Reminding method and system of AI voice medical advice wearable device

Abstract

The invention relates to the technical field of voice reminding, in particular to a reminding method and system of an AI voice medical advice wearable device, and the method comprises the following steps: extracting a medical advice text based on an identity field, analyzing a behavior keyword combination, generating an execution feature set, judging an execution frequency and precautions field, extracting grade parameters, and generating a response grade interval; and matching the voice identity model to screen broadcast characteristics, scheduling a voice unit to push audio, and monitoring broadcast data to generate a reminding completion identification field. According to the method, the electronic doctor's advice text is matched based on the identity field, continuous instruction phrases and behavior action phrases are extracted, semantic judgment of the execution frequency field and the attention item field in the behavior instruction is combined, and the voice identity is selected according to the grade label; a reminding completion identification field is generated based on indexes such as a buffer state and interruption times in a broadcasting process, so that the response accuracy of voice reminding, the behavior guide effect of information transmission and the tracking capability of a user to confirm an action can be enhanced.

Description

Technical Field

[0001] This invention relates to the field of voice reminder technology, and in particular to a reminder method and system for an AI voice medical advice wearable device. Background Technology

[0002] The field of voice prompt technology involves voice interaction systems built upon artificial intelligence technologies such as speech recognition, speech synthesis, and natural language processing, aiming to provide audio prompts, warnings, or guidance to specific targets. This field includes speech generation algorithms, personalized speech simulation technology, semantic recognition and context matching mechanisms, and communication and control interfaces with external hardware devices. By combining voice with functions such as scene awareness, identity recognition, and task scheduling, this technology can achieve multi-role, multi-task voice output, exhibiting broad adaptability and responsiveness in various application environments such as healthcare, home, transportation, and education.

[0003] The AI-powered voice-guided medical advice wearable device aims to provide users with medical advice reminders by combining wearable devices with artificial intelligence voice technology. Its applications include simulating the voices of doctors, experts, health managers, or family members, issuing trustworthy and authoritative voice medical advice through the wearable device, thereby increasing user acceptance and adherence to health management suggestions. This method can be used in scenarios such as medication reminders for chronic disease patients, post-operative rehabilitation management, and health guidance for the elderly, achieving personalized, multi-source, and highly recognizable voice-guided medical advice interaction.

[0004] Traditional reminder methods rely solely on preset character audio for content prompts, lacking semantic recognition and hierarchical evaluation mechanisms for behavioral instructions in medical orders. This makes it difficult to adjust broadcast strategies based on the urgency and complexity of the task. When faced with medical orders requiring time-limited responses or intervention, they cannot trigger a voice broadcast model with recognizability and a sense of urgency. Consequently, the reminder voice fails to reflect behavioral priority in terms of both auditory perception and content, resulting in weak user recognition of the task's importance. In some situations, delays may occur due to ambiguous broadcast information or insufficient reminder timeliness. This is particularly true in postoperative rehabilitation and high-frequency medication reminder scenarios, where the lack of hierarchical broadcasting can easily lead to information fatigue and execution omissions. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a reminder method and system for AI voice medical advice wearable devices.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a reminder method for an AI voice medical advice wearable device, comprising the following steps: S1: Based on the electronic medical order database, detect the continuous arrangement of instruction phrases and action phrases in the text, analyze the frequency and combination relationship of three types of keywords involving medication, examination, and rehabilitation behavior in the field position, and generate a set of medical order execution features; S2: Based on the set of medical order execution features, determine whether there are three semantic fragments such as "immediately", "time-limited" and "dependent on intervention" in the execution frequency field and precaution field marked for each behavioral instruction, compare the level boundary thresholds, establish corresponding level labels, and generate medical order response level ranges; S3: Based on the medical order response level range, call the stimulus perception level range of the four types of voice identities (doctor, expert, family member, health worker) preset in the AI ​​voice identity library, filter the identity models that match the current range, and generate reminder voice selection results; S4: Call the reminder voice selection result and the medical order execution feature set, match the scheduling queue of the voice broadcasting unit, the speaker audio channel control parameters and the unique device identifier of the wearable device, schedule the voice broadcasting unit to implement task distribution and push audio to the terminal device, and generate wearable terminal broadcasting record data.

[0007] The present invention is improved in that the set of medical order execution features includes execution behavior category labels, behavior target content fragments, time association fields, behavior timeliness classification identifiers, and precaution aggregation items; the medical order response level range includes response intensity level labels, time sensitivity level identifiers, and semantic urgency level identifiers; the reminder voice selection result includes voice identity type number, voice broadcast text segment, tone adjustment parameter group, speech rate matching index field, and broadcast time trigger field; and the wearable terminal broadcast record data includes broadcast task status identifier code, device communication log item, and voice broadcast timing value.

[0008] The present invention is improved in that the step of obtaining the medical order execution feature set is specifically as follows: S111: Based on the electronic medical order database, using the ID card number and medical record number in the user identity field, detect the mapping relationship between the identity identifier field and the record identifier field in the data index group, locate the data nodes with associated markers, call the medical order text data frame corresponding to the located data node, and generate the identity-bound medical order data frame. S112: Based on the text content field in the identity-bound medical order data frame, retrieve text fragments that begin with a verb phrase and end with a noun representing the action object. Combine the action verbs and the following nouns in the text fragments continuously, and identify the total number of occurrences and positional order of the combination patterns in the paragraph to obtain the action semantic arrangement feature group. S113: Based on the distribution information of behavioral verbs and the content of noun objects in the behavioral semantic arrangement feature group, call the predefined keyword classification dictionary, match the keyword grouping in each text segment with the field position relationship, determine whether the keyword belongs to the category of medication, examination or rehabilitation behavior, filter the text paragraph structure features that meet the dual matching conditions of keyword and classification category, and establish a set of medical order execution features.

[0009] The present invention is improved in that the step of obtaining the medical order response level range is specifically as follows: S211: Based on the behavioral instruction field in the medical order execution feature set, extract the execution frequency field and precaution field bound to the behavioral instruction, detect whether the field includes immediate, time-limited and intervention-dependent semantic fragments, and mark and map the field labels with semantic expression to establish an emergency feature indication mapping set; S212: Based on the field marking results in the emergency feature indication mapping set, collect the corresponding behavioral instructions' preset severity level value, task load score, and response time limit value in the database, using the formula: ; The response level value is obtained through calculation. By comparing it with the level boundary threshold, the level range of each behavioral instruction is determined. Instructions belonging to the same range are classified and grouped to establish the medical order response level range. in, Indicates the severity level of the illness. This represents the arithmetic mean of the severity levels in the set of behavioral instructions. This represents the normalized value of the task load score. This represents the normalized value of the response time limit. Represents the semantic density coefficient. , , The first The normalized values ​​of the severity rating, task load score, and response time limit corresponding to each action instruction. Indicates the response level value.

[0010] The present invention is improved in that the step of obtaining the reminder voice selection result is specifically as follows: S311: Based on the medical order response level range, call the AI ​​voice identity library, detect the stimulus perception level value mapped by the identity label field in the model group, and index and aggregate the stimulus value range of the voice model to establish the response level structure corresponding to the identity model and generate the identity stimulus level mapping set. S312: Call the combined level label field in the medical order response level range, compare the response level value marked by the field with the level range boundary corresponding to each identity model in the identity stimulus level mapping set, determine whether there is an overlapping range, extract the cross-matching voice identity model identifier, and establish an identity level matching structure group. S313: Based on the voice identity model identifier in the identity level matching structure group, extract the timbre spectrum index, speech rate control parameter field and broadcast intensity field corresponding to the matching model, and combine them with the semantic instruction content in the medical order execution feature set to perform text segment combination and broadcast parameter integration processing to establish the reminder voice selection result.

[0011] The present invention is improved in that the step of obtaining the recorded data broadcast by the wearable terminal is specifically as follows: S411: Call the reminder voice selection result and the behavior action content and execution time field in the medical order execution feature set, retrieve the broadcast voice template segment corresponding to the behavior verb, extract the speech rate adjustment parameter and timbre index value that match the execution time and the speaker identity field, and splice and combine the parameters and semantic segments to generate a voice broadcast combination structure; S412: Based on the broadcast timestamp field and the voice identity index field in the voice broadcast combination structure, match the time channel mapping relationship, identity priority allocation index, and channel resource table entries in the voice broadcast unit scheduling table, and combine the registration status table of the wearable terminal identification code field to establish the identity time channel allocation mapping and obtain the broadcast task scheduling parameter set; S413: Call the identity, time, and device index fields in the broadcast task scheduling parameter set, execute the voice broadcast unit scheduling control, allocate broadcast resources according to the task parameters, start the corresponding terminal channel, push the combined audio command data to the audio output interface of the wearable device, and establish wearable terminal broadcast record data.

[0012] The present invention has an improvement, wherein the method further includes the following steps: S5: Based on the broadcast record data of the wearable terminal, monitor whether the audio output buffer status, voice interruption number and user confirmation action event are completely recorded in the broadcast task, the response duration of the combination process and the instruction confirmation identifier, and generate a medical order reminder completion identifier field; The medical order reminder completion identifier field includes the broadcast status feedback code, the instruction response confirmation mark, and the broadcast task tracking number.

[0013] The present invention is improved in that the specific steps for obtaining the medical order reminder completion identifier field are as follows: S511: Based on the broadcast recording data of the wearable terminal, detect the real-time status parameters of the audio output buffer unit in the broadcast task, monitor the synchronization stability between the buffer fill rate field and the output flow rate field, determine the continuity of the waveform frame, extract the record index value of the interruption or abnormal delay, and establish audio output status information. S512: Based on the interruption marker index in the audio output status information, retrieve the voice interruption count field and user confirmation action event field under the same batch of tasks, calculate the integrity ratio of the two types of field records, compare with the preset integrity threshold, filter the record sample set that meets the threshold condition, and generate a broadcast interaction integrity parameter set. S513: Call the time duration field and confirmation flag field in the broadcast interaction integrity parameter set, combine and encode the response duration and user confirmation flag in the task, write the encoding result into the task execution tracking index, construct a structured status mark record, and generate a medical order reminder completion flag field.

[0014] The present invention improves upon this invention by setting the integrity threshold by statistically analyzing the recording frequency of the voice interruption count field and the user confirmation action event field in the same batch of broadcast task records, calculating the average and standard deviation of the recording frequency respectively, and setting the integrity threshold as the value corresponding to the sum of the average and standard deviation. When the recording frequency of both the voice interruption count field and the user confirmation action event field is greater than or equal to the integrity threshold, the corresponding record samples are selected and written into the broadcast interaction integrity parameter set.

[0015] A reminder system for an AI voice-guided medical advice wearable device, the reminder system for the AI ​​voice-guided medical advice wearable device being used to implement the aforementioned reminder method for the AI ​​voice-guided medical advice wearable device, the system comprising: The execution feature extraction module, based on the electronic medical order database, detects the continuous arrangement of instruction phrases and behavioral action phrases in the text, analyzes the frequency and combination relationship of three types of keywords involving medication, examination, and rehabilitation behavior in the field positions, and generates a set of medical order execution features; The emergency assessment module for medical orders, based on the set of medical order execution features, determines whether there are three semantic fragments: "immediately," "time-limited," and "dependent on intervention" in the execution frequency field and precaution field marked for each behavioral instruction, compares the level boundary thresholds, establishes corresponding level labels, and generates a range of medical order response levels. The voice model filtering module, based on the medical order response level range, calls the stimulus perception level range of four preset voice identity labels (doctor, expert, family member, and health worker) in the AI ​​voice identity library, filters identity models that match the current range, and generates reminder voice selection results. The terminal broadcast execution module calls the reminder voice selection result and the medical order execution feature set, matches the scheduling queue of the voice broadcast unit, the speaker audio channel control parameters and the unique device identifier of the wearable device, schedules the voice broadcast unit to implement task distribution and pushes audio to the terminal device, and generates wearable terminal broadcast record data; The medical order reminder recording module, based on the broadcast recording data of the wearable terminal, monitors whether the audio output buffer status, the number of voice interruptions and user confirmation action events are completely recorded in the broadcast task, combines the response duration of the process and the instruction confirmation identifier, and generates a medical order reminder completion identifier field.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by matching electronic medical order text based on identity fields and extracting continuous instruction phrases and behavioral action phrases, and combining semantic judgment of execution frequency fields and precaution fields in behavioral instructions, a graded assessment of medication, examination, and rehabilitation behavioral tasks is achieved. Voice identities with role-awareness are selected according to grade labels, and broadcast voice content with timbre differentiation and speech rate control is generated based on voice identity parameters. Broadcast task scheduling is performed and broadcast record data is generated by combining broadcast time and unique device identification information. A reminder completion identifier field is generated based on indicators such as buffer status and number of interruptions during broadcasting. This enhances the responsiveness of voice reminders, the behavioral guidance effect of information delivery, and the tracking ability of user confirmation actions, achieving a dynamic feedback loop of task importance perception, voice broadcast matching, and broadcast results. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the process of obtaining a set of medical order execution features according to the present invention; Figure 3 This is a flowchart illustrating the process of obtaining the medical order response level range in this invention; Figure 4 This is a flowchart illustrating the process of obtaining the selection results for the reminder voice in this invention; Figure 5 This is a flowchart illustrating the process of acquiring wearable terminal broadcast record data according to the present invention; Figure 6 This is a flowchart illustrating the process of obtaining the medical order reminder completion identifier field in this invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0020] Please see Figure 1 This invention provides a technical solution, a reminder method for an AI voice medical advice wearable device, comprising the following steps: S1: Based on the electronic medical order database, using the ID card number and medical record number in the user identity field, call the electronic medical order text content that matches the identity field, detect the continuous arrangement of instruction phrases and behavioral action phrases in the text, analyze the frequency of occurrence and combination relationship of three types of keywords involving medication, examination, and rehabilitation behavior in the field position, and generate a set of medical order execution features. S2: Based on the set of medical order execution features, determine whether there are three types of semantic fragments such as "immediate", "time-limited" and "dependent on intervention" in the execution frequency field and precaution field marked for each behavioral instruction. Extract the severity level value, task load score and response time limit value set under the corresponding field, compare the level boundary threshold, establish the level label corresponding to the execution feature, and generate the medical order response level range. S3: Based on the medical order response level range, call the stimulus perception level range of the four types of voice identities (doctor, expert, family member, health worker) preset in the AI ​​voice identity library, determine the overlap of the level range, filter the identity model that matches the current range, read the timbre spectrum index, speech rate value and tone intensity under the identity model, and generate the reminder voice selection result. Stimulus perception level range is a psychoacoustic indicator used in HCI speech interaction, often measured by the SAM self-rating emotion scale or the "arousal" dimension in Russell's two-dimensional emotion space; timbre spectrum index is a timbre difference parameter characterized by Mel frequency cepstral coefficients, which is widely used in speech synthesis; speech rate value measures the number of syllables spoken per unit time; pitch intensity is based on the amplitude of F0 fundamental frequency variation and envelope power, serving as an indicator of emotional expression intensity; S4: Invoke the reminder voice selection results and the medical order execution feature set, assemble the broadcast voice content segment and voice feature index according to the corresponding behavioral action content and execution time expression content, combine the broadcast timestamp field and voice identity index field, match the scheduling queue of the voice broadcast unit, the speaker audio channel control parameters and the unique device identifier of the wearable device, schedule the voice broadcast unit to implement task distribution and push audio to the terminal device, and generate wearable terminal broadcast record data; S5: Based on the broadcast record data of the wearable terminal, monitor whether the audio output buffer status, voice interruption number and user confirmation action event are completely recorded in the broadcast task, combine the response duration and instruction confirmation flag of the process, and generate a medical order reminder completion flag field; The set of features for medical order execution includes execution behavior category labels, behavior target content fragments, time-related fields, behavior timeliness classification identifiers, and precaution aggregation items. The range of medical order response levels includes response intensity level labels, time sensitivity level identifiers, and semantic urgency level identifiers. The result of the reminder voice selection includes voice identity type number, voice broadcast text segment, tone adjustment parameter group, speech rate adjustment index field, and broadcast time trigger field. The data recorded by the wearable terminal broadcast includes broadcast task status identifier code, device communication log item, and voice broadcast time value. The medical order reminder completion identifier field includes broadcast status feedback code, instruction response confirmation mark, and broadcast task tracking number.

[0021] Please see Figure 2 The specific steps for obtaining the feature set of medical order execution are as follows: S111: Based on the electronic medical order database, using the ID card number and medical record number in the user identity field, detect the mapping relationship between the identity identifier field and the record identifier field in the data index group, locate the data nodes with associated markers, call the medical order text data frame corresponding to the located data node, and generate the identity-bound medical order data frame. The system retrieves the ID number "33010619700307XXXX" and medical record number "A0001" from the user identity field. Using these two numbers, it checks the mapping relationship between the "Patient Unique ID" field and the "Medical Order Record Number" field in the data index group. It then performs data indexing and precisely locates data nodes with associated markers using the medical order record number "OR12345". For example, it locates the medical order data node with the storage address "db_addr_001". It then calls the corresponding medical order text data frame for this located medical order data node, which contains the text content "Intravenous injection of ceftriaxone sodium 2.0g, once daily". Finally, it generates an identity-bound medical order data frame containing the ID number "33010619700307XXXX", medical record number "A0001", medical order record number "OR12345", and the medical order text content "Intravenous injection of ceftriaxone sodium 2.0g, once daily".

[0022] S112: Based on the text content field in the identity-bound medical order data frame, retrieve text fragments that begin with a verb phrase and end with a noun representing the action object. Combine the action verbs and the following nouns in the text fragments continuously, and identify the total number of occurrences and positional order of the combination patterns in the paragraph to obtain the action semantic arrangement feature group. Based on the content field "intravenous injection of ceftriaxone sodium 2.0g, once daily", the search retrieves text fragments that begin with a verb phrase and end with a noun representing the action object. For example, it identifies "intravenous injection" as a verb phrase and "ceftriaxone sodium" as the noun representing the action object. The action verb "injection" in the text fragment is continuously combined with the following noun "ceftriaxone sodium" to obtain "injecting ceftriaxone sodium". The combination pattern "injecting ceftriaxone sodium" in the paragraph is identified as having a total occurrence count of 1 and being the first action semantic fragment. This results in an action semantic arrangement feature group, which includes the medical order record number "OR12345", the action verb "injection", the action object noun "ceftriaxone sodium", the occurrence count of 1, and the position order of 1.

[0023] S113: Based on the distribution information of behavioral verbs and the content of noun objects in the behavioral semantic arrangement feature group, call the predefined keyword classification dictionary, match the keyword grouping in each text segment with the field position relationship, determine whether the keyword belongs to the category of medication, examination or rehabilitation behavior, filter the text paragraph structure features that meet the dual matching conditions of keyword and classification category, and establish a set of medical order execution features. Based on the distribution information of the action verb "inject" and the noun object content "cefotaxime sodium", a predefined keyword classification dictionary is invoked. This dictionary sets "injection" and "take" as keywords for "medication behavior", "electrocardiogram" and "chest X-ray" as keywords for "examination behavior", and "massage" and "physiotherapy" as keywords for "rehabilitation behavior". The keywords "injection" and "cefotaxime sodium" in each text segment are grouped and matched with their corresponding field positions. For example, "injection" is matched to the verb position and "cefotaxime sodium" is matched to the object noun position. It is determined whether the category of the keyword "injection" belongs to medication, examination, or rehabilitation behavior. Specifically, it is determined that "injection" belongs to the "medication behavior" category. Text segment structural features that meet the dual matching conditions of "injection" belonging to "medication behavior" and "cefotaxime sodium" being the medication object are selected. For example, the structural feature "medication behavior: injecting ceftriaxone sodium" is selected. A set of medical order execution features is established, which includes the medical order record number "OR12345", the action instruction "injecting ceftriaxone sodium", and the action type "medication".

[0024] Please see Figure 3 The specific steps for obtaining the medical order response level range are as follows: S211: Based on the behavioral instruction field in the medical order execution feature set, extract the execution frequency field and precaution field bound to the behavioral instruction, detect whether the fields include immediate, time-limited and intervention-dependent semantic fragments, and mark and map the semantically expressive field labels to establish an emergency feature instruction mapping set; Based on the behavioral instruction field "inject ceftriaxone sodium", the execution frequency field "once daily" and the precaution field "beware of allergic reactions" bound to the behavioral instruction are extracted. The execution frequency field "once daily" and the precaution field "beware of allergic reactions" are checked to see if they contain any semantic fragments related to immediate, time-limited, or dependent intervention. The semantically expressive field labels are then marked and mapped. For example, fields that do not contain semantic fragments related to immediate, time-limited, or dependent intervention are marked as "routine". An emergency feature indication mapping set is established, which includes the medical order record number "OR12345", the behavioral instruction "inject ceftriaxone sodium", and the emergency feature label "routine".

[0025] S212: Based on the field labeling results in the emergency feature indication mapping set, collect the corresponding behavioral instructions' preset severity level value, task load score, and response time limit value in the database, using the formula: ; The response level value is obtained through calculation. By comparing it with the level boundary threshold, the level range of each behavioral instruction is determined. Instructions belonging to the same range are classified and grouped to establish the medical order response level range. in, This represents the severity level classification value, a dimensionless score that quantifies the level of medical intervention corresponding to each medical order. The data type is integer. The arithmetic mean of the severity ratings in the set of behavioral instructions is derived from a comprehensive analysis of all... The average calculation, The normalized value representing the task load score is derived from the NASA-TLX task load survey score results and obtained through max-min normalization transformation. The normalized value representing the response time limit is derived from the second-level time difference after conversion of the execution time field in the medical order, and is obtained after normalization calculation based on the maximum response time. The semantic density coefficient is the ratio of the number of semantic segments in a behavioral instruction to the total number of words in the field, reflecting the degree of semantic clustering. , , The first The normalized values ​​of the severity classification, task load score, and response time limit corresponding to each action instruction, with data sourced from... , , same, The response level value represents the urgency of the medical order and is used to determine the boundary interval of the level. The level boundary threshold is set based on the preset emergency level mapping table inside the medical system. The baseline value is divided with reference to the 5-level criticality classification standard of CTCAE and the corresponding task load and time limit interval. The boundary is the distribution boundary of the three scores in 95% of the behavioral instruction samples as the threshold set to generate the interval boundary value of each level. Based on the field label result "routine", the criticality level value of the corresponding action instruction "inject ceftriaxone sodium" is collected from the database. Task load scoring With response time limit Criticality level classification The criteria for setting the severity level are based on the medical intervention level corresponding to the doctor's order, and are quantified with reference to the 5-level severity classification standard of CTCAE (Common Terminology Standard for Adverse Events). The value range is as follows: Integer values, such as the severity grading value for a routine medication order like "intravenous ceftriaxone sodium". A score of 3 indicates moderate to critical illness. This is a dimensionless score that quantifies the medical intervention level corresponding to each medical order. The data type is integer, and it represents the task load score. The normalized value is derived from the NASA-TLX mission workload survey score, obtained through max-min normalization transformation. For example, if the original NASA-TLX score is 60 points, the max-min normalization formula is: If the highest historical rating is set to 100 and the lowest rating to 20, then... Response time limit The normalized value is derived from the second-level time difference after conversion of the execution time field in the medical order, and is obtained after normalization calculation based on the maximum response time. For example, if the execution time limit of this medical order is 8 hours (28800 seconds), and the historical maximum response time is set to 24 hours (86400 seconds), then... semantic density coefficient This is the ratio of the number of semantic segments in a behavioral instruction to the total number of words in the field, reflecting the degree of semantic clustering. For example, if the behavioral instruction "inject ceftriaxone sodium" contains 3 words ("inject", "ceftriaxone sodium", where "ceftriaxone sodium" is a proper noun and considered as one word), and the number of semantic segments (the behavioral verb and the noun of the behavioral object) is 2, then... , The arithmetic mean of the severity ratings in the set of behavioral instructions is derived from a comprehensive analysis of all... The average calculation is used to set the total number of actions in the current action instruction set. The severity levels of the medical orders are as follows: (Injection of ceftriaxone sodium) (Immediate rescue) (Conventional physical therapy), then The formula used is: ; Perform calculations to obtain response level values The logical approach of this formula lies in using the molecule... The semantic density coefficient is used to measure the urgency of a single medical order. Reflects the clarity of the instructions. It reflects the degree to which the severity of the medical order deviates from the average severity of critical illness. Taking into account the workload and response time The higher the value of each term in the numerator, the greater the urgency. The higher the value of each term in the denominator, the greater the urgency. This represents the total urgency of the entire set of medical orders, used to normalize the urgency of individual medical orders, thereby obtaining the relative level of urgency of each order within the overall set. ,Should The value is used for subsequent level boundary interval determination. ; First, calculate the denominator; the remaining data needs to be collected. One doctor's order Value, setting (Immediate rescue) (Conventional physical therapy), then the denominator Then calculate the molecule Finally, the response level value is calculated. The threshold for severity levels is set based on a pre-defined emergency level mapping table within the medical system. The baseline values ​​are based on the CTCAE's 5-level criticality grading standard, with the boundaries defined by the 95% threshold in the behavioral instruction samples. The distribution boundaries of the three ratings serve as a threshold set, for example, by statistically analyzing historical data to obtain 95% of the samples. Values ​​distributed in Within this range, the range is divided into 5 level intervals, and the set of level boundary thresholds is defined as follows: The level range is then divided into: Level 1 Level 2 Level 3 Level 4 Level 5 The calculated response level value Compare with the level boundary threshold to determine The level range it belongs to, due to If the instruction belongs to the Level 2 range, then the instructions belonging to the same range will be classified and grouped to establish a medical order response level range. This range includes the medical order record number "OR12345", the behavioral instruction "inject ceftriaxone sodium", and the response level "Level 2".

[0026] Please see Figure 4 The specific steps for obtaining the voice selection results are as follows: S311: Based on the medical order response level range, call the AI ​​voice identity library, detect the stimulus perception level value mapped by the identity label field in the model group, index and aggregate the stimulus value range of the voice model, establish the response level structure corresponding to the identity model, and generate the identity stimulus level mapping set. Based on the level range "Level 2", the AI ​​voice identity library is invoked. This library stores different voice identity models, such as Model A, Model B, and Model C. The stimulus perception level values ​​mapped to the identity label fields in the model group are detected. For example, Model A is mapped to "Low stimulus (corresponding to Level 1-2)", Model B is mapped to "Medium stimulus (corresponding to Level 2-3)", and Model C is mapped to "High stimulus (corresponding to Level 4-5)". The stimulus value ranges of the voice models are indexed and aggregated. For example, the stimulus value ranges of Model A are aggregated into... The interval aggregation of model B is The interval aggregation of model C is as follows: Establish a response level structure corresponding to the identity model, and generate an identity stimulus level mapping set, which includes model A (level interval). Model B (Level Range) Model C (Level Range) ).

[0027] S312: Call the combined level label field in the medical order response level range, compare the response level value marked by the field with the level range boundary corresponding to each identity model in the identity stimulus level mapping set, determine whether there is an overlapping range, extract the cross-matching voice identity model identifier, and establish an identity level matching structure group. By comparing the response level value 2, as labeled in the field, with the level interval boundaries corresponding to each identity model in the identity stimulus level mapping set, it is determined whether there are any overlapping intervals. Specifically, the intervals of level value 2 and model A are compared. There is an overlap point 2. Compare the interval of the grade value 2 with that of model B. There is an overlap point 2. Compare the interval of the grade value 2 with that of model C. Without overlap, the voice identity model identifiers of cross-matching are extracted, that is, the identifiers of model A and model B are extracted, and an identity level matching structure group is established. This structure group includes the medical order response level "level 2" and the matching model identifiers "model A, model B".

[0028] S313: Based on the voice identity model identifier in the identity level matching structure group, extract the timbre spectrum index, speech rate control parameter field and broadcast intensity field corresponding to the matching model, and combine them with the semantic instruction content in the medical order execution feature set to perform text segment combination and broadcast parameter integration processing to establish the reminder voice selection result; Extract the timbre spectrum index, speech rate control parameter field, and broadcast intensity field corresponding to the matching model. For example, extract the timbre spectrum index of model A as "IDX_A", the speech rate control parameter as "1.0x (standard speech rate)", and the broadcast intensity as "0.7 (medium intensity)", and extract the timbre spectrum index of model B as "IDX_B", the speech rate control parameter as "1.2x (slightly faster speech rate)", and the broadcast intensity as "0.9 (high intensity)". Combine this with the semantic instructions in the medical order execution feature set. The content "inject ceftriaxone sodium" is processed by combining text segments and integrating broadcast parameters. For example, it is combined into "Patient A0001, please note, doctor's order: inject ceftriaxone sodium". The timbre spectrum index, speech rate and intensity parameters are integrated into the text segment to establish the reminder voice selection result. This content includes the doctor's order record number "OR12345", the semantic instruction "inject ceftriaxone sodium", model A (IDX_A, 1.0x, 0.7) and model B (IDX_B, 1.2x, 0.9).

[0029] Please see Figure 5 The specific steps for obtaining the data recorded by the wearable terminal broadcast are as follows: S411: Call the reminder voice selection result and the behavior action content and execution time field in the medical order execution feature set, retrieve the broadcast voice template segment corresponding to the behavior verb, extract the speech rate adjustment parameter and timbre index value matching the execution time and speaker identity field, and splice and combine the parameters and semantic segments to generate a voice broadcast combination structure; The system invokes the semantic instruction "inject ceftriaxone sodium" and the action content "inject ceftriaxone sodium" and execution time field "once daily" (e.g., the specific timestamp is "2024-11-01 08:00:00") from the medical order execution feature set. It retrieves the broadcast speech template fragment corresponding to the action verb "inject," for example, the template fragment "Please inject immediately...". It extracts the speech rate adjustment parameter "1.0x" and the timbre index value "IDX_A" that match the execution time "2024-11-01 08:00:00" with the speaker identity field (e.g., selecting model A). These parameters are then concatenated with the semantic fragment "Patient A0001, please note, medical order: inject ceftriaxone sodium," for example, to form a structure containing a timestamp, identity index, speech rate parameter, timbre index, and broadcast text. This generates a speech broadcast combination structure containing the timestamp "2024-11-01". 08:00:00”, Voice identity index “IDX_A”, Speech rate “1.0x”, Voice timbre “IDX_A”, Announcement text “Patient A0001, please note, doctor’s order: inject ceftriaxone sodium”.

[0030] S412: Based on the broadcast timestamp field and voice identity index field in the voice broadcast combination structure, match the time channel mapping relationship, identity priority allocation index, and channel resource table entries in the voice broadcast unit scheduling table, and combine the registration status table of the wearable terminal identification code field to establish the identity time channel allocation mapping and obtain the broadcast task scheduling parameter set; Based on the broadcast timestamp field "2024-11-01 08:00:00" and the voice identity index field "IDX_A", the time channel mapping relationship, identity priority allocation index, and channel resource table entries in the voice broadcast unit scheduling table are matched. For example, if the scheduling table indicates that channel C is available during the 8:00:00 time period, the priority of IDX_A is medium (P2), and the resource table entry for channel C is "wearable terminal T001", combined with the registration status table of the wearable terminal identification code field "T001" (e.g., status is "online"), an identity time channel allocation mapping is established. For example, the mapping is: T001 terminal, 8:00:00 time, channel C, P2 priority. The broadcast task scheduling parameter set is obtained, which includes the device index "T001", the time index "2024-11-01 08:00:00", the identity index "IDX_A", the channel "C", and the priority "P2".

[0031] S413: Call the identity, time, and device index fields in the broadcast task scheduling parameter set, execute the voice broadcast unit scheduling control, allocate broadcast resources according to the task parameters, start the corresponding terminal channel, push the combined audio command data to the wearable device audio output interface, and establish wearable terminal broadcast record data. The system invokes the identity "IDX_A", time "2024-11-01 08:00:00", and device index "T001" fields to execute the voice broadcast unit scheduling control. It allocates broadcast resources based on task parameters, such as allocating audio stream resources for channel C, and starts the corresponding terminal channel, such as starting the audio output interface of T001. It then pushes the combined audio command data (e.g., an audio stream containing IDX_A timbre, 1.0x speech rate, 0.7 intensity, and the text "Patient A0001, please note, doctor's order: inject ceftriaxone sodium") to the audio output interface of the wearable device T001, establishing wearable terminal broadcast record data. This data includes the task ID "TASK001", device ID "T001", broadcast time "2024-11-01 08:00:00", broadcast content summary "inject ceftriaxone sodium", and status "push".

[0032] Please see Figure 6 The specific steps for obtaining the identification field of the doctor's order reminder are as follows: S511: Based on the broadcast recording data of the wearable terminal, detect the real-time status parameters of the audio output buffer unit in the broadcast task, monitor the synchronization stability between the buffer fill rate field and the output flow rate field, determine the continuity of the waveform frame, extract the record index value of the interruption or abnormal delay, and establish audio output status information. The system detects the real-time status parameters of the audio output buffer unit in the broadcast task. For example, it monitors the synchronization stability between the real-time value of the buffer fill rate field "95%" and the real-time value of the output flow rate field "100kbps". For example, it determines the continuity of the waveform frame by calculating the rate of change of the fill rate over time and the degree of matching with the flow rate. For example, it determines that the waveform frame is "continuous". It extracts the record index value of the interruption or abnormal delay. For example, if there is no interruption or abnormal delay in this record, the record index value is empty. It establishes audio output status information, which includes the task ID "TASK001", the status "continuous", and the interruption marker index "none".

[0033] S512: Based on the interruption marker index in the audio output status information, retrieve the voice interruption count field and user confirmation action event field under the same batch of tasks, calculate the integrity ratio of the two types of field records, compare with the preset integrity threshold, filter the record sample set that meets the threshold condition, and generate the broadcast interaction integrity parameter set. The integrity threshold is set by counting the recording frequency of the voice interruption number field and the user confirmation action event field in the same batch of broadcast task records, calculating the average and standard deviation of the recording frequency respectively, and setting the integrity threshold as the value corresponding to the sum of the average and standard deviation. When the recording frequency of the voice interruption number field and the user confirmation action event field is greater than or equal to the integrity threshold, the corresponding record samples are selected and written into the broadcast interaction integrity parameter set. Based on the interruption marker index "None", retrieve the voice interruption count field and user confirmation action event field under the same batch of tasks (e.g., batch "BATCH001"). Calculate the completeness ratio of these two types of field records. For example, if the batch has 100 broadcast tasks, with 90 samples showing 0 voice interruption counts and 85 samples showing 1 user confirmation action event, the completeness ratios are 90% and 85% respectively. Compare this to a preset completeness threshold. The completeness threshold is set by calculating the recording frequency of the voice interruption count field and the user confirmation action event field in the same batch of broadcast task records, calculating the average and standard deviation of each recording frequency, and setting the completeness threshold as the sum of the average and standard deviation. For example, for the voice interruption count field, after counting 100 task records, the average interruption count is... Number of tasks / tasks, standard deviation is Next / task, then the voice interruption threshold For the "times / task" field, based on 100 task records, the average frequency of confirmation event records is [value missing]. Number of tasks / tasks, standard deviation is If the event threshold is confirmed per task, then the event is confirmed. Count / task, filter records based on the frequency of voice interruption (times / tasks). (Sub-task) And the frequency of user confirmation action event field recording ( (Sub-task) (here) use judge, use (To reflect completeness), a set of record samples is selected that meets a threshold condition, such as the task ID "TASK001". and Generate a broadcast interaction integrity parameter set, which includes task ID "TASK001", interruption count "0", and user confirmation "1".

[0034] S513: Call the time duration field and confirmation flag field in the broadcast interaction integrity parameter set, combine and encode the response duration and user confirmation flag in the task, write the encoding result into the task execution tracking index, build a structured status mark record, and generate the medical order reminder completion flag field; Call the duration field (e.g., broadcast duration is "5 seconds") and the confirmation flag field "1" to encode the response duration "5 seconds" and the user confirmation flag "1" in the task. For example, encode it as "5_1" in the format of "duration_confirmation status". Write the encoded result "5_1" into the task execution tracking index to build a structured status flag record, for example, the record is {task ID: TASK001, status flag: 5_1}. Generate the medical order reminder completion flag field, for example, the completion flag is "completed (5_1)".

[0035] A reminder system for an AI voice-guided medical advice wearable device, the system being used to implement the aforementioned reminder method for the AI ​​voice-guided medical advice wearable device, the system comprising: The execution feature extraction module, based on the electronic medical order database, detects the continuous arrangement of instruction phrases and behavioral action phrases in the text, analyzes the frequency and combination relationship of three types of keywords involving medication, examination, and rehabilitation behavior in the field positions, and generates a set of medical order execution features; The emergency assessment module for medical orders, based on the set of medical order execution features, determines whether there are three semantic fragments: "immediate," "time-limited," and "dependent on intervention" in the execution frequency field and precaution field of each behavioral instruction. It compares the level boundary thresholds, establishes corresponding level labels, and generates a range of medical order response levels. The voice model filtering module, based on the medical order response level range, calls the stimulus perception level range of four preset voice identity labels (doctor, expert, family member, health worker) in the AI ​​voice identity library, filters identity models that match the current range, and generates reminder voice selection results. The terminal broadcast execution module calls the reminder voice selection result and the medical order execution feature set, matches the scheduling queue of the voice broadcast unit, the speaker audio channel control parameters and the unique device identifier of the wearable device, schedules the voice broadcast unit to implement task distribution and pushes audio to the terminal device, and generates wearable terminal broadcast record data; The medical order reminder recording module monitors the audio output buffer status, the number of voice interruptions, and whether the user confirmation action events are completely recorded during the broadcast task, based on the broadcast record data of the wearable terminal. It combines the response duration of the process and the instruction confirmation flag to generate a medical order reminder completion flag field.

[0036] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A reminder method for an AI voice-guided medical advice wearable device, characterized in that, Includes the following steps: S1: Based on the electronic medical order database, detect the continuous arrangement of instruction phrases and action phrases in the text, analyze the frequency and combination relationship of three types of keywords involving medication, examination, and rehabilitation behavior in the field position, and generate a set of medical order execution features; S2: Based on the set of medical order execution features, determine whether there are three semantic fragments, namely "immediate", "time-limited" and "dependent on intervention", in the execution frequency field and precaution field marked for each behavioral instruction; compare the level boundary thresholds, establish corresponding level labels, and generate medical order response level ranges; S3: Based on the medical order response level range, call the stimulus perception level range of the four types of voice identities (doctor, expert, family member, health worker) preset in the AI ​​voice identity library, filter the identity models that match the current range, and generate the reminder voice selection result; S4: Call the reminder voice selection result and the medical order execution feature set, match the scheduling queue of the voice broadcasting unit, the speaker audio channel control parameters and the unique device identifier of the wearable device, schedule the voice broadcasting unit to implement task distribution and push audio to the terminal device, and generate wearable terminal broadcasting record data.

2. The reminder method of the AI ​​voice medical order wearable device according to claim 1, characterized in that, The set of medical order execution features includes execution behavior category labels, behavior target content fragments, time-related fields, behavior timeliness classification identifiers, and precaution aggregation items. The medical order response level range includes response intensity level labels, time sensitivity level identifiers, and semantic urgency level identifiers. The reminder voice selection result includes voice identity type number, voice broadcast text segment, tone adjustment parameter group, speech rate matching index field, and broadcast time trigger field. The wearable terminal broadcast record data includes broadcast task status identifier code, device communication log item, and voice broadcast timing value.

3. The reminder method of the AI ​​voice medical order wearable device according to claim 2, characterized in that, The specific steps for obtaining the set of medical order execution features are as follows: S111: Based on the electronic medical order database, using the ID card number and medical record number in the user identity field, detect the mapping relationship between the identity identifier field and the record identifier field in the data index group, locate the data nodes with associated markers, call the medical order text data frame corresponding to the located data node, and generate the identity-bound medical order data frame. S112: Based on the text content field in the identity-bound medical order data frame, retrieve text fragments that begin with a verb phrase and end with a noun representing the action object. Combine the action verbs and the following nouns in the text fragments continuously, and identify the total number of occurrences and positional order of the combination patterns in the paragraph to obtain the action semantic arrangement feature group. S113: Based on the distribution information of behavioral verbs and the content of noun objects in the behavioral semantic arrangement feature group, call the predefined keyword classification dictionary, match the keyword grouping in each text segment with the field position relationship, determine whether the keyword belongs to the category of medication, examination or rehabilitation behavior, filter the text paragraph structure features that meet the dual matching conditions of keyword and classification category, and establish a set of medical order execution features.

4. The reminder method of the AI ​​voice medical order wearable device according to claim 3, characterized in that, The specific steps for obtaining the medical order response level range are as follows: S211: Based on the behavioral instruction field in the medical order execution feature set, extract the execution frequency field and precaution field bound to the behavioral instruction, detect whether the field includes immediate, time-limited and intervention-dependent semantic fragments, and mark and map the field labels with semantic expression to establish an emergency feature indication mapping set; S212: Based on the field marking results in the emergency feature indication mapping set, collect the corresponding behavioral instructions' preset severity level value, task load score, and response time limit value in the database, using the formula: ； The response level value is obtained through calculation. By comparing it with the level boundary threshold, the level range of each behavioral instruction is determined. Instructions belonging to the same range are classified and grouped to establish the medical order response level range. in, Indicates the severity level of the illness. This represents the arithmetic mean of the severity levels in the set of behavioral instructions. This represents the normalized value of the task load score. This represents the normalized value of the response time limit. Represents the semantic density coefficient. , , The first The normalized values ​​of the severity rating, task load score, and response time limit corresponding to each action instruction. Indicates the response level value.

5. The reminder method of the AI ​​voice medical order wearable device according to claim 4, characterized in that, The specific steps for obtaining the reminder voice selection result are as follows: S311: Based on the medical order response level range, call the AI ​​voice identity library, detect the stimulus perception level value mapped by the identity label field in the model group, and index and aggregate the stimulus value range of the voice model to establish the response level structure corresponding to the identity model and generate the identity stimulus level mapping set. S312: Call the combined level label field in the medical order response level range, compare the response level value marked by the field with the level range boundary corresponding to each identity model in the identity stimulus level mapping set, determine whether there is an overlapping range, extract the cross-matching voice identity model identifier, and establish an identity level matching structure group. S313: Based on the voice identity model identifier in the identity level matching structure group, extract the timbre spectrum index, speech rate control parameter field and broadcast intensity field corresponding to the matching model, and combine them with the semantic instruction content in the medical order execution feature set to perform text segment combination and broadcast parameter integration processing to establish the reminder voice selection result.

6. The reminder method of the AI ​​voice medical order wearable device according to claim 5, characterized in that, The specific steps for obtaining the wearable terminal broadcast record data are as follows: S411: Call the reminder voice selection result and the behavior action content and execution time field in the medical order execution feature set, retrieve the broadcast voice template segment corresponding to the behavior verb, extract the speech rate adjustment parameter and timbre index value that match the execution time and the speaker identity field, and splice and combine the parameters and semantic segments to generate a voice broadcast combination structure; S412: Based on the broadcast timestamp field and the voice identity index field in the voice broadcast combination structure, match the time channel mapping relationship, identity priority allocation index, and channel resource table entries in the voice broadcast unit scheduling table, and combine the registration status table of the wearable terminal identification code field to establish the identity time channel allocation mapping and obtain the broadcast task scheduling parameter set; S413: Call the identity, time, and device index fields in the broadcast task scheduling parameter set, execute the voice broadcast unit scheduling control, allocate broadcast resources according to the task parameters, start the corresponding terminal channel, push the combined audio command data to the audio output interface of the wearable device, and establish wearable terminal broadcast record data.

7. The reminder method of the AI ​​voice medical order wearable device according to claim 6, characterized in that, The method further includes the following steps: S5: Based on the broadcast record data of the wearable terminal, monitor whether the audio output buffer status, voice interruption number and user confirmation action event are completely recorded in the broadcast task, the response duration of the combination process and the instruction confirmation identifier, and generate a medical order reminder completion identifier field; The medical order reminder completion identifier field includes the broadcast status feedback code, the instruction response confirmation mark, and the broadcast task tracking number.

8. The reminder method of the AI ​​voice medical order wearable device according to claim 7, characterized in that, The specific steps for obtaining the medical order reminder completion identifier field are as follows: S511: Based on the broadcast recording data of the wearable terminal, detect the real-time status parameters of the audio output buffer unit in the broadcast task, monitor the synchronization stability between the buffer fill rate field and the output flow rate field, determine the continuity of the waveform frame, extract the record index value of the interruption or abnormal delay, and establish audio output status information. S512: Based on the interruption marker index in the audio output status information, retrieve the voice interruption count field and user confirmation action event field under the same batch of tasks, calculate the integrity ratio of the two types of field records, compare with the preset integrity threshold, filter the record sample set that meets the threshold condition, and generate a broadcast interaction integrity parameter set. S513: Call the time duration field and confirmation flag field in the broadcast interaction integrity parameter set, combine and encode the response duration and user confirmation flag in the task, write the encoding result into the task execution tracking index, construct a structured status mark record, and generate a medical order reminder completion flag field.

9. The reminder method of the AI ​​voice medical order wearable device according to claim 8, characterized in that, The integrity threshold is set by counting the recording frequency of the voice interruption count field and the user confirmation action event field in the same batch of broadcast task records, calculating the average and standard deviation of the recording frequency respectively, and setting the integrity threshold as the value corresponding to the sum of the average and standard deviation. When the recording frequency of both the voice interruption count field and the user confirmation action event field is greater than or equal to the integrity threshold, the corresponding record samples are selected and written into the broadcast interaction integrity parameter set.

10. A reminder system for an AI voice-guided wearable medical order device, characterized in that, The system is used to implement the reminder method of the AI ​​voice medical advice wearable device according to any one of claims 1-9, the system comprising: The execution feature extraction module, based on the electronic medical order database, detects the continuous arrangement of instruction phrases and behavioral action phrases in the text, analyzes the frequency and combination relationship of three types of keywords involving medication, examination, and rehabilitation behavior in the field positions, and generates a set of medical order execution features; The emergency assessment module for medical orders, based on the set of medical order execution features, determines whether there are three semantic fragments: "immediately," "time-limited," and "dependent on intervention" in the execution frequency field and precaution field marked for each behavioral instruction, compares the level boundary thresholds, establishes corresponding level labels, and generates a medical order response level range; The voice model filtering module, based on the medical order response level range, calls the stimulus perception level range of four preset voice identity labels (doctor, expert, family member, and health worker) in the AI ​​voice identity library, filters identity models that match the current range, and generates reminder voice selection results. The terminal broadcast execution module calls the reminder voice selection result and the medical order execution feature set, matches the scheduling queue of the voice broadcast unit, the speaker audio channel control parameters and the unique device identifier of the wearable device, schedules the voice broadcast unit to implement task distribution and pushes audio to the terminal device, and generates wearable terminal broadcast record data; The medical order reminder recording module, based on the broadcast recording data of the wearable terminal, monitors whether the audio output buffer status, the number of voice interruptions and user confirmation action events are completely recorded in the broadcast task, combines the response duration of the process and the instruction confirmation identifier, and generates a medical order reminder completion identifier field.

Images