Physiological index self-adaptive health assessment method and device driven by user feedback

Abstract

The invention relates to the technical field of physiological index data processing, in particular to a user feedback driven physiological index adaptive health assessment method and device. The method comprises the following steps: acquiring a BCG signal of a user based on a non-contact sensor, and acquiring index monitoring data of a plurality of indexes based on the BCG signal; the user inputs a user feedback text representing the abnormal state of the body, matching is conducted on the basis of the user feedback text and the disease words in the lexicon, and if the disease words are matched, indexes corresponding to the disease words serve as final monitoring indexes; if the disease word cannot be matched, screening out an initial monitoring index based on the user feedback text, and determining a final monitoring index based on the initial monitoring index; and obtaining daily index monitoring data based on the final monitoring index. In this way, mapping from coarse-grained subjective perception to refined objective indexes can be realized, and monitoring blind areas and misjudgment risks are remarkably reduced while the early-stage anomaly detection rate is increased.

Description

Technical Field

[0001] This invention relates to the field of physiological indicator data processing technology, specifically to a user feedback-driven adaptive health assessment method and apparatus for physiological indicators. Background Technology

[0002] With the rapid development of wearable devices and smart healthcare, non-contact physiological monitoring technology has become a core direction in the field of home health management due to its advantages of being non-invasive and capable of long-term continuous monitoring. Smart beds are equipped with sensors such as piezoelectric films, which collect human cardiac impaction (BCG) signals to achieve real-time extraction of multi-dimensional indicators such as heart rate, respiratory rate, heart rate variability (HRV), apnea events, arrhythmias, snoring, and body movement status. Monitoring can be completed solely through the micro-vibration of the bed, without requiring the user to wear the device.

[0003] While existing technologies have achieved the extraction and preliminary assessment of physiological parameters from BCG signals, they still have the following shortcomings: 1. Existing technologies are mostly limited to monitoring single diseases or isolated indicators (e.g., CN204169842U only provides heart rate alerts, and US6078261A only provides out-of-bed detection), making it difficult to cover multi-dimensional abnormal states (e.g., coronary heart disease accompanied by stress and sleep disorders). User subjective sensations (e.g., chest tightness, fatigue) cannot be accurately mapped to objective indicators, lacking keyword parsing and closed-loop response mechanisms, creating monitoring blind spots. 2. Monitoring indicators are mostly fixed configurations and cannot be dynamically adjusted based on real-time user sensations (e.g., by selecting from an indicator library). Warning thresholds rely on group statistics, failing to consider individual differences (e.g., age, underlying diseases, lifestyle habits) and health status migration, leading to over-warning or under-warning. Existing technologies lack forward-looking optimization, resulting in low matching between indicators and user sensations. 3. Existing technologies have not established a hierarchical process of "preliminary assessment - steady-state determination - continuous tracking," and indicator thresholds remain static, unable to adapt to the evolution of health states (e.g., from abnormal to a new steady state). The lack of daily trend forecasts and interpretable feedback makes it difficult to achieve adaptive baseline updates and indicator re-optimization, thus hindering the formation of a complete closed loop of "monitoring-evaluation-intervention". Summary of the Invention

[0004] The purpose of this invention is to provide a user feedback-driven adaptive health assessment method and device for physiological indicators, which can realize the mapping from coarse-grained subjective perception to refined objective indicators, thereby improving the early detection rate of abnormalities and significantly reducing monitoring blind spots and the risk of misjudgment.

[0005] In a first aspect of the embodiments of the present invention, a user feedback-driven adaptive health assessment method for physiological indicators is provided, comprising:

[0006] The user's BCG signal is collected based on a non-contact sensor, and the monitoring data of multiple indicators are obtained based on the BCG signal.

[0007] The system inputs user feedback text representing abnormal physical conditions. The system matches the user feedback text with disease terms in the lexicon. If a disease term is matched, the indicator corresponding to the disease term is used as the final monitoring indicator. If no disease term is matched, the system filters out initial monitoring indicators based on the user feedback text and determines the final monitoring indicator based on the initial monitoring indicators.

[0008] Daily indicator monitoring data is obtained based on the final monitoring indicators. Indicator trend vector sequences are obtained based on the indicator monitoring data. Daily predicted health status is determined based on the indicator trend vector sequences. Daily explanatory prompts are generated based on the predicted health status.

[0009] As a preferred embodiment of the present invention, matching user feedback text with disease terms in the lexicon specifically includes:

[0010] Step 11. Obtain the semantic vector of the user feedback text based on the BioBERT model;

[0011] Step 12. Obtain the semantic vector of each symptom word;

[0012] Step 13. Calculate the semantic similarity value between semantic vector one and each semantic vector two;

[0013] Step 14. Determine the maximum semantic similarity value from all semantic similarity values;

[0014] Step 15. Determine whether the maximum semantic similarity value is greater than or equal to the first threshold. If the maximum semantic similarity value is greater than or equal to the first threshold, the user feedback text matches the disease word; otherwise, the user feedback text does not match the disease word.

[0015] As a preferred embodiment of the present invention, the initial monitoring indicators are selected based on user feedback text, specifically including:

[0016] Step 21. Obtain an indicator from the indicator library, obtain the third semantic vector of the indicator, and calculate the second semantic similarity value between the first semantic vector and the third semantic vector;

[0017] Step 22. Obtain the indicator monitoring data for the most recent N days, and calculate the current trend slope, current fluctuation range, and current number of jump points of the indicator based on the indicator monitoring data;

[0018] Step 23. Obtain the historical indicator monitoring data of the indicator over the past M days, group the historical indicator monitoring data of the indicator by N days, and calculate multiple historical trend slopes, multiple historical fluctuation amplitudes and multiple historical jump points of the indicator based on multiple groups of historical indicator monitoring data;

[0019] Step 24. Calculate the comprehensive normalized volatility intensity value of the indicator based on the indicator's current trend slope, current volatility amplitude, current number of jump points, and multiple historical trend slopes, multiple historical volatility amplitudes, and multiple historical number of jump points.

[0020] Step 25. Calculate the index score based on the semantic similarity value and the comprehensive normalized fluctuation intensity value of the index;

[0021] Repeat steps 21 to 25 until an indicator score has been calculated for each indicator in the indicator library;

[0022] Step 26. Select several indicators based on their scores to serve as initial monitoring indicators.

[0023] As a preferred embodiment of the present invention, step 24 specifically includes:

[0024] Step 241. Calculate the mean slope and standard deviation of the slope based on multiple historical trend slopes, calculate the mean amplitude and standard deviation of the amplitude based on multiple historical fluctuation amplitudes, and calculate the mean number of points and standard deviation of the number of points based on multiple historical jump points;

[0025] Step 242. Calculate the slope deviation based on the current trend slope and the average slope, calculate the amplitude deviation based on the current fluctuation amplitude and the average amplitude, and calculate the point deviation based on the current number of jump points and the average number of points;

[0026] Step 243. Calculate the normalized slope deviation based on the slope deviation and the slope standard deviation, calculate the normalized amplitude deviation based on the amplitude deviation and the amplitude standard deviation, and calculate the normalized point deviation based on the point deviation and the point standard deviation.

[0027] Step 244. Calculate the comprehensive normalized fluctuation intensity value of the index based on the normalized slope deviation, normalized amplitude deviation, and normalized point number deviation.

[0028] As a preferred embodiment of the present invention, step 25 specifically calculates the index score using the following formula:

[0029] , in, The score for a certain indicator. The preset weight value for semantic similarity value two. The semantic similarity value of a certain indicator is two. The preset weight values ​​are used to comprehensively normalize the fluctuation intensity values. This represents the comprehensive normalized volatility intensity value of a certain indicator.

[0030] As a preferred embodiment of the present invention, determining the final monitoring indicators based on the initial monitoring indicators specifically includes:

[0031] Step 31. Obtain monitoring data for subsequent indicators based on the initial monitoring indicators;

[0032] Step 32. Determine whether the user has not entered a daily symptom severity score for Q consecutive days. If so, use the initial monitoring indicator as the final monitoring indicator; otherwise, proceed to step 33.

[0033] Step 33. Obtain the comprehensive objective function value based on subsequent indicator monitoring data and the daily symptom severity score input by the user. If the comprehensive objective function value cannot be obtained, return to step 31; otherwise, proceed to step 34.

[0034] Step 34. Determine whether the comprehensive objective function value is greater than or equal to the second threshold. If the comprehensive objective function value is greater than or equal to the second threshold, the initial monitoring indicator is used as the final monitoring indicator. If the comprehensive objective function value is less than the third threshold or less than the historical maximum comprehensive objective function value, one of the initial monitoring indicators is replaced to obtain a new initial monitoring indicator, and the process returns to step 31. Otherwise, the process returns directly to step 31.

[0035] As a preferred embodiment of the present invention, step 33 specifically includes:

[0036] Step 331. Obtain the indicator change sequence based on the subsequent indicator monitoring data for at least Y consecutive days, form the symptom score sequence based on the daily symptom severity score values ​​for at least Y consecutive days, obtain the Spearman rank correlation coefficient based on the indicator change sequence and the symptom score sequence, and use the Spearman rank correlation coefficient as the diurnal fluctuation consistency value.

[0037] Step 332. Based on subsequent indicator monitoring data and regression model, obtain the predicted score value, determine the actual score value based on the obtained daily symptom severity score value, obtain the total sum of squares and residual sum of squares based on the predicted score value and the actual score value, obtain the coefficient of determination based on the total sum of squares and residual sum of squares, and use the coefficient of determination as the physiological explanatory value.

[0038] Step 333. Obtain the correlation coefficient between indicators based on subsequent indicator monitoring data, and obtain the indicator diversity value based on the correlation coefficient between indicators;

[0039] Step 334. Calculate the comprehensive objective function value based on the consistency value of diurnal fluctuations, the physiological explanatory power value, and the indicator diversity value.

[0040] As a preferred embodiment of the present invention, step 34, which involves replacing one of the initial monitoring indicators to obtain a new initial monitoring indicator, specifically includes:

[0041] Step 341. Remove the indicator with the lowest contribution to the comprehensive objective function value from the initial monitoring indicators to obtain the remaining indicators;

[0042] Step 342. Remove the indicators that are the same as the initial monitoring indicators from the indicator library or the remaining indicator library to obtain the remaining indicator library, and select the indicators with the highest indicator scores in the remaining indicator library as candidate indicators.

[0043] Step 343. Based on the monitoring data of the retained indicators and each candidate indicator, several comprehensive objective function values ​​are calculated. The candidate indicator corresponding to the largest comprehensive objective function value is combined with the retained indicators to obtain new initial monitoring indicators.

[0044] As a preferred embodiment of the present invention, obtaining the indicator trend vector sequence based on indicator monitoring data specifically includes:

[0045] Step 41. Obtain the monitoring data of a specific indicator from the final monitoring indicators for the previous 3 days;

[0046] Step 42. Calculate the velocity value, acceleration value, and volatility based on the indicator monitoring data from the previous 3 days;

[0047] Step 43. Obtain the indicator monitoring data for the previous 90 days for the corresponding indicator, and obtain the historical standard deviation based on the indicator monitoring data for the previous 90 days;

[0048] Step 44. Obtain normalized velocity, normalized acceleration, and normalized volatility based on velocity, acceleration, volatility, and historical standard deviation, and use these values ​​as a trend vector for the corresponding indicators.

[0049] Repeat steps 41 to 44 to obtain multiple indicator trend vectors for the corresponding indicator, and use the multiple indicator trend vectors as the indicator trend vector sequence for the corresponding indicator.

[0050] In a second aspect of the present invention, a user feedback-driven adaptive health assessment device for physiological indicators is provided, comprising:

[0051] The indicator monitoring data acquisition module is configured to collect the user's BCG signal based on a non-contact sensor, and acquire indicator monitoring data for multiple indicators based on the BCG signal.

[0052] The final monitoring indicator determination module is configured to accept user feedback text representing abnormal physical conditions as input by the user; it matches the user feedback text with disease words in the thesaurus; if a disease word is matched, the indicator corresponding to the disease word is used as the final monitoring indicator; if no disease word is matched, the initial monitoring indicators are filtered based on the user feedback text, and the final monitoring indicator is determined based on the initial monitoring indicators.

[0053] The explanatory prompt generation module is configured to obtain daily indicator monitoring data based on the final monitoring indicators, obtain indicator trend vector sequences based on the indicator monitoring data, determine the daily predicted health status based on the indicator trend vector sequences, and generate daily explanatory prompts based on the predicted health status.

[0054] In summary, the present invention has the following beneficial effects:

[0055] 1. The method described in this specification addresses the challenge of health risk identification caused by vague subjective descriptions from users, biased initial feedback, or atypical symptoms. Through a hierarchical, progressive strategy combining semantic initial screening and dynamic optimization, the BioBERT model is first used to perform high-confidence symptom word matching on user sensations to quickly obtain the final monitoring indicators. If no match is found, initial monitoring indicators are obtained by fusing semantic similarity and recent physiological fluctuation features from the full indicator library. Inefficient indicators are continuously iterated and replaced based on a comprehensive objective function value, ultimately adaptively locking in the final monitoring indicators. This method achieves a mapping from coarse-grained subjective perception to refined objective indicators, significantly reducing monitoring blind spots and misjudgment risks while improving the early anomaly detection rate, providing reliable technical support for personalized and dynamic digital health early warning.

[0056] 2. The method described in this specification constructs a phased, closed-loop adaptive health state evolution prediction system. After the final monitoring indicators are locked, starting from the day of user feedback, a 3-day sliding window is used to extract the trend vector of each indicator. The trend vector sequence of the indicators is input into the GRU model to obtain the daily predicted health status. At the same time, the contribution of the indicators and the actual data changes are integrated to generate interpretable prompts for users. This method abandons the static judgment logic of traditional fixed thresholds. Through the closed loop of "daily prediction - interpretation intervention - steady state migration", the health monitoring system has the ability to continuously learn, dynamically calibrate and adapt in the long term, which significantly improves the timeliness, personalization and clinical credibility of the early warning.

[0057] Further or more detailed beneficial effects will be described in conjunction with specific embodiments in the detailed implementation. Attached Figure Description

[0058] Figure 1 A flowchart of a user feedback-driven adaptive health assessment method for physiological indicators according to an embodiment of the present invention is shown;

[0059] Figure 2 A block diagram of a user feedback-driven adaptive health assessment device for physiological indicators according to an embodiment of the present invention is shown. Detailed Implementation

[0060] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the invention. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the invention.

[0061] In the description of embodiments of the present invention, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0062] Figure 1 A flowchart of a user feedback-driven adaptive health assessment method 100 based on physiological indicators, according to an embodiment of the present invention, is shown. The method 100 includes:

[0063] 102. Collect user's BCG signal based on non-contact sensor, and obtain indicator monitoring data of multiple indicators based on the BCG signal.

[0064] Taking user a as an example, the device can collect user a's BCG signal each night while sleeping (user a only needs to lie on the smart bed and sleep normally; the smart bed is part of the device). In this embodiment, user a's BCG signal for one night is used as a set of raw data. Each time the device acquires a set of raw data, it processes the raw data and calculates monitoring data for multiple indicators such as heart rate, respiratory rate, heart rate variability (HRV) (time domain and frequency domain), sleep stages, heart rate fluctuations, and respiratory fluctuations. For ease of understanding, this embodiment assumes there are a total of 10 indicators (indicator A, indicator B, indicator C... indicator J). Each time the device acquires a set of raw data, it will obtain a set of indicator monitoring data for indicator A, a set of indicator monitoring data for indicator B, a set of indicator monitoring data for indicator C... a set of indicator monitoring data for indicator J.

[0065] For each night user A sleeps in the smart bed, there will be several sets of raw data, corresponding to several sets of indicator monitoring data for indicators A, B, C, ..., J. As long as user A uses the smart bed to sleep every day, a new set of raw data will be added each day, which in turn will add a new set of indicator monitoring data for indicator A, a new set of indicator monitoring data for indicator B, a new set of indicator monitoring data for indicator C, ..., a new set of indicator monitoring data for indicator J.

[0066] 104. The user inputs user feedback text representing an abnormal physical state. The user feedback text is matched with disease words in the thesaurus. If a disease word is matched, the indicator corresponding to the disease word is used as the final monitoring indicator. If no disease word is matched, the initial monitoring indicators are selected based on the user feedback text, and the final monitoring indicators are determined based on the initial monitoring indicators.

[0067] Suppose on a certain day (e.g., October 15th), user A feels unwell. User A can input their discomfort in text form into a mobile app (which is also part of the device). For example, user A might input, "For the past three days, I've had chest tightness, a very fast heartbeat, and it beats erratically in waves, which is more noticeable when lying down" (this text is user feedback describing the abnormal physical condition). Once the device receives user feedback, it will match the feedback text with symptom terms in its lexicon. If a symptom term is matched, the corresponding indicator will be used as the final monitoring indicator.

[0068] In this embodiment, matching user feedback text with disease terms in the lexicon specifically includes:

[0069] Step 11. Obtain the semantic vector one of the user feedback text based on the BioBERT model. In this embodiment, the BioBERT model directly adopts existing technology. In this step, it is only necessary to input the user feedback text into the BioBERT model, and the BioBERT model will automatically output a semantic vector one.

[0070] Step 12. Obtain the semantic vector II for each disease term. For ease of understanding, this embodiment assumes that there are 5 disease terms in the vocabulary, and the semantic vector II for each disease term has been determined in advance by the BioBERT model. This step only requires taking out the 5 semantic vector II directly.

[0071] Step 13. Calculate the semantic similarity value 1 between semantic vector 1 and each semantic vector 2. This step calculates the cosine similarity between semantic vector 1 and semantic vector 2, which is the semantic similarity value 1. Assume the semantic similarity value 1 calculated between semantic vector 1 and the semantic vector 2 of the first symptom word is 0.55; the semantic similarity value 1 calculated between semantic vector 1 and the semantic vector 2 of the second symptom word is 0.66; the semantic similarity value 1 calculated between semantic vector 1 and the semantic vector 2 of the third symptom word is 0.33; the semantic similarity value 1 calculated between semantic vector 1 and the semantic vector 2 of the fourth symptom word is 0.77; and the semantic similarity value 1 calculated between semantic vector 1 and the semantic vector 2 of the fifth symptom word is 0.81.

[0072] Step 14. Determine the maximum semantic similarity value from all semantic similarity values. Assume that in this step, the maximum semantic similarity value is the semantic similarity value calculated between semantic vector one and semantic vector two of the 5th symptom word (i.e., 0.81).

[0073] Step 15. Determine whether the maximum semantic similarity value is greater than or equal to the first threshold. If the maximum semantic similarity value is greater than or equal to the first threshold, the user feedback text matches the symptom word; otherwise, the user feedback text does not match the symptom word. In this embodiment, the first threshold can be 0.85. Since the maximum semantic similarity value calculated in step 14 is 0.81, the maximum semantic similarity value is less than the first threshold, thus determining that the user feedback text does not match the symptom word.

[0074] Each symptom term in the lexicon has several pre-defined indicators. Let's assume the pre-defined indicators for the fourth symptom term are indicators B and D. If the user feedback text matches the fourth symptom term in the lexicon, then the final monitoring indicators will be indicators B and D. If the user feedback text does not match any of the symptom terms in the lexicon, then step 104 needs to filter out initial monitoring indicators based on the user feedback text.

[0075] In this embodiment, the initial monitoring indicators selected based on user feedback text specifically include:

[0076] Step 21. Obtain one indicator from the indicator library, obtain the semantic vector three of the indicator, and calculate the semantic similarity value two between semantic vector one and semantic vector three. In this embodiment, there are 10 indicators in the indicator library, and the semantic vector three of each indicator has been determined in advance using the BioBERT model. This step only requires directly using the semantic vector three from the indicator library. Assuming that the first indicator A in the indicator library is obtained at this time, then the semantic similarity value two can be obtained by calculating the cosine similarity between the semantic vector three of indicator A and the semantic vector one of the user feedback text.

[0077] Step 22. Obtain the indicator monitoring data for the most recent N days, and calculate the current trend slope, current fluctuation range, and current number of jump points of the indicator based on the indicator monitoring data.

[0078] This step obtains the monitoring data of indicator A for the most recent 7 days (i.e., from October 9th to October 15th). The monitoring data of indicator A has already been obtained in step 102, so this step only needs to use it.

[0079] The current trend slope for indicator A can be calculated using the following formula:

[0080] , in, The current trend slope of indicator A. The variable is the number of days (t can be day 1, day 2, ..., day 7). This is the average of the total number of days (for example, the average total number of days for the last 7 days is 4). Let A be the specific data value for indicator A on day t. This represents the average data of indicator A over the past 7 days.

[0081] The current fluctuation range of indicator A can be calculated using the following formula:

[0082] , in, This represents the current fluctuation range of indicator A. The variable is the number of days (t can be day 1, day 2, ..., day 7). Let A be the specific data value for indicator A on day t. This represents the average data of indicator A over the past 7 days.

[0083] For the current jump points of indicator A, first obtain the standard deviation of the indicator monitoring data based on the indicator monitoring data of the most recent 7 days. Then determine whether the absolute value of the difference between the monitoring data of day 2 and the monitoring data of day 1 (if the monitoring data of indicator A has only 1 value per day, then simply subtract the monitoring data value of day 1 from the monitoring data value of day 2; if the monitoring data of indicator A has multiple values ​​per day, then first calculate the average value of the monitoring data of each day, and then subtract the average value of the monitoring data of day 1 from the average value of the monitoring data of day 2) is greater than 3 times the standard deviation. If it is greater than 3 times the standard deviation, then the cumulative jump points (the initial cumulative jump points are 0) are incremented by 1; otherwise, the cumulative jump points remain unchanged. Next, it determines whether the absolute value of the difference between the monitoring data on day 3 and the monitoring data on day 2 is greater than 3 times the standard deviation. If it is greater than 3 times the standard deviation, the cumulative jump point is incremented by 1; otherwise, the cumulative jump point remains unchanged. Finally, it determines whether the absolute value of the difference between the monitoring data on day 7 and the monitoring data on day 6 is greater than 3 times the standard deviation. If it is greater than 3 times the standard deviation, the cumulative jump point is incremented by 1; otherwise, the cumulative jump point remains unchanged. The final cumulative jump point is the current jump point of indicator A.

[0084] Step 23. Obtain the historical indicator monitoring data of the indicator over the past M days, group the historical indicator monitoring data of the indicator into N days, and calculate multiple historical trend slopes, multiple historical fluctuation amplitudes and multiple historical jump points of the indicator based on multiple groups of historical indicator monitoring data.

[0085] This step obtains the historical indicator monitoring data for indicator A over the past 90 days (i.e., from July 17 to October 15). Then, the historical indicator monitoring data for indicator A is grouped into 7-day groups. Specifically, the historical indicator monitoring data from July 17 to July 23 can be grouped into one group, the historical indicator monitoring data from July 18 to July 24 can be grouped into one group, the historical indicator monitoring data from July 19 to July 25 can be grouped into one group, and so on, with the historical indicator monitoring data from October 9 to October 15 grouped into another group.

[0086] For each set of historical indicator monitoring data, one corresponding historical trend slope, one historical fluctuation range, and one historical jump point are calculated. The specific calculation method is the same as that in step 22. Assuming there are a total of 100 sets of historical indicator monitoring data, then 100 historical trend slopes, 100 historical fluctuation ranges, and 100 historical jump points can be obtained.

[0087] Step 24. Calculate the comprehensive normalized volatility intensity value of the indicator based on the indicator's current trend slope, current volatility amplitude, current number of jump points, and multiple historical trend slopes, multiple historical volatility amplitudes, and multiple historical number of jump points.

[0088] In this embodiment, step 24 specifically includes:

[0089] Step 241. Calculate the mean slope and standard deviation of the slope based on multiple historical trend slopes, the mean amplitude and standard deviation of the amplitude based on multiple historical fluctuation amplitudes, and the mean number of points and standard deviation of the number of points based on multiple historical jump points. For the mean slope, assuming 100 historical trend slopes were obtained in Step 23, the mean slope in this step is simply obtained by summing the 100 historical trend slopes and then dividing by 100. For the standard deviation of the slope, simply obtain the squared deviation of each historical trend slope from the mean slope, then average all the squared deviations and take the square root to obtain the standard deviation of the slope. The calculation methods for the mean amplitude and the mean number of points are the same as those for the mean slope, and the calculation methods for the standard deviation of amplitude and the standard deviation of the number of points are the same as those for the standard deviation of the slope.

[0090] Step 242. Calculate the slope deviation based on the current trend slope and the average slope, the amplitude deviation based on the current fluctuation amplitude and the average amplitude, and the point deviation based on the current number of jump points and the average number of points. This step simply involves subtracting the average slope from the current trend slope and taking the absolute value to obtain the slope deviation. Similarly, this step involves subtracting the average amplitude from the current fluctuation amplitude and taking the absolute value to obtain the amplitude deviation, and subtracting the average number of points from the current number of jump points and taking the absolute value to obtain the point deviation.

[0091] Step 243. Calculate the normalized slope deviation based on the slope deviation and slope standard deviation; calculate the normalized amplitude deviation based on the amplitude deviation and amplitude standard deviation; and calculate the normalized point deviation based on the point deviation and point standard deviation. This step simply requires dividing the slope deviation by the slope standard deviation to obtain the normalized slope deviation. Similarly, this step requires dividing the amplitude deviation by the amplitude standard deviation to obtain the normalized amplitude deviation, and dividing the point deviation by the point standard deviation to obtain the normalized point deviation.

[0092] Step 244. Calculate the comprehensive normalized volatility value of the index based on the normalized slope deviation, normalized amplitude deviation, and normalized point number deviation. In this step, it is only necessary to first sum the normalized slope deviation, normalized amplitude deviation, and normalized point number deviation, and then divide by 3 to obtain the comprehensive normalized volatility value of index A.

[0093] Step 25. Calculate the indicator score based on the semantic similarity value 2 and the comprehensive normalized volatility value of indicator A. At this point, we know the semantic similarity value 2 and the comprehensive normalized volatility value of indicator A. Then, we calculate the indicator score using the following formula:

[0094] , in, The score for a certain indicator. The preset weight value for semantic similarity value two. The semantic similarity value of a certain indicator is two. The preset weight values ​​are used to comprehensively normalize the fluctuation intensity values. This represents the comprehensive normalized fluctuation intensity value of a certain indicator. In this embodiment, and The values ​​are all 0.5, and the score of indicator A can be calculated using this formula.

[0095] Repeat steps 21 through 25 until a score has been calculated for each indicator in the indicator library. For other indicators in the indicator library, simply repeat steps 21 through 25 to obtain their scores.

[0096] Step 26. Select several indicators as initial monitoring indicators based on their scores. In this embodiment, the three indicators with the highest scores can be selected as initial monitoring indicators. Assuming that two of the three initial monitoring indicators have a correlation coefficient greater than 0.8, the indicator with the lower score is removed. Then, the indicator with the highest score is selected from the remaining indicators in the indicator library to recombine into three initial monitoring indicators. This process continues until the correlation coefficient between any two of the three initial monitoring indicators is less than or equal to 0.8. For example, if the correlation coefficient between the first indicator (with the highest score) and the third indicator (with the third highest score) is greater than 0.8, the third indicator is removed. Then, the fourth highest score indicator is selected from the indicator library to recombine into three initial monitoring indicators. If the correlation coefficient between any two of the three recombine initial monitoring indicators is less than or equal to 0.8, then the selection of the three initial monitoring indicators is complete. This embodiment assumes that the three initial monitoring indicators finally selected are Indicator A, Indicator C, and Indicator E.

[0097] Step 104: After selecting the initial monitoring indicators, it is also necessary to determine the final monitoring indicators based on the initial monitoring indicators.

[0098] In this embodiment, determining the final monitoring indicators based on the initial monitoring indicators specifically includes:

[0099] Step 31. Obtain monitoring data for subsequent indicators based on the initial monitoring indicators.

[0100] In this embodiment, after October 15th, the device will still obtain the monitoring data of each indicator through step 102. This step only needs to obtain the monitoring data of indicator A, indicator C, and indicator E as the subsequent monitoring data.

[0101] Step 32. Determine if the user has not entered a daily symptom severity score for Q consecutive days. If so, use the initial monitoring indicator as the final monitoring indicator; otherwise, proceed to Step 33. This step could be to determine if the user has not entered a daily symptom severity score for 3 consecutive days. Normally, after October 15th, user A needs to enter a daily symptom severity score through the mobile app. Specifically, based on user A's feedback text, "I've had chest tightness for the past three days, my heart is beating very fast, and it's erratic in waves, especially when lying down," user A needs to provide a symptom severity score for each day after October 15th. The score can be from 1 to 10, with a higher score indicating more severe symptoms.

[0102] Scenario 1: Suppose that on October 17th, user A entered the daily symptom severity score for October 16th. In this case, this step determines that the user has not failed to enter a daily symptom severity score for three consecutive days, so proceed to step 33.

[0103] Scenario 2: Suppose it's just the beginning of October 18th, and user A did not enter the daily symptom severity score for October 16th and 17th. In this case, this step determines that the user has not failed to enter the daily symptom severity score for three consecutive days, so proceed to step 33.

[0104] Scenario 3: Suppose that on October 19th, user A did not input daily symptom severity scores for October 16th, 17th, and 18th. In this case, this step determines that the user has not input daily symptom severity scores for three consecutive days, so the initial monitoring indicators (indicators A, C, and E) are directly used as the final monitoring indicators.

[0105] Scenario 4: Suppose that on October 19th, user A entered daily symptom severity scores for October 16th, 17th, and 18th. In this case, this step determines that the user has not failed to enter daily symptom severity scores for three consecutive days, so proceed to step 33.

[0106] Step 33. Obtain the comprehensive objective function value based on subsequent indicator monitoring data and the daily symptom severity score input by the user. If the comprehensive objective function value cannot be obtained, return to step 31; otherwise, proceed to step 34.

[0107] In this embodiment, step 33 specifically includes:

[0108] Step 331. Obtain the indicator change sequence based on at least Y consecutive days of subsequent indicator monitoring data; form a symptom score sequence based on at least Y consecutive days of daily symptom severity scores; obtain the Spearman rank correlation coefficient based on the indicator change sequence and the symptom score sequence, and use the Spearman rank correlation coefficient as the diurnal volatility consistency value. In this embodiment, the value of Y can be 3.

[0109] Scenario 1: Suppose we have just entered October 17th. At this time, we have only obtained the monitoring data of one day (i.e., October 16th), so we cannot obtain the sequence of indicator changes. Therefore, we cannot calculate the comprehensive objective function value. So we directly return to step 31.

[0110] Scenario 2: Assuming it's just the beginning of October 19th, we have acquired monitoring data for three consecutive days (October 16th, 17th, and 18th). We further assume that user a entered a daily symptom severity score on October 16th and 17th, but not on October 18th. Therefore, we cannot obtain the symptom score sequence, and consequently, we cannot calculate the comprehensive objective function value. So, we directly return to step 31.

[0111] Scenario 3: Suppose it is just the beginning of October 19th. At this time, we have obtained subsequent indicator monitoring data for three consecutive days (October 16th, October 17th and October 18th). Furthermore, we assume that user A input daily symptom severity scores on October 16th, October 17th and October 18th, so we can obtain the symptom score sequence.

[0112] For indicator A, simply subtract the indicator monitoring data value of October 15 from the indicator monitoring data value of October 16 (if there are multiple indicator monitoring data values ​​for each day, process them (e.g., calculate the average) to obtain a final indicator monitoring data value), subtract the indicator monitoring data value of October 16 from the indicator monitoring data value of October 17, and subtract the indicator monitoring data value of October 17 from the indicator monitoring data value of October 18 to obtain the indicator change sequence for indicator A. Similarly, for indicator C, simply subtract the indicator monitoring data value of October 15 from the indicator monitoring data value of October 16, subtract the indicator monitoring data value of October 16 from the indicator monitoring data value of October 17, and subtract the indicator monitoring data value of October 17 from the indicator monitoring data value of October 18 to obtain the indicator change sequence for indicator C. For indicator E, we only need to subtract the indicator monitoring data value of October 15 from the indicator monitoring data value of October 16, subtract the indicator monitoring data value of October 16 from the indicator monitoring data value of October 17, and subtract the indicator monitoring data value of October 17 from the indicator monitoring data value of October 18 to obtain the indicator change sequence of indicator E.

[0113] First, the absolute intensity of change in indicator A is calculated based on the sequence of changes in indicator A (i.e., the sum of squares of the values ​​in the sequence of changes in indicator A, followed by the square root). Simultaneously, the absolute intensity of change in indicator C is calculated based on the sequence of changes in indicator C, and the absolute intensity of change in indicator E is calculated based on the sequence of changes in indicator E. Then, the absolute intensity of change in indicators A, C, and E is used as a sequence and compared with the symptom score sequence to calculate the Spearman rank correlation coefficient. This Spearman rank correlation coefficient is used as the concurrency consistency value for diurnal fluctuations in this step.

[0114] Step 332. Obtain the predicted score based on subsequent indicator monitoring data and regression model, determine the actual score based on the obtained daily symptom severity score, obtain the total sum of squares and residual sum of squares based on the predicted score and the actual score, obtain the coefficient of determination based on the total sum of squares and residual sum of squares, and use the coefficient of determination as the physiological explanatory value.

[0115] The regression model in this step is pre-set. This step only requires inputting the subsequent monitoring data for indicators A, C, and E into the regression model to obtain the predicted score values. Inputting the subsequent monitoring data from October 16th into the regression model will yield the predicted score value for October 16th; inputting the subsequent monitoring data from October 17th will yield the predicted score value for October 17th; inputting the subsequent monitoring data from October 18th will yield the predicted score value for October 18th. The daily symptom severity score value entered by the user on October 16th is determined as the actual score value for October 16th; the daily symptom severity score value entered by the user on October 17th is determined as the actual score value for October 17th; the daily symptom severity score value entered by the user on October 18th is determined as the actual score value for October 18th.

[0116] The sum of squares of the differences between the predicted score on October 16th and the actual score on October 16th is calculated. The sum of squares of the differences between the predicted score on October 17th and the actual score on October 17th is calculated. The sum of squares of the differences between the predicted score on October 18th and the actual score on October 18th is calculated. Finally, the sum of the three squared values ​​is calculated to obtain the residual sum of squares.

[0117] The average score is calculated by averaging the actual scores for October 16th, October 17th, and October 18th. Then, the difference between the actual score for October 16th and the average score is squared, the difference between the actual score for October 17th and the average score is squared, and the difference between the actual score for October 18th and the average score is squared. Finally, the three squared values ​​are added together to obtain the total sum of squares.

[0118] Finally, subtracting the sum of squared residuals from 1 and dividing by the sum of squared totals gives the coefficient of determination. This step uses the coefficient of determination as the physiological explanatory value.

[0119] Step 333. Obtain the correlation coefficient between indicators based on subsequent indicator monitoring data, and obtain the indicator diversity value based on the correlation coefficient between indicators.

[0120] First, calculate the Spearman rank correlation coefficient between the subsequent monitoring data of indicator A and indicator C to obtain correlation coefficient one. Simultaneously, calculate the Spearman rank correlation coefficient between the subsequent monitoring data of indicator A and indicator E to obtain correlation coefficient two, and calculate the Spearman rank correlation coefficient between the subsequent monitoring data of indicator C and indicator E to obtain correlation coefficient three. Then, average the absolute values ​​of correlation coefficient one, correlation coefficient two, and correlation coefficient three to obtain the average correlation coefficient. Finally, subtract the average correlation coefficient from 1 to obtain the indicator diversity value for this step.

[0121] Step 334. Calculate the comprehensive objective function value based on the consistency value of diurnal fluctuations, the physiological explanatory power value, and the indicator diversity value.

[0122] In this step, the overall objective function value equals the intraday volatility consistency value multiplied by its weight, plus the physiological explanatory power value multiplied by its weight, plus the indicator diversity value multiplied by its weight. In this step, the weight of the intraday volatility consistency value can be 0.5, the weight of the physiological explanatory power value can be 0.3, and the weight of the indicator diversity value can be 0.2.

[0123] Step 34. Determine whether the comprehensive objective function value is greater than or equal to the second threshold. If the comprehensive objective function value is greater than or equal to the second threshold, the initial monitoring indicator is used as the final monitoring indicator. If the comprehensive objective function value is less than the third threshold or less than the historical maximum comprehensive objective function value, one of the initial monitoring indicators is replaced to obtain a new initial monitoring indicator, and the process returns to step 31. Otherwise, the process returns directly to step 31.

[0124] In this step, the second threshold can be 0.75, and the third threshold can be 0.6. If the comprehensive objective function value calculated in step 33 is greater than or equal to 0.75, then the initial monitoring indicators are directly used as the final monitoring indicators, i.e., indicators A, C, and E are used as the final monitoring indicators. If the comprehensive objective function value calculated in step 33 is less than 0.6 or the historical maximum comprehensive objective function value (assuming this is the first execution of step 34, then there is no historical maximum comprehensive objective function value; if this is the second execution of step 34, then the comprehensive objective function value calculated in the first execution is the historical maximum comprehensive objective function value; if this is the fifth execution of step 34, then the largest comprehensive objective function value among the comprehensive objective function values ​​calculated in the 1st, 2nd, 3rd, and 4th executions is used as the historical maximum comprehensive objective function value), then one of the initial monitoring indicators is replaced to obtain a new initial monitoring indicator, and then the process returns to step 31. If the comprehensive objective function value calculated in step 33 is neither greater than or equal to 0.75, nor less than 0.6, nor the historical maximum comprehensive objective function value, then the process directly returns to step 31.

[0125] In this embodiment, step 34, which involves replacing one of the initial monitoring indicators to obtain a new initial monitoring indicator, specifically includes:

[0126] Step 341. Remove the indicator with the lowest contribution to the overall objective function value from the initial monitoring indicators to obtain the remaining indicators. Assume the initial monitoring indicators are indicator A, indicator C, and indicator E. If removing indicator A has the least impact on the overall objective function value, then indicator A is the indicator with the lowest contribution to the overall objective function value. The remaining indicators after removing indicator A are indicator C and indicator E.

[0127] Step 342. Remove indicators from the indicator library or the remaining indicator library that are identical to the initial monitoring indicators to obtain the remaining indicator library. Select the indicators with the highest scores in the remaining indicator library as candidate indicators. The original indicator library contained 10 indicators: Indicator A, Indicator B, Indicator C, Indicator D, Indicator E, Indicator F, Indicator G, Indicator H, Indicator I, and Indicator J. After removing the initial monitoring indicators (Indicator A, Indicator C, and Indicator E) from the indicator library, the remaining indicator library is obtained: Indicator B, Indicator D, Indicator F, Indicator G, Indicator H, Indicator I, and Indicator J.

[0128] This step can select the five indicators with the highest scores in the remaining indicator library as candidate indicators. For example, indicators F, G, H, I, and J can be selected as candidate indicators (the scores of each indicator have been calculated in the previous steps).

[0129] Step 343. Based on the monitoring data of the retained indicators and each candidate indicator, several comprehensive objective function values ​​are calculated. The candidate indicator corresponding to the largest comprehensive objective function value is combined with the retained indicators to obtain new initial monitoring indicators.

[0130] This step requires calculating the first comprehensive objective function value based on indicators C, E, and F; the second based on indicators C, E, and G; the third based on indicators C, E, and H; the fourth based on indicators C, E, and I; and the fifth based on indicators C, E, and J. Assuming the third comprehensive objective function value is the largest, then indicator H (corresponding to the third comprehensive objective function value) from the candidate indicators is combined with the retained indicators (i.e., indicators C and E) to obtain new initial monitoring indicators: indicators C, E, and H.

[0131] After determining the new initial monitoring indicators, continue to repeat steps 31 to 34 until the required final monitoring indicators are obtained.

[0132] 106. Obtain daily indicator monitoring data based on the final monitoring indicators, obtain indicator trend vector sequences based on the indicator monitoring data, determine the daily predicted health status based on the indicator trend vector sequences, and generate daily explanatory prompts based on the predicted health status.

[0133] This step assumes that the final monitoring indicators—indicators C, E, and H—were determined on October 19th. This step only requires directly retrieving the monitoring data for indicators C, E, and H from step 102.

[0134] In this embodiment, obtaining the indicator trend vector sequence based on indicator monitoring data specifically includes:

[0135] Step 41. Obtain the monitoring data for a specific indicator from the final monitoring indicators for the previous 3 days. Assume today is October 19th, and further assume that the monitoring data for indicator C is for October 16th, October 17th, and October 18th.

[0136] Step 42. Calculate the velocity, acceleration, and volatility values ​​based on the indicator monitoring data from the previous 3 days. This step calculates the velocity, acceleration, and volatility values ​​based on the indicator monitoring data of indicator C on October 16th, 17th, and 18th.

[0137] The speed value of indicator C can be obtained by subtracting the indicator monitoring data value of indicator C on October 16 from the indicator monitoring data value of indicator C on October 18 (if there are multiple indicator monitoring data values ​​for each day, they are processed (e.g., averaged) to obtain a final indicator monitoring data value).

[0138] Subtracting the monitoring data value of indicator C on October 17 from the monitoring data value of indicator C on October 18 yields the first data difference. Subtracting the monitoring data value of indicator C on October 16 from the monitoring data value of indicator C on October 17 yields the second data difference. Subtracting the second data difference from the first data difference yields the acceleration value of indicator C.

[0139] Subtract the velocity value from the first data difference and then square it to get the first squared value. Subtract the velocity value from the second data difference and then square it to get the second squared value. Take the square root of the first squared value and the second squared value to get the volatility of indicator C.

[0140] Step 43. Obtain the indicator monitoring data for the previous 90 days for the corresponding indicator, and obtain the historical standard deviation based on the indicator monitoring data for the previous 90 days.

[0141] This step assumes that indicator C's monitoring data for October 18th and the preceding 89 days is obtained. First, the average of the 90 monitoring data points for indicator C is calculated. Then, the squared deviation of each monitoring data point from the average is calculated. Next, the total average of all squared deviations is calculated. Finally, the square root of the total average is taken to obtain the historical standard deviation of indicator C.

[0142] Step 44. Obtain normalized velocity, normalized acceleration, and normalized volatility based on velocity, acceleration, volatility, and historical standard deviation. Use these values ​​as a trend vector for the corresponding indicators.

[0143] Dividing the velocity value by the historical standard deviation yields the normalized velocity value of indicator C; dividing the acceleration value by the historical standard deviation yields the normalized acceleration value of indicator C; and dividing the volatility by the historical standard deviation yields the normalized volatility of indicator C. This step uses the normalized velocity value, normalized acceleration value, and normalized volatility of indicator C obtained on October 19th as a trend vector for indicator C on October 19th.

[0144] Repeat steps 41 to 44 to obtain multiple indicator trend vectors for the corresponding indicator, and use the multiple indicator trend vectors as the indicator trend vector sequence for the corresponding indicator.

[0145] Still focusing on indicator C, if today is October 20th, then we can obtain indicator monitoring data for indicator C on October 17th, October 18th, and October 19th, as well as indicator monitoring data for indicator C on October 19th and the previous 89 days. Through this data, we can calculate a trend vector for indicator C on October 20th.

[0146] By repeatedly executing steps 41 to 44, a trend vector for indicator C on October 21st, a trend vector for indicator C on October 22nd, and so on, can be calculated. Combining these indicator trend vectors yields the indicator trend vector sequence for indicator C.

[0147] Similarly, by continuously executing steps 41 to 44, we can also obtain the trend vector sequence of indicator E and the trend vector sequence of indicator H.

[0148] This embodiment pre-sets a health status prediction model. By inputting the trend vector sequences of indicators C, E, and H into the health status prediction model, a status probability can be output. Assuming a status probability greater than or equal to 0.75, it indicates "improved status"; a status probability less than 0.75 but greater than or equal to 0.5 indicates "stable status"; and a status probability less than 0.5 indicates "abnormal status".

[0149] The device can generate daily explanatory prompts based on daily status probabilities and display them via a mobile app. Explanatory prompts could be, for example, "Heart rate decreased by 2.5 bpm / day, the trend is improving, it is recommended to maintain this," or "Respiratory rate increased by 2 breaths / day, the trend is abnormal, it is recommended to seek medical attention," or "Indicators stabilized at new levels, no significant changes in physical sensation, continue to observe."

[0150] Figure 2 A block diagram of a time-series-based government data processing device 200 according to an embodiment of the present invention is shown. The device 200 includes:

[0151] The indicator monitoring data acquisition module 202 is configured to collect the user's BCG signal based on a non-contact sensor, and acquire indicator monitoring data of multiple indicators based on the BCG signal.

[0152] The final monitoring indicator determination module 204 is configured to accept user feedback text representing abnormal physical conditions as input by the user; it matches the user feedback text with disease words in the thesaurus; if a disease word is matched, the indicator corresponding to the disease word is used as the final monitoring indicator; if no disease word is matched, the initial monitoring indicators are selected based on the user feedback text, and the final monitoring indicator is determined based on the initial monitoring indicators.

[0153] The explanatory prompt generation module 206 is configured to obtain daily indicator monitoring data based on the final monitoring indicators, obtain indicator trend vector sequences based on the indicator monitoring data, determine the daily predicted health status based on the indicator trend vector sequences, and generate daily explanatory prompts based on the predicted health status.

[0154] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A user feedback-driven adaptive health assessment method based on physiological indicators, characterized in that, include: The user's BCG signal is collected based on a non-contact sensor, and the monitoring data of multiple indicators are obtained based on the BCG signal. The system inputs user feedback text representing abnormal physical conditions. The system matches the user feedback text with disease terms in the lexicon. If a disease term is matched, the indicator corresponding to the disease term is used as the final monitoring indicator. If no disease term is matched, the system filters out initial monitoring indicators based on the user feedback text and determines the final monitoring indicator based on the initial monitoring indicators. Daily indicator monitoring data is obtained based on the final monitoring indicators. Indicator trend vector sequences are obtained based on the indicator monitoring data. Daily predicted health status is determined based on the indicator trend vector sequences. Daily explanatory prompts are generated based on the predicted health status.

2. The method according to claim 1, characterized in that, Matching user feedback text with disease-related terms in the lexicon specifically includes: Step 11. Obtain the semantic vector of the user feedback text based on the BioBERT model; Step 12. Obtain the semantic vector of each symptom word; Step 13. Calculate the semantic similarity value between the semantic vector one and each of the semantic vectors two; Step 14. Determine the maximum semantic similarity value from all the semantic similarity values; Step 15. Determine whether the maximum semantic similarity value is greater than or equal to the first threshold. If the maximum semantic similarity value is greater than or equal to the first threshold, then the user feedback text matches the symptom word; otherwise, the user feedback text does not match the symptom word.

3. The method according to claim 2, characterized in that, The initial monitoring metrics selected based on user feedback text specifically include: Step 21. Obtain an indicator from the indicator library, obtain the semantic vector three of the indicator, and calculate the semantic similarity value two between the semantic vector one and the semantic vector three; Step 22. Obtain the indicator monitoring data for the most recent N days, and calculate the current trend slope, current fluctuation range, and current number of jump points of the indicator based on the indicator monitoring data; Step 23. Obtain the historical indicator monitoring data of the indicator over the past M days, group the historical indicator monitoring data of the indicator into N days, and calculate multiple historical trend slopes, multiple historical fluctuation amplitudes and multiple historical jump points of the indicator based on multiple groups of historical indicator monitoring data; Step 24. Calculate the comprehensive normalized volatility intensity value of the indicator based on the current trend slope, current fluctuation amplitude, current number of jump points, and multiple historical trend slopes, multiple historical fluctuation amplitudes, and multiple historical number of jump points. Step 25. Calculate the index score of the index based on the semantic similarity value and the comprehensive normalized fluctuation intensity value of the index; Repeat steps 21 to 25 until an indicator score has been calculated for each indicator in the indicator library; Step 26. Select several indicators based on their scores to serve as initial monitoring indicators.

4. The method according to claim 3, characterized in that, Step 24 specifically includes: Step 241. Calculate the mean slope and standard deviation of the slope based on multiple historical trend slopes, calculate the mean amplitude and standard deviation of the amplitude based on multiple historical fluctuation amplitudes, and calculate the mean number of points and standard deviation of the number of points based on multiple historical jump points; Step 242. Calculate the slope deviation based on the current trend slope and the average slope, calculate the amplitude deviation based on the current fluctuation amplitude and the average amplitude, and calculate the point deviation based on the current number of jump points and the average number of points; Step 243. Calculate the normalized slope deviation based on the slope deviation and the slope standard deviation, calculate the normalized amplitude deviation based on the amplitude deviation and the amplitude standard deviation, and calculate the normalized point deviation based on the point deviation and the point standard deviation. Step 244. Calculate the comprehensive normalized fluctuation intensity value of the index based on the normalized slope deviation, normalized amplitude deviation, and normalized point number deviation.

5. The method according to claim 3, characterized in that, Step 25 specifically calculates the indicator score using the following formula: ， in, The score for a certain indicator. The preset weight value for semantic similarity value two. The semantic similarity value of a certain indicator is two. The preset weight values ​​are used to comprehensively normalize the fluctuation intensity values. This represents the comprehensive normalized volatility intensity value of a certain indicator.

6. The method according to claim 1, characterized in that, Determining the final monitoring indicators based on the initial monitoring indicators specifically includes: Step 31. Obtain monitoring data for subsequent indicators based on the initial monitoring indicators; Step 32. Determine whether the user has not entered a daily symptom severity score for Q consecutive days. If so, use the initial monitoring indicator as the final monitoring indicator; otherwise, proceed to step 33. Step 33. Obtain the comprehensive objective function value based on subsequent indicator monitoring data and the daily symptom severity score input by the user. If the comprehensive objective function value cannot be obtained, return to step 31; otherwise, proceed to step 34. Step 34. Determine whether the comprehensive objective function value is greater than or equal to the second threshold. If the comprehensive objective function value is greater than or equal to the second threshold, then the initial monitoring indicator is used as the final monitoring indicator. If the comprehensive objective function value is less than the third threshold or less than the historical maximum comprehensive objective function value, then one of the initial monitoring indicators is replaced to obtain a new initial monitoring indicator, and the process returns to step 31. Otherwise, the process returns directly to step 31.

7. The method according to claim 6, characterized in that, Step 33 specifically includes: Step 331. Obtain the indicator change sequence based on the subsequent indicator monitoring data for at least Y consecutive days, form a symptom score sequence based on the daily symptom severity score values ​​for at least Y consecutive days, obtain the Spearman rank correlation coefficient based on the indicator change sequence and the symptom score sequence, and use the Spearman rank correlation coefficient as the diurnal fluctuation consistency value. Step 332. Obtain the predicted score based on subsequent indicator monitoring data and regression model, determine the actual score based on the obtained daily symptom severity score, obtain the total sum of squares and residual sum of squares based on the predicted score and the actual score, obtain the coefficient of determination based on the total sum of squares and residual sum of squares, and use the coefficient of determination as the physiological explanatory value. Step 333. Obtain the correlation coefficient between indicators based on subsequent indicator monitoring data, and obtain the indicator diversity value based on the correlation coefficient between indicators; Step 334. Calculate the comprehensive objective function value based on the consistency value of diurnal fluctuations, the physiological explanatory power value, and the indicator diversity value.

8. The method according to claim 6, characterized in that, The replacement of one of the initial monitoring indicators in step 34 to obtain a new initial monitoring indicator specifically includes: Step 341. Remove the indicator with the lowest contribution to the comprehensive objective function value from the initial monitoring indicators to obtain the remaining indicators; Step 342. Remove the indicators that are the same as the initial monitoring indicators from the indicator library or the remaining indicator library to obtain the remaining indicator library, and select the indicators with the highest indicator scores in the remaining indicator library as candidate indicators. Step 343. Based on the monitoring data of the retained indicators and each candidate indicator, several comprehensive objective function values ​​are calculated. The candidate indicator corresponding to the largest comprehensive objective function value is combined with the retained indicators to obtain new initial monitoring indicators.

9. The method according to claim 1, characterized in that, Obtaining the indicator trend vector sequence based on the aforementioned indicator monitoring data specifically includes: Step 41. Obtain the monitoring data of a specific indicator from the final monitoring indicators for the previous 3 days; Step 42. Calculate the velocity value, acceleration value, and volatility based on the indicator monitoring data from the previous 3 days; Step 43. Obtain the indicator monitoring data for the previous 90 days for the corresponding indicator, and obtain the historical standard deviation based on the indicator monitoring data for the previous 90 days; Step 44. Obtain normalized velocity value, normalized acceleration value, and normalized volatility based on velocity value, acceleration value, volatility, and historical standard deviation, and use the normalized velocity value, normalized acceleration value, and normalized volatility as a trend vector of the corresponding indicator; Repeat steps 41 to 44 to obtain multiple indicator trend vectors for the corresponding indicator, and use the multiple indicator trend vectors as the indicator trend vector sequence for the corresponding indicator.

10. A user feedback-driven adaptive health assessment method and apparatus for physiological indicators, characterized in that, include: The indicator monitoring data acquisition module is configured to collect the user's BCG signal based on a non-contact sensor, and acquire indicator monitoring data for multiple indicators based on the BCG signal. The final monitoring indicator determination module is configured to accept user feedback text that represents abnormal physical conditions. The system matches user feedback text with disease terms in the thesaurus. If a disease term is matched, the indicator corresponding to the disease term is used as the final monitoring indicator. If no disease term is matched, the system filters out initial monitoring indicators based on user feedback text and determines the final monitoring indicator based on the initial monitoring indicators. The explanatory prompt generation module is configured to obtain daily indicator monitoring data based on the final monitoring indicators, obtain indicator trend vector sequences based on the indicator monitoring data, determine the daily predicted health status based on the indicator trend vector sequences, and generate daily explanatory prompts based on the predicted health status.

Images