated alarms and timeout escalation. After handling repeated alarm suppression and timeout escalation management, the store event classification and scoring model is optimized by combining historical data networks.

[0007] Furthermore, in a preferred embodiment of the present invention, the step of collecting multi-source sensing data of the store through the store's omni-channel sensing terminal, and preprocessing the collected data to construct a core data feature library, specifically includes: Identify stores, designate them as target stores, and determine the terminal control system within the target stores; The terminal control system controls all types of sensing terminals in the target store, including security sensing terminals, operational sensing terminals, and environmental and abnormal sound sensing terminals. The terminal control system controls all types of sensing terminals to collect corresponding types of data, obtains store sensing terminal data, and filters and removes invalid information from the store sensing terminal data. Among them, invalid information in the store sensing terminal data includes the shape of data collected by different types of sensing terminals and redundant data; Meanwhile, after filtering and removing invalid information, the data from the store sensing terminals is standardized, including format unification and feature retrieval, to obtain the feature data of the store sensing terminals. By introducing a historical data network, all core feature types for the target store are retrieved. Based on all core feature types of the target store, core features are extracted from the feature data of the store's sensing terminal and a feature database is established to obtain a core data feature library.

[0008] Furthermore, in a preferred embodiment of the present invention, the step of combining the core data feature library to perform weight initialization processing and constructing a store event grading and scoring model in the terminal control system specifically includes: The core features of the store perception terminal feature data in the core data feature library are summarized into event type, occurrence area, duration, numerical threshold deviation, associated device status, and on-site environment, which serve as the basic data for the store event classification model. The core features of the store perception terminal feature data in the core data feature library are mapped to the five evaluation dimensions of the store event classification model, which are labeled as the degree of security impact, the degree of business impact, the degree of risk diffusion, the degree of continuous accumulation, and the degree of environmental interference. At the same time, retrieve the historical weight percentage of all evaluation dimensions and sort them in reverse order to achieve the initial weight allocation of all evaluation dimensions; By using historical data networks, the frequency, false alarm probability, and response efficiency of historical event types of target stores are retrieved. This information is used to normalize and correct the initial weights of each evaluation dimension until the sum of the weights of all evaluation dimensions is always equal to 1. At the same time, the event comprehensive risk scoring algorithm is retrieved. The event comprehensive risk scoring algorithm quantifies the core features of different evaluation dimensions and then combines them with the normalized initial weights of the evaluation dimensions to construct an event comprehensive risk scoring model. Define the historical events that occurred in the target store, and output the comprehensive risk score value of the historical events through the event comprehensive risk scoring model. This value is used to divide the scoring range into multiple levels and to update the event comprehensive risk scoring model, thus obtaining the store event classification scoring model. The multi-level rating range includes four levels: dangerous, emergency, warning, and normal.

[0009] Furthermore, in a preferred embodiment of the present invention, the step of dynamically calculating the classification and determining the proactive alert trigger based on the store event classification and scoring model, and generating standardized alarm basic information, specifically includes: Through the terminal control system, real-time sensing data from security sensing terminals, operation sensing terminals, and environmental and abnormal sound sensing terminals are collected and transmitted to the core data feature library for real-time updates, generating the core features of the real-time sensing data. The core features of real-time sensing data are imported into the store event classification and scoring model, which outputs the event type and corresponding comprehensive risk score of the real-time events of the target store, and generates the current level of the real-time events. An active reminder mechanism is assigned to the four levels in the multi-level scoring range. Specifically, real-time events at the danger and emergency levels are set to require active reminders. For real-time events at the warning level, the duration is calculated. If the duration exceeds the preset value, it is set to require active reminders; if the duration is not greater than the preset value, it is set to not require active reminders. For normal-level real-time events, no active reminders are required. The core characteristics of the real-time sensing data corresponding to the active reminder mechanism are recorded in the terminal control system. Combined with the current level of the real-time event, standardized alarm basic information is encapsulated and generated.

[0010] Furthermore, in a preferred embodiment of the present invention, the step of allocating reminder permissions and classifying target store reminder objects in the terminal control system based on standardized alarm basic information, and coordinating the push of text reminder information and voice reminder information, specifically includes: Determine the personnel structure of the target store and the regional permissions of different personnel in the target store, enter the data into the terminal control system, obtain the mobile terminals of different personnel in the target store, and connect the mobile terminals of different personnel in the target store to the terminal control system. In the terminal control system, based on standardized alarm information, the target branch reminder objects are defined. The rule for defining the target branch reminder objects is that if the current level of a real-time event is dangerous, then in the terminal control system, the target branch reminder objects are determined to be all personnel. If the current level of a real-time event is emergency, the terminal control system determines the target audience to be personnel within the target store area where the event occurred. The system then detects the response of the corresponding personnel's mobile terminals. If no response is received within a predetermined time, the target audience is expanded to include the personnel with the highest area access in the target store. If the current level of a real-time event is an early warning level, then the terminal control system will determine the target of the alert as the personnel in the target store area where the target event occurred. If the current level of a real-time event is normal, then it is not necessary to determine the recipient of the alert. Based on the classification of target stores and standardized alarm information, reminder instructions are encapsulated and generated. Then, based on the reminder instructions, text reminders and voice reminders are coordinated and pushed.

[0011] Furthermore, in a preferred embodiment of the present invention, the step of coordinating the push of text reminder information and voice reminder information based on the reminder instruction specifically includes: Based on the terminal control system, the system reads the reminder command and, according to the classification of the target store reminder object in the reminder command, synchronously pushes standardized alarm basic information to the mobile terminal of the corresponding personnel. During the synchronous push process, standardized alarm basic information is packaged into text reminders and voice reminders; For text-based reminders, the system directly pushes them to the mobile devices of the corresponding personnel via the terminal control system. After the text reminder is pushed, the terminal control system detects the mobile terminal of the corresponding person. If the mobile terminal of the corresponding person responds to the text reminder, there is no need to push the voice reminder. If there is no response text notification on the mobile terminal of the corresponding personnel, the text notification is classified into target events. If it is a dangerous level event or an emergency level event, the terminal control system is used to control the prompt volume of the corresponding personnel's mobile terminal to be turned up to the maximum and play the voice reminder message in a loop until the corresponding personnel's mobile terminal responds with text notification messages. If it is a warning-level event, the terminal control system will keep the volume of the prompts on the mobile terminals of the corresponding personnel constant and play the voice reminder information in a loop until the mobile terminals of the corresponding personnel all have a response text reminder information.

[0012] Furthermore, in a preferred embodiment of the present invention, the continuous push of text and voice reminder information, and the handling of repeated alarm suppression and timeout escalation management, specifically include: Set up a recurring alarm suppression mechanism and a timeout escalation management mechanism within the terminal control system; The repeated alarm mechanism is as follows: during the process of pushing text reminder information and voice reminder information, a predetermined time window is set. If the same real-time event generates reminder instructions repeatedly within the predetermined time window, a repeated alarm password is generated in the terminal control system, and the push of text reminder information and voice reminder information corresponding to the same real-time event by the terminal control system is suppressed. The timeout escalation management mechanism involves setting timeout thresholds for real-time events of different levels through a terminal control system, including timeout thresholds for warning levels, timeout thresholds for emergency levels, and timeout thresholds for danger levels. Simultaneously, the terminal control system calculates in real time the time it takes for the person receiving the text notification to respond to the text notification on their mobile terminal, and marks it as the text response time. If a real-time event is at the warning level, but the text response time exceeds the warning level timeout threshold, then the notification volume on the corresponding personnel's mobile terminal will be increased to the maximum, and the notification recipients will be extended to the personnel with the highest area permissions in the target store. If a real-time event is classified as urgent, but the text response time exceeds the urgent timeout threshold, the alert will be extended to all personnel in the target store. If a real-time event is classified as dangerous, but the text response time exceeds the dangerous level timeout threshold, then the repeated alarm mechanism should be canceled.

[0013] Furthermore, in a preferred embodiment of the present invention, after the repeated alarm suppression and timeout escalation management processing, the optimization of the store event classification and scoring model by combining historical data networks specifically includes: Based on the terminal control system, relevant data during the process of repetitive alarm suppression and timeout escalation management are recorded in real time and stored in the core data feature library for data updates. Based on the updated core data feature library, the store event classification and scoring model is optimized. The optimization target is to combine historical data network to retrieve relevant data on the historical repeated alarm suppression and timeout escalation management process of the target store, and integrate and adjust the normalized weights of all evaluation dimensions. During the adjustment of the normalized weights of all evaluation dimensions, the weights are checked in real time to ensure that the sum of the weights of all evaluation dimensions is always equal to 1, thus completing the optimization of the store event classification scoring model and obtaining the optimized store event classification scoring model. For the optimized store event grading and scoring model, standardized alarm basic information generation is repeatedly invested.

[0014] This invention addresses the technical deficiencies in the existing technology and offers the following advantages: It collects and preprocesses multi-source data from stores through a comprehensive sensing terminal to construct a core data feature library; based on this feature library, it builds an event grading and scoring model, initializing dimension weights and quantifying risk levels; it dynamically scores real-time events and determines whether proactive alerts are triggered, generating standardized alarm information; it categorizes alert recipients according to event levels and collaboratively pushes text and voice alerts; it simultaneously implements repeated alarm suppression and timeout escalation mechanisms to avoid alarm overload and delayed response; and finally, it iteratively optimizes the grading model based on the processed data. This invention achieves accurate event grading, differentiated proactive alert pushes, and intelligent voice and text linkage, effectively solving the problems of numerous false alarms, confusing alerts, untimely responses, and lack of closed-loop optimization in traditional store alarms, thus improving alarm effectiveness and processing efficiency. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0016] Figure 1 A flowchart of an event-based intelligent alarm and voice reminder method for stores is shown. Figure 2 A flowchart illustrating the method for coordinating push text and voice notifications is provided. Detailed Implementation

[0017] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0018] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0019] Figure 1 A flowchart illustrating an event-based intelligent alarm and voice reminder method for stores is provided, including the following steps: S102: Collect multi-source sensing data from stores through the store's omni-channel sensing terminals, preprocess the collected data, and build a core data feature library. S104: Combining the core data feature library, perform weight initialization processing and construct a store event grading and scoring model in the terminal control system; S106: Based on the store event classification and scoring model, dynamically classify and calculate real-time events and determine the trigger for proactive reminders, and generate standardized alarm basic information; S108: Based on standardized alarm information, the terminal control system allocates reminder permissions and classifies target stores for reminders, and coordinates the push of text reminder information and voice reminder information. S110: Continuously pushes text and voice reminders, and performs repeated alarm suppression and timeout escalation management; S112: After handling repeated alarm suppression and timeout escalation management, optimize the store event classification and scoring model by combining historical data networks.

[0020] Furthermore, in a preferred embodiment of the present invention, the step of collecting multi-source sensing data of the store through the store's omni-channel sensing terminal, and preprocessing the collected data to construct a core data feature library, specifically includes: Identify stores, designate them as target stores, and determine the terminal control system within the target stores; The terminal control system controls all types of sensing terminals in the target store, including security sensing terminals, operational sensing terminals, and environmental and abnormal sound sensing terminals. The terminal control system controls all types of sensing terminals to collect corresponding types of data, obtains store sensing terminal data, and filters and removes invalid information from the store sensing terminal data. Among them, invalid information in the store sensing terminal data includes the shape of data collected by different types of sensing terminals and redundant data; Meanwhile, after filtering and removing invalid information, the data from the store sensing terminals is standardized, including format unification and feature retrieval, to obtain the feature data of the store sensing terminals. By introducing a historical data network, all core feature types for the target store are retrieved. Based on all core feature types of the target store, core features are extracted from the feature data of the store's sensing terminal and a feature database is established to obtain a core data feature library.

[0021] It's important to note that a unified control system enables centralized management of the entire domain's perception capabilities. The sensing terminals cover all scenarios, including security, operations, and environmental sound, achieving comprehensive, blind-spot-free detection of the store. Invalid information is filtered and removed from the store's sensing terminal data to ensure the data input to the model is accurate and reliable, preventing noise from affecting the accuracy of subsequent model building. Finally, time, format, units, and location identifiers are standardized, and effective features are extracted. Six core features—event type, region, duration, deviation, equipment status, and environment—are extracted from the standardized data to form a dedicated feature library, ensuring stronger adaptability for subsequent modeling.

[0022] Furthermore, in a preferred embodiment of the present invention, the step of combining the core data feature library to perform weight initialization processing and constructing a store event grading and scoring model in the terminal control system specifically includes: The core features of the store perception terminal feature data in the core data feature library are summarized into event type, occurrence area, duration, numerical threshold deviation, associated device status, and on-site environment, which serve as the basic data for the store event classification model. The core features of the store perception terminal feature data in the core data feature library are mapped to the five evaluation dimensions of the store event classification model, which are labeled as the degree of security impact, the degree of business impact, the degree of risk diffusion, the degree of continuous accumulation, and the degree of environmental interference. At the same time, retrieve the historical weight percentage of all evaluation dimensions and sort them in reverse order to achieve the initial weight allocation of all evaluation dimensions; By using historical data networks, the frequency, false alarm probability, and response efficiency of historical event types of target stores are retrieved. This information is used to normalize and correct the initial weights of each evaluation dimension until the sum of the weights of all evaluation dimensions is always equal to 1. At the same time, the event comprehensive risk scoring algorithm is retrieved. The event comprehensive risk scoring algorithm quantifies the core features of different evaluation dimensions and then combines them with the normalized initial weights of the evaluation dimensions to construct an event comprehensive risk scoring model. Define the historical events that occurred in the target store, and output the comprehensive risk score value of the historical events through the event comprehensive risk scoring model. This value is used to divide the scoring range into multiple levels and to update the event comprehensive risk scoring model, thus obtaining the store event classification scoring model. The multi-level rating range includes four levels: dangerous, emergency, warning, and normal.

[0023] It should be noted that the six categories of event type, occurrence area, duration, numerical threshold deviation, associated equipment status, and on-site environment form the basic data range for model input. This systematically categorizes scattered multi-source features, providing a clear and standardized input foundation for subsequent dimensional mapping and quantitative scoring, avoiding model calculation bias caused by feature confusion. Correspondingly, the input data is mapped to five evaluation dimensions: safety impact, business impact, risk diffusion, cumulative impact, and environmental interference. This clarifies the model's grading evaluation criteria, used to measure the comprehensive impact of events on stores, avoiding misjudgments due to single-dimensional grading. Historical data networks are used to retrieve the historical weight percentages of the five evaluation dimensions, avoiding blind weight allocation and ensuring data-supported weight distribution, thus improving the rationality of the initial model grading. The advantages are more targeted weight allocation, eliminating the need for manual initial value setting, reducing manual intervention costs, and providing a basis for subsequent weight correction, improving model adaptability. Normalization correction of weights ensures that the sum of all dimension weights is always equal to 1, avoiding grading bias caused by unreasonable weights, such as underestimating the danger level due to excessively low weights for the safety dimension. The comprehensive event risk scoring algorithm quantifies and scores core features across five evaluation dimensions. The algorithm is adapted to store-specific evaluation dimensions and adjusted weights, featuring clear scoring logic, accurate calculations, and rapid output of the comprehensive event risk score. Finally, multi-level scoring intervals are defined and used to update the comprehensive event risk scoring model, resulting in a store-level event scoring model.

[0024] Furthermore, in a preferred embodiment of the present invention, the step of dynamically calculating the classification and determining the proactive alert trigger based on the store event classification and scoring model, and generating standardized alarm basic information, specifically includes: Through the terminal control system, real-time sensing data from security sensing terminals, operation sensing terminals, and environmental and abnormal sound sensing terminals are collected and transmitted to the core data feature library for real-time updates, generating the core features of the real-time sensing data. The core features of real-time sensing data are imported into the store event classification and scoring model, which outputs the event type and corresponding comprehensive risk score of the real-time events of the target store, and generates the current level of the real-time events. An active reminder mechanism is assigned to the four levels in the multi-level scoring range. Specifically, real-time events at the danger and emergency levels are set to require active reminders. For real-time events at the warning level, the duration is calculated. If the duration exceeds the preset value, it is set to require active reminders; if the duration is not greater than the preset value, it is set to not require active reminders. For normal-level real-time events, no active reminders are required. The core characteristics of the real-time sensing data corresponding to the active reminder mechanism are recorded in the terminal control system. Combined with the current level of the real-time event, standardized alarm basic information is encapsulated and generated.

[0025] It should be noted that after real-time updates to the core data feature library and extraction of six core features of real-time events, a comprehensive risk score is calculated using a store event grading and scoring model. This automatically classifies events into one of four levels: dangerous, urgent, warning, or normal. This achieves automatic, objective, and quantitative grading of real-time events, replacing manual judgment. Dangerous / urgent events receive mandatory proactive alerts, while warnings are based on duration, with alerts only issued after a timeout. Normal events receive no alerts; only the core features of the real-time sensing data corresponding to the proactive alert mechanism's output are recorded. Standardized alarm basic information is encapsulated and generated. This standardized alarm basic information provides a unified data foundation for subsequent text / voice push notifications, repetition suppression, timeout escalation, and model iteration.

[0026] Furthermore, in a preferred embodiment of the present invention, the continuous push of text and voice reminder information, and the handling of repeated alarm suppression and timeout escalation management, specifically include: Set up a recurring alarm suppression mechanism and a timeout escalation management mechanism within the terminal control system; The repeated alarm mechanism is as follows: during the process of pushing text reminder information and voice reminder information, a predetermined time window is set. If the same real-time event generates reminder instructions repeatedly within the predetermined time window, a repeated alarm password is generated in the terminal control system, and the push of text reminder information and voice reminder information corresponding to the same real-time event by the terminal control system is suppressed. The timeout escalation management mechanism involves setting timeout thresholds for real-time events of different levels through a terminal control system, including timeout thresholds for warning levels, timeout thresholds for emergency levels, and timeout thresholds for danger levels. Simultaneously, the terminal control system calculates in real time the time it takes for the person receiving the text notification to respond to the text notification on their mobile terminal, and marks it as the text response time. If a real-time event is at the warning level, but the text response time exceeds the warning level timeout threshold, then the notification volume on the corresponding personnel's mobile terminal will be increased to the maximum, and the notification recipients will be extended to the personnel with the highest area permissions in the target store. If a real-time event is classified as urgent, but the text response time exceeds the urgent timeout threshold, the alert will be extended to all personnel in the target store. If a real-time event is classified as dangerous, but the text response time exceeds the dangerous level timeout threshold, then the repeated alarm mechanism should be canceled.

[0027] It should be noted that the terminal control system pre-configures two control mechanisms: a duplicate alarm suppression mechanism and a timeout escalation management mechanism. These two mechanisms are activated simultaneously and work collaboratively, addressing the two core issues of "solving duplicate alarms" and "solving response delays," respectively. The mechanism design is comprehensive, simultaneously resolving the two major pain points of duplicate alarms and response delays. The control logic is clear, requiring no manual intervention, reducing store personnel management costs, and improving the intelligence level of reminder control. The specific rules of the duplicate alarm suppression mechanism are as follows: during text and voice reminder pushes, a predetermined time window is set. Within this time window, if the same real-time event repeatedly triggers the reminder command, the terminal control system automatically generates a duplicate alarm password and simultaneously suppresses subsequent text and voice reminder pushes for that event, retaining only the initial reminder. The aim is to prevent repeated pushes within a short period from causing staff to become desensitized to or ignore the reminder information, while also reducing information overload on mobile terminals and ensuring the effectiveness of the reminder information. The timeout escalation management mechanism sets differentiated response time limits based on the urgency level of the event, ensuring rapid response to high-risk events (dangerous, urgent) and providing reasonable response time for low-risk events (early warnings).

[0028] Furthermore, in a preferred embodiment of the present invention, after the repeated alarm suppression and timeout escalation management processing, the optimization of the store event classification and scoring model by combining historical data networks specifically includes: Based on the terminal control system, relevant data during the process of repetitive alarm suppression and timeout escalation management are recorded in real time and stored in the core data feature library for data updates. Based on the updated core data feature library, the store event classification and scoring model is optimized. The optimization target is to combine historical data network to retrieve relevant data on the historical repeated alarm suppression and timeout escalation management process of the target store, and integrate and adjust the normalized weights of all evaluation dimensions. During the adjustment of the normalized weights of all evaluation dimensions, the weights are checked in real time to ensure that the sum of the weights of all evaluation dimensions is always equal to 1, thus completing the optimization of the store event classification scoring model and obtaining the optimized store event classification scoring model. For the optimized store event grading and scoring model, standardized alarm basic information generation is repeatedly invested.

[0029] It should be noted that the terminal control system automatically records all operational data generated—including the number of repeated alarms, suppression time, timeout counts, timeout levels, response duration, escalation records, and expanded alert records—and synchronously updates it to the core data feature library, expanding historical data and ensuring the completeness and traceability of data sources. Using the updated core data feature library, combined with the historical data network, the system analyzes which events are prone to repeated alarms, which events are prone to timeouts, and which dimensions lead to misclassification. It then adjusts the normalized weights of the five evaluation dimensions accordingly, allowing the model to correct its judgment logic based on actual operational results, reducing repeated triggers, reducing misclassification, and improving sensitivity to high-risk events. When adjusting the weights of the five dimensions, the system automatically performs normalization verification to ensure that the total weight is strictly equal to 1, avoiding calculation overflow, abnormal scoring, and incorrect classification. Finally, an optimized model is obtained and re-integrated into the real-time event classification process, forming a closed-loop iteration.

[0030] Figure 2 A flowchart illustrating a method for coordinating push text and voice notifications is shown, including the following steps: S202: Based on standardized alarm information, the terminal control system allocates reminder permissions and classifies target stores for reminders, and coordinates the push of text reminder information and voice reminder information. S204: Based on reminder commands, coordinate the push of text reminder messages and voice reminder messages.

[0031] Furthermore, in a preferred embodiment of the present invention, the step of allocating reminder permissions and classifying target store reminder objects in the terminal control system based on standardized alarm basic information, and coordinating the push of text reminder information and voice reminder information, specifically includes: Determine the personnel structure of the target store and the regional permissions of different personnel in the target store, enter the data into the terminal control system, obtain the mobile terminals of different personnel in the target store, and connect the mobile terminals of different personnel in the target store to the terminal control system. In the terminal control system, based on standardized alarm information, the target branch reminder objects are defined. The rule for defining the target branch reminder objects is that if the current level of a real-time event is dangerous, then in the terminal control system, the target branch reminder objects are determined to be all personnel. If the current level of a real-time event is emergency, the terminal control system determines the target audience to be personnel within the target store area where the event occurred. The system then detects the response of the corresponding personnel's mobile terminals. If no response is received within a predetermined time, the target audience is expanded to include the personnel with the highest area access in the target store. If the current level of a real-time event is an early warning level, then the terminal control system will determine the target of the alert as the personnel in the target store area where the target event occurred. If the current level of a real-time event is normal, then it is not necessary to determine the recipient of the alert. Based on the classification of target stores and standardized alarm information, reminder instructions are encapsulated and generated. Then, based on the reminder instructions, text reminders and voice reminders are coordinated and pushed.

[0032] It's important to first clarify the personnel structure of the target store, such as the store manager, staff, and regional managers, and assign different management permissions to different personnel. For example, staff member A is responsible for the fresh produce section, and staff member B is responsible for the checkout section. This information is then entered into the terminal control system. Simultaneously, the mobile devices of all relevant personnel are acquired, and the mobile devices are bound to the terminal control system to ensure that reminders are accurately pushed to the corresponding personnel. The significance of this is to establish a "personnel-permission-terminal" association system, laying the foundation for subsequent "on-demand push and precise reminders," clarifying "who should receive reminders and who has the authority to handle them," avoiding misdirected reminder pushes. Furthermore, clear personnel permissions and stable terminal binding ensure that subsequent reminders can be directly matched to the corresponding personnel, preventing irrelevant personnel from receiving invalid reminders, and ensuring that personnel with handling authority receive reminders in a timely manner, improving response efficiency. Subsequently, personnel are categorized according to the current level of the real-time event, ensuring full coverage of all personnel at each level of danger. This avoids delays in handling due to missed reminder recipients and minimizes the harm of dangerous events. Emergency level incidents do not require full staff involvement; they are prioritized for notification to the area manager to improve efficiency. If there is no response, notification is extended to the store manager to ensure the incident is addressed and not overlooked. Warning level incidents are minor and have a narrow impact, requiring only the area manager's attention to avoid disturbing the store manager and other unrelated personnel, thus reducing the information burden on all staff. Normal level incidents require no notification.

[0033] Furthermore, in a preferred embodiment of the present invention, the step of coordinating the push of text reminder information and voice reminder information based on the reminder instruction specifically includes: Based on the terminal control system, the system reads the reminder command and, according to the classification of the target store reminder object in the reminder command, synchronously pushes standardized alarm basic information to the mobile terminal of the corresponding personnel. During the synchronous push process, standardized alarm basic information is packaged into text reminders and voice reminders; For text-based reminders, the system directly pushes them to the mobile devices of the corresponding personnel via the terminal control system. After the text reminder is pushed, the terminal control system detects the mobile terminal of the corresponding person. If the mobile terminal of the corresponding person responds to the text reminder, there is no need to push the voice reminder. If there is no response text notification on the mobile terminal of the corresponding personnel, the text notification is classified into target events. If it is a dangerous level event or an emergency level event, the terminal control system is used to control the prompt volume of the corresponding personnel's mobile terminal to be turned up to the maximum and play the voice reminder message in a loop until the corresponding personnel's mobile terminal responds with text notification messages. If it is a warning-level event, the terminal control system will keep the volume of the prompts on the mobile terminals of the corresponding personnel constant and play the voice reminder information in a loop until the mobile terminals of the corresponding personnel all have a response text reminder information.

[0034] It should be noted that, based on the pre-defined target audience, standardized alarm information is precisely pushed to the mobile devices of the corresponding personnel, avoiding unrelated individuals. The push process is divided into text and voice messages. Text messages are quiet and viewable, while voice messages are mandatory. Text and voice messages are generated from the same source, synchronously, and with the same structured content, rather than from two independent systems, ensuring information consistency. Text messages are prioritized to avoid disturbing residents, being jarring, or affecting the store's operating environment. If everyone responds to the text message, there is no need for voice broadcast. If someone does not respond, and the event is of a dangerous or urgent level, the system forcibly increases the volume of the mobile devices to maximum and loops the voice message until everyone responds. Similarly, warning events use a normal volume loop of voice messages without increasing the volume, providing a reminder without excessive interference. This refined differentiation of reminder intensity balances the effectiveness of the reminder with the business environment, avoiding disturbance, jarring sounds, and causing employee stress.

[0035] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A store intelligent alarm and voice reminder method based on event hierarchy, characterized in that, Includes the following steps: By using the store's omni-channel sensing terminals, multi-source sensing data is collected from the store, and the collected data is preprocessed to build a core data feature library. By combining the core data feature library, weight initialization processing and store event classification and scoring model construction are carried out in the terminal control system. Based on the store event classification and scoring model, dynamic classification calculation and proactive reminder trigger determination are performed on real-time events, and standardized alarm basic information is generated. Based on standardized alarm information, the terminal control system allocates reminder permissions and classifies target stores for reminders, and coordinates the push of text and voice reminder information. Continuously push text and voice reminders, and manage and handle repeated alarms and timeout escalation. After handling repeated alarm suppression and timeout escalation management, the store event classification and scoring model is optimized by combining historical data networks.

2. The store intelligent alarm and voice reminder method based on event hierarchy as described in claim 1, characterized in that, The process involves collecting multi-source sensing data from the store through its omni-channel sensing terminals, preprocessing the collected data, and constructing a core data feature library. Specifically: Identify stores, designate them as target stores, and determine the terminal control system within the target stores; The terminal control system controls all types of sensing terminals in the target store, including security sensing terminals, operational sensing terminals, and environmental and abnormal sound sensing terminals. The terminal control system controls all types of sensing terminals to collect corresponding types of data, obtains store sensing terminal data, and filters and removes invalid information from the store sensing terminal data. Among them, invalid information in the store sensing terminal data includes the shape of data collected by different types of sensing terminals and redundant data; Meanwhile, after filtering and removing invalid information, the data from the store sensing terminals is standardized, including format unification and feature retrieval, to obtain the feature data of the store sensing terminals. By introducing a historical data network, all core feature types for the target store are retrieved. Based on all core feature types of the target store, core features are extracted from the feature data of the store's sensing terminal and a feature database is established to obtain a core data feature library.

3. The store intelligent alarm and voice reminder method based on event hierarchy as described in claim 1, characterized in that, The process of combining the core data feature library with the terminal control system for weight initialization and the construction of a store event grading and scoring model is as follows: The core features of the store perception terminal feature data in the core data feature library are summarized into event type, occurrence area, duration, numerical threshold deviation, associated device status, and on-site environment, which serve as the basic data for the store event classification model. The core features of the store perception terminal feature data in the core data feature library are mapped to the five evaluation dimensions of the store event classification model, which are labeled as the degree of security impact, the degree of business impact, the degree of risk diffusion, the degree of continuous accumulation, and the degree of environmental interference. At the same time, retrieve the historical weight percentage of all evaluation dimensions and sort them in reverse order to achieve the initial weight allocation of all evaluation dimensions; By using historical data networks, the frequency, false alarm probability, and response efficiency of historical event types of target stores are retrieved. This information is used to normalize and correct the initial weights of each evaluation dimension until the sum of the weights of all evaluation dimensions is always equal to 1. At the same time, the event comprehensive risk scoring algorithm is retrieved. The event comprehensive risk scoring algorithm quantifies the core features of different evaluation dimensions and then combines them with the normalized initial weights of the evaluation dimensions to construct an event comprehensive risk scoring model. Define the historical events that occurred in the target store, and output the comprehensive risk score value of the historical events through the event comprehensive risk scoring model. This value is used to divide the scoring range into multiple levels and to update the event comprehensive risk scoring model, thus obtaining the store event classification scoring model. The multi-level rating range includes four levels: dangerous, emergency, warning, and normal.

4. The store intelligent alarm and voice reminder method based on event hierarchy as described in claim 1, characterized in that, The store-based event grading and scoring model dynamically grades and calculates real-time events, determines proactive alert triggers, and generates standardized alarm information, specifically: Through the terminal control system, real-time sensing data from security sensing terminals, operation sensing terminals, and environmental and abnormal sound sensing terminals are collected and transmitted to the core data feature library for real-time updates, generating the core features of the real-time sensing data. The core features of real-time sensing data are imported into the store event classification and scoring model, which outputs the event type and corresponding comprehensive risk score of the real-time events of the target store, and generates the current level of the real-time events. An active reminder mechanism is assigned to the four levels in the multi-level scoring range. Specifically, real-time events at the danger and emergency levels are set to require active reminders. For real-time events at the warning level, the duration is calculated. If the duration exceeds the preset value, it is set to require active reminders; if the duration is not greater than the preset value, it is set to not require active reminders. For normal-level real-time events, no active reminders are required. The core characteristics of the real-time sensing data corresponding to the active reminder mechanism are recorded in the terminal control system. Combined with the current level of the real-time event, standardized alarm basic information is encapsulated and generated.

5. The store intelligent alarm and voice reminder method based on event hierarchy as described in claim 1, characterized in that, Based on standardized alarm information, the terminal control system allocates alert permissions and classifies target stores for alerts, and coordinates the push of text and voice alert messages. Specifically: Determine the personnel structure of the target store and the regional permissions of different personnel in the target store, enter the data into the terminal control system, obtain the mobile terminals of different personnel in the target store, and connect the mobile terminals of different personnel in the target store to the terminal control system. In the terminal control system, based on standardized alarm information, the target branch reminder objects are defined. The rule for defining the target branch reminder objects is that if the current level of a real-time event is dangerous, then in the terminal control system, the target branch reminder objects are determined to be all personnel. If the current level of a real-time event is emergency, the terminal control system determines the target audience to be personnel within the target store area where the event occurred. The system then detects the response of the corresponding personnel's mobile terminals. If no response is received within a predetermined time, the target audience is expanded to include the personnel with the highest area access in the target store. If the current level of a real-time event is an early warning level, then the terminal control system will determine the target of the alert as the personnel in the target store area where the target event occurred. If the current level of a real-time event is normal, then it is not necessary to determine the recipient of the alert. Based on the classification of target stores and standardized alarm information, reminder instructions are encapsulated and generated. Then, based on the reminder instructions, text reminders and voice reminders are coordinated and pushed.

6. The store intelligent alarm and voice reminder method based on event hierarchy as described in claim 5, characterized in that, The coordinated push of text and voice reminders based on reminder commands specifically includes: Based on the terminal control system, the system reads the reminder command and, according to the classification of the target store reminder object in the reminder command, synchronously pushes standardized alarm basic information to the mobile terminal of the corresponding personnel. During the synchronous push process, standardized alarm basic information is packaged into text reminders and voice reminders; For text-based reminders, the system directly pushes them to the mobile devices of the corresponding personnel via the terminal control system. After the text reminder is pushed, the terminal control system detects the mobile terminal of the corresponding person. If the mobile terminal of the corresponding person responds to the text reminder, there is no need to push the voice reminder. If there is no response text notification on the mobile terminal of the corresponding personnel, the text notification is classified into target events. If it is a dangerous level event or an emergency level event, the terminal control system is used to control the prompt volume of the corresponding personnel's mobile terminal to be turned up to the maximum and play the voice reminder message in a loop until the corresponding personnel's mobile terminal responds with text notification messages. If it is a warning-level event, the terminal control system will keep the volume of the prompts on the mobile terminals of the corresponding personnel constant and play the voice reminder information in a loop until the mobile terminals of the corresponding personnel all have a response text reminder information.

7. The store intelligent alarm and voice reminder method based on event hierarchy as described in claim 1, characterized in that, The continuous push of text and voice reminders, along with the suppression of repeated alarms and timeout escalation management, specifically involves: Set up a recurring alarm suppression mechanism and a timeout escalation management mechanism within the terminal control system; The repeated alarm mechanism is as follows: during the process of pushing text reminder information and voice reminder information, a predetermined time window is set. If the same real-time event generates reminder instructions repeatedly within the predetermined time window, a repeated alarm password is generated in the terminal control system, and the push of text reminder information and voice reminder information corresponding to the same real-time event by the terminal control system is suppressed. The timeout escalation management mechanism involves setting timeout thresholds for real-time events of different levels through a terminal control system, including timeout thresholds for warning levels, timeout thresholds for emergency levels, and timeout thresholds for danger levels. Simultaneously, the terminal control system calculates in real time the time it takes for the person receiving the text notification to respond to the text notification on their mobile terminal, and marks it as the text response time. If a real-time event is at the warning level, but the text response time exceeds the warning level timeout threshold, then the notification volume on the corresponding personnel's mobile terminal will be increased to the maximum, and the notification recipients will be extended to the personnel with the highest area permissions in the target store. If a real-time event is classified as urgent, but the text response time exceeds the urgent timeout threshold, the alert will be extended to all personnel in the target store. If a real-time event is classified as dangerous, but the text response time exceeds the dangerous level timeout threshold, then the repeated alarm mechanism should be canceled.

8. The store intelligent alarm and voice reminder method based on event hierarchy as described in claim 1, characterized in that, Following the handling of repeated alarm suppression and timeout escalation management, and combining historical data networks, the store event classification and scoring model is optimized, specifically as follows: Based on the terminal control system, relevant data during the process of repetitive alarm suppression and timeout escalation management are recorded in real time and stored in the core data feature library for data updates. Based on the updated core data feature library, the store event classification and scoring model is optimized. The optimization target is to combine historical data network to retrieve relevant data on the historical repeated alarm suppression and timeout escalation management process of the target store, and integrate and adjust the normalized weights of all evaluation dimensions. During the adjustment of the normalized weights of all evaluation dimensions, the weights are checked in real time to ensure that the sum of the weights of all evaluation dimensions is always equal to 1, thus completing the optimization of the store event classification scoring model and obtaining the optimized store event classification scoring model. For the optimized store event grading and scoring model, standardized alarm basic information generation is repeatedly invested.

Images