Sleep temperature regulation and control method and system based on non-invasive brain-computer interface

Abstract

The invention provides a sleep temperature regulation and control method and system based on a non-intrusive brain-computer interface, and belongs to the technical field of data processing, and the method specifically comprises the steps: determining available marketing selling points in agricultural products based on a matching condition, and determining the sleep temperature of the agricultural products based on composition data of the available marketing selling points in the marketing selling points; the method comprises the following steps: obtaining an agricultural product, determining a marketing association type of a marketing account in the agricultural product in combination with matching conditions of available marketing selling points and different user portraits, and determining a data processing method of the marketing account for updating processing of the user portraits according to the marketing association type of the marketing account in different agricultural products. Based on the updating results of the user portraits of the different marketing accounts, the generation processing method of the marketing selling points of the agricultural products is determined, and the identification processing efficiency of the marketing selling points is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of temperature regulation, and particularly relates to a sleep temperature regulation method and system based on a non-invasive brain-computer interface. BACKGROUND

[0002] Brain-computer interface data can be used to obtain the brain electrical signals of a user, and the brain electrical signals can be used for sleep staging processing. Specifically, in the application patent application CN202511147316.8 “Intelligent sleep detection and intervention eyeshade system and method combined with a brain-computer interface”, an accurate sleep detection result is obtained by real-time analysis based on brain-computer interface technology through a core control module, so as to provide personalized sleep detection for a user, and real-time sleep intervention is realized according to the personalized sleep detection result, so as to improve the sleep intervention effect for the user. However, the following technical problems exist: In order to adopt different temperature regulation rates in different sleep stages, so as to avoid the occurrence of the technical problem of poor user experience caused by too fast temperature regulation rate in the deep sleep stage, the recognition accuracy of the sleep staging model is crucial. Therefore, how to determine the update strategy of the sleep staging model according to the adjustment delay of the user in the environmental temperature adjustment process and the deviation of the sleep staging of the user with a faster adjustment rate, so as to improve the reliability of the temperature regulation process of the user, becomes a technical problem to be solved.

[0003] Therefore, there is an urgent need for a sleep temperature regulation method and system based on a non-invasive brain-computer interface. SUMMARY

[0004] To achieve the purpose of the application, the application adopts the following technical solutions: Specifically, the application provides a sleep temperature regulation method based on a non-invasive brain-computer interface, which specifically comprises the following steps: S1 determining an updated brain-computer staging model based on the update data of the brain-computer staging model based on brain-computer interface data, determining a suspected deviation user in the sleep staging result of the user, determining the constituent data of the suspected deviation user, and determining that the adjustment processing of the environmental temperature can be performed using the sleep staging result, and entering the next step based on the feature similarity of the brain electrical signals of different suspected deviation users; S2 determining the response delay data of the temperature control device in different environmental temperature intervals according to the temperature monitoring data of the suspected deviation user, determining a temperature regulation requirement user in the suspected deviation user based on the response delay data, and determining the update management strategy of the brain-computer staging model according to the adjustment data of the temperature regulation requirement user in different environmental temperature intervals and the coincidence of the temperature regulation requirement user and the suspected deviation user.

[0005] The present application has the advantages of: Based on the response delay data, the temperature adjustment demand user in the suspected deviation user is determined, so as to realize the screening of the suspected deviation user with faster adjustment rate in different environmental temperature intervals, that is, the screening of the temperature adjustment demand user, and lay a foundation for further determining the update management strategy of the brain-computer staging model according to the composition of the temperature adjustment user in the suspected deviation user.

[0006] According to the adjustment data of the temperature adjustment demand user in different environmental temperature intervals and the coincidence of the adjustment demand user and the suspected deviation user, the update management strategy of the brain-computer staging model is determined, which not only considers the difference in the update processing demand of the brain-computer staging model due to the difference in the adjustment delay, but also considers the difference in the update processing demand of the brain-computer staging model caused by the coincidence of the adjustment demand user and the suspected deviation user, so as to realize the determination of the update management strategy of the brain-computer staging model from the perspective of the update processing demand, and improve the efficiency and pertinence of the update processing of the brain-computer staging model.

[0007] Further, the update data of the brain-computer staging model is determined according to the update record of the brain-computer staging model.

[0008] Further, the update data of the brain-computer staging model includes the update period of the brain-computer staging model, that is, in which period the brain-computer staging model is updated.

[0009] Further, the variation of the sleep staging result of the user is determined according to the variation of the sleep staging result under similar brain-computer signal characteristics after the brain-computer staging model is updated in different update periods.

[0010] It should be noted that the brain-computer signal characteristics include frequency characteristics and amplitude characteristics, and the variation is determined according to whether the brain-computer staging result under similar brain-computer signal characteristics in a unit time is consistent.

[0011] Further, the method for determining the suspected deviation user in the user is: Based on the brain-computer signal data of the user in different sleep processes, the time periods with similar brain-computer signal characteristics in a unit time are divided into the same brain-computer characteristic combination; Based on the updated brain-computer staging model, the different historical versions of the brain-computer staging model are determined, and the variation of the sleep staging result of the user in different time periods under the brain-computer characteristic combination is determined to determine the sleep staging result of the different historical versions. determine whether the user is a suspected deviation user based on sleep staging results of different historical versions of sleep staging results in different brain-computer feature combinations.

[0012] Further, the method for determining the update management strategy of the brain-computer staging model is: determine the number of adjustments of the temperature adjustment demand user in different environmental temperature intervals based on the adjustment data of the temperature adjustment demand user in different environmental temperature intervals; determine the proportion of the number of adjustment demand users in the suspected deviation users based on the coincidence of the adjustment demand users and the suspected deviation users, and take it as the demand user number proportion; determine the update management strategy of the brain-computer staging model according to the demand user number proportion, the number of adjustments of the temperature adjustment demand user in different environmental temperature intervals, and the delay coefficient.

[0013] In a second aspect, the present application provides a computer system, comprising a memory and a processor connected in communication, and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to perform the above-mentioned sleep temperature regulation method based on the non-invasive brain-computer interface.

[0014] Other features and advantages will be set forth in the following description of the application, and in part will be apparent from the description and the drawings, or can be learned by practice of the application as claimed in the claims.

[0015] In order to make the above-mentioned purpose, features and advantages of the present application more obvious and easy to understand, the following preferred embodiments are specifically described, and the accompanying drawings are described in detail as follows. BRIEF DESCRIPTION OF DRAWINGS

[0016] The above and other features and advantages of the present application will become more apparent by describing in detail example embodiments thereof with reference to the attached drawings.

[0017] Figure 1 is a flowchart of a sleep temperature regulation method based on a non-invasive brain-computer interface; Figure 2 is a flowchart of a method for determining suspected deviation users among users; Figure 3 is a flowchart of a method for determining temperature adjustment demand users among suspected deviation users. DETAILED DESCRIPTION

[0018] In order for those skilled in the art to better understand the technical solutions in the specification, the technical solutions in the specification will be clearly and completely described below in combination with the drawings in the specification. Obviously, the described embodiments are only part of the embodiments of the specification, not all. Based on the embodiments of the specification, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the specification.

[0019] Embodiment 1 As Figure 1 shown, the application provides a sleep temperature regulation method based on a non-invasive brain-computer interface, specifically comprising: S1 determines the brain-computer staging model after updating based on the updating data of the brain-computer interface data brain-computer staging model, determines the suspected deviation user in the user under the change of the sleep staging result of the user, determines the sleep staging result that can be used for environmental temperature adjustment processing based on the constituent data of the suspected deviation user, and combines the feature similarity of the suspected deviation user's electroencephalogram signal, and enters the next step when the sleep staging result can be used for environmental temperature adjustment processing; S2 determines the response delay data of the temperature control device in different environmental temperature intervals according to the temperature monitoring data of the suspected deviation user, determines the temperature adjustment demand user in the suspected deviation user based on the response delay data, and determines the updating management strategy of the brain-computer staging model according to the adjustment data of the temperature adjustment demand user in different environmental temperature intervals and the coincidence of the adjustment demand user and the suspected deviation user.

[0020] It should be noted that when the brain-computer signal features of the user are acquired, the authorization and consent of the user are required.

[0021] Further, the updating data of the brain-computer staging model is determined according to the updating record of the brain-computer staging model.

[0022] Further, the updating data of the brain-computer staging model includes the updating period of the brain-computer staging model, that is, in which period the brain-computer staging model is updated.

[0023] Further, the change of the sleep staging result of the user is determined according to the change of the sleep staging result under similar brain-computer signal features after the brain-computer staging model is updated in different updating periods.

[0024] It should be noted that the brain-computer signal features include frequency features and amplitude features, and specifically, whether the brain-computer staging result under similar brain-computer signal features in a unit time is consistent is determined to determine the change.

[0025] Note that the brain-machine signal feature: extracted from electroencephalogram and other signals, used to represent the index of sleep stage, frequency feature: such as the power of Delta wave, Theta wave in different frequency bands, amplitude feature: such as the amplitude, peak value of brain wave, etc.

[0026] Brain-machine staging model: an AI model, the input is the brain-machine signal of the user, and the output is the sleep staging result (such as wakefulness, REM, N1, N2, N3 stage).

[0027] Historical version: different iteration versions of the model saved after multiple data updates in history (such as V1.0, V1.1, V1.2...).

[0028] Core problem: whether the same user, in similar physiological state (i.e. similar brain-machine signal features), different versions of the model give consistent sleep staging results? If the results are inconsistent, it means that the user's data or physiological mode may cause the model to be unstable in judgment, and the user is marked as a "suspected deviation user".

[0029] Specifically, as shown in Figure 2 The method for determining the suspected deviation user in the user is: S11 based on the brain-machine signal data of the user in different sleep processes, the time periods with similar brain-machine signal features in a unit time are divided into the same brain-machine feature combination; In the above step, a cross-cycle "physiological state" profile - brain-machine feature combination is constructed. Sleep is periodic, and the same period (such as when falling asleep, deep sleep period) of different nights may be in similar physiological state. The goal of this step is to break the boundaries of single sleep record, and classify the time periods with similar physiological state in all sleep records of the user together to form a unified "state profile".

[0030] Unit time: a time window for analysis, for example, a data epoch of 30 seconds, brain-machine feature combination: a set containing all time periods with highly similar brain-machine signal features of the user in different nights and different dates.

[0031] Specific method: Iterate through the sleep process data of each user, divide the data into multiple time periods of "unit time", calculate the similarity coefficient (such as cosine similarity) of brain-machine signal features between any two time periods, if the similarity coefficient is greater than the preset coefficient threshold (such as 0.9), it is considered that the two time periods represent the same physiological state, and they are classified into the same "brain-machine feature combination".

[0032] User Xiao Wang has 3 nights of sleep data. The system finds: "Combination A": contains 00:15-00:16 of the 1st night, 00:20-00:21 of the 2nd night, and 00:18-00:19 of the 3rd night. These time periods are characterized by extremely high delta wave power, smooth signals, "Combination B": contains 02:05-02:06 of the 1st night and 02:10-02:11 of the 3rd night. These time periods are characterized by prominent theta and sawtooth waves.

[0033] S12 determines the different historical versions of the brain-machine staging model based on the updated brain-machine staging model, the changes in the sleep staging results of the user in different time periods under the brain-machine feature combination, and determines the sleep staging results of the different historical versions; Specifically, the stability of the evaluation model under different "states" is evaluated, which is the core of detecting model consistency. All saved model historical versions (such as V1.0, V1.1, V1.2, V1.3) are "re-run" on each "physiological state profile" (brain-machine feature combination) constructed in step S11 to determine whether their judgments for the same state are consistent.

[0034] Sleep staging results of historical versions: refer to a series of results obtained by using a specific historical model (such as V1.1) to stage each time period under a certain "brain-machine feature combination".

[0035] Embodiment (continued): For "Combination A" (high delta wave period): Model V1.0's judgment result: [N3, N3, N3] -> consistent as "deep sleep period", Model V1.1's judgment result: [N3, N3, N3] -> consistent as "deep sleep period", Model V1.2's judgment result: [N3, N2, N3] -> inconsistent, Model V1.3's judgment result: [N3, N3, N3] -> consistent as "deep sleep period" S13 determines whether the user is a suspected deviation user based on the sleep staging results of the different historical versions of the sleep staging results in different brain-machine feature combinations.

[0036] Specifically, the suspected deviation user is determined by consistency voting, which requires a quantitative index to judge the degree of instability. By counting the distribution of the models' "votes" under the same state, it can accurately measure which state of the user will cause the models to disagree.

[0037] Version combination: the set of historical version models that output the same staging result under a certain "brain-machine feature combination", identify matching factor: for a "version combination", the number of versions / total number of historical versions. This factor represents "what proportion of models hold the same opinion under this physiological state".

[0038] Specific method: Grouping: under the same "brain-machine feature combination", historical versions with the same output result are grouped into the same "version combination", and influence is calculated: calculate the identification matching factor of each "version combination".

[0039] Judgment: if there is a "brain-machine feature combination" in which the identification matching factor of all "version combinations" is less than a certain threshold (for example, 0.6), the user is determined to be a suspected deviation user.

[0040] Analysis of "combination B" (period of prominent theta waves): version 1: contains V1.0, staging result is REM, version 2: contains V1.1, staging result is N1, version 3: contains V1.2, staging result is REM, version 4: contains V1.3, staging result is N1; Calculation: the identification matching factor of each version combination is 1 / 2 = 0.5, i.e. V1.0 and V1.2 are in the same version combination.

[0041] Conclusion: under the physiological state of "combination B", the four models are divided into two factions, and neither faction has a "vote rate" exceeding 0.6. The judgment of the model is seriously divergent and unstable.

[0042] Final ruling: because there is a "combination B" brain-machine feature combination in which the identification matching factor of all version combinations is less than the preset threshold (0.6), the system determines user Xiao Wang to be a "suspected deviation user".

[0043] It can be understood that the specific division method of the brain-machine feature combination is: If the similarity coefficient of the brain-machine signal feature in a certain unit time period in the current sleep process is greater than the preset coefficient threshold compared with that in a certain unit time period in other sleep processes, then the certain unit time period in the current sleep process and the certain unit time period in other sleep processes are divided into the same brain-machine feature combination.

[0044] Specifically, the brain-machine staging model updated in history is regarded as a historical version.

[0045] It should be noted that the historical version of the sleep staging result is determined according to the sleep staging result of the historical version of the brain-machine staging model under the brain-machine feature combination after the update is completed.

[0046] It can be understood that, based on the sleep staging result of the different historical versions of the sleep staging result in different brain-machine feature combinations, whether the user is a suspected deviation user is determined, specifically including: Based on the sleep staging result under the brain-machine feature combination, the historical version with consistent sleep staging result is determined, and the historical version with consistent sleep staging result is divided into the same version combination. According to the number of historical versions in different version combinations under different brain-machine feature combinations, whether the user is a suspected deviation user is determined.

[0047] It can be understood that, according to the number of historical versions in different version combinations under different brain-machine feature combinations, whether the user is a suspected deviation user is determined, specifically including: Based on the proportion of the number of historical versions in the version combination in all historical versions under the brain-machine feature combination, the identification matching factor of the version combination is determined. When there are different version combination identification matching factors of brain-machine feature combinations that are less than the preset matching factor threshold, it means that the sleep staging results of different historical versions in the brain-machine feature combination have a high degree of change, and therefore the user is determined to be a suspected deviation user.

[0048] Focus on consistency rather than absolute correctness: It does not care which model is "correct", but only cares whether the models are "consistent". Inconsistency means high uncertainty. Compare based on physiological state: Ensure the fairness of comparison, which is "compare apples to apples". Accurately locate the problem: Not only can it find that the user has a problem, but also can locate which specific physiological state (such as REM / N1 transition period) of the user is most likely to cause model disagreement. This has a very high guiding value for subsequent model repair and optimization.

[0049] Specifically, it is determined that the sleep staging result can be used for adjustment processing of the environmental temperature, specifically including: When it is found that the sleep staging result of a part of users (suspected deviation users) is unstable, whether this instability is an exception or a general systematic defect? This directly determines whether the system can still be trusted to automatically control important parameters such as environmental temperature.

[0050] Application scenario setting Background: A smart home company deploys a "sleep optimization system" that monitors users' sleep stages using a brain-computer interface and automatically adjusts the bedroom air conditioning temperature accordingly (e.g., reducing the temperature adjustment rate during deep sleep to save energy and improve sleep quality, and maintaining a higher temperature adjustment rate during REM sleep to improve sleep quality).

[0051] Current problem: Through the previous detection process, the system found that a part of the total users were marked as "suspected deviation users", meaning that the model's sleep staging results for them were inconsistent and unreliable. The company's management needs to decide whether to immediately suspend the temperature adjustment function of the entire system or continue to trust the system.

[0052] S21 determines the proportion of suspected deviation users in the total number of users based on the composition data of the suspected deviation users, and takes it as the suspected deviation proportion; Step S21: Evaluate the impact of the problem - Calculate the suspected deviation proportion. The system found that 50 users were determined to be "suspected deviation users". Calculate: suspected deviation proportion = 50 / 1000 = 5%.

[0053] S22 determines the brain-computer feature combination of different version combinations of suspected deviation users that is less than the preset matching factor threshold, and takes it as the staging deviation combination. According to the similarity of different staging deviation combinations of suspected deviation users, determine the similarity coefficient of electroencephalogram feature data of different staging deviation combinations of suspected deviation users; S23 determines whether the sleep staging results can be used for environmental temperature adjustment processing according to the suspected deviation proportion and the similarity coefficient of electroencephalogram feature data of different staging deviation combinations of suspected deviation users.

[0054] Specifically, the similarity coefficient of electroencephalogram feature data of the staging deviation combination is determined according to the maximum value of the similarity coefficient between the time periods in the staging deviation combination, wherein the similarity coefficient is determined according to the calculation result of the Euclidean distance function between brain-computer signal features.

[0055] It can be understood that determining whether the sleep staging results can be used for environmental temperature adjustment processing according to the suspected deviation proportion and the similarity coefficient of electroencephalogram feature data of different staging deviation combinations of suspected deviation users specifically includes: S231 judges whether the suspected deviation proportion is greater than the preset deviation proportion threshold. If yes, it is determined that the sleep staging results cannot be used for environmental temperature adjustment processing. If not, go to the next step; Judgment (S231): The suspected deviation proportion (5%) is not greater than the preset deviation proportion threshold (set to 10%), which is macro risk assessment. If the proportion of problem users is high, it means that this is likely to be a systematic model defect, and all automated operations must be stopped immediately. On the contrary, if it is only a small range of problems, further in-depth analysis can be carried out.

[0056] Suspected deviation proportion: reflects the universality of model instability. Decision: Since the proportion is not high (5% <10%), the problem is not universal, and the system is not stopped immediately, and the next step of in-depth analysis is entered.

[0057] S232 determines whether there is a staging deviation combination between the suspected deviation user and other suspected deviation users based on the similarity coefficient of the electroencephalogram feature data of the different suspected deviation user staging deviation combinations, if yes, it is determined that the sleep staging result cannot be used for environmental temperature adjustment processing, if no, it enters the next step; In the above steps, the problem homogeneity is evaluated - check if the deviation sources are consistent, the system extracts the "staging deviation combination" (i.e. the electroencephalogram feature that causes the model to diverge) of each of the 50 suspected deviation users, and calculates the similarity coefficient of the electroencephalogram feature data of any two users "staging deviation combination", check if there is at least one pair of "staging deviation combination" between all suspected deviation users, the similarity coefficient is less than the preset similarity coefficient threshold (set to 0.8), Example and judgment: After analysis, among the 50 users, the "staging deviation combination" of 45 users is very similar (similarity coefficient > 0.9), and they all show model confusion in the "REM sleep and N1 period transition" electroencephalogram mode, but there are 5 users whose deviation combination is different from the other 45 users, and the similarity coefficient is less than 0.8. For example, user U46's deviation occurs when "deep sleep period wakes up".

[0058] Judgment: Since there is such a user U46, he does not have a highly similar deviation combination with most other suspected deviation users. Therefore, the condition that "there is a highly similar deviation combination between all suspected deviation users" is not met.

[0059] This is a root heterogeneity analysis. If all problem users have problems in the same electroencephalogram mode, the root is single, and it may be a specific defect of the model. But if the problems are diverse, it means that the model may have a deeper problem of insufficient generalization, which is more risky.

[0060] The similarity coefficient of the EEG feature data of the different staging deviation combinations: an index for measuring whether the physiological patterns leading to the failure of the model of different users are similar. High similarity means strong repairability. Decision: The condition is not met, and the system is not immediately stopped. Go to the next step.

[0061] S233 determines the comprehensive similarity coefficient between the suspected deviation user and other suspected deviation users based on the average of the similarity coefficients of the EEG feature data of different staging deviation combinations between the suspected deviation user and other suspected deviation users, and determines whether the comprehensive similarity coefficients between different suspected deviation users are all less than a preset comprehensive similarity coefficient threshold. If yes, it is determined that the sleep staging result cannot be used for environmental temperature adjustment processing. If no, go to the next step. In the above step, the overall similarity is evaluated - the comprehensive similarity coefficient is calculated. The system calculates the average of the similarity coefficients of all "staging deviation combinations" between each two suspected deviation users as their "comprehensive similarity coefficient".

[0062] Judgment: Is the comprehensive similarity coefficient between all suspected deviation users less than the preset comprehensive similarity coefficient threshold (assuming 0.7)? The comprehensive similarity coefficient between the 45 users with the same problem is very high (~ 0.95), but the comprehensive similarity coefficient between user U46 and any of the 45 users is very low (~ 0.3).

[0063] Judgment: Due to the existence of U46, the condition that "the similarity coefficient between all users is greater than 0.7" is not met, which is a quantitative enhancement of S232. S232 looks at whether there are high-similarity pairs, and S233 looks at whether the overall relationship network is tight. It further confirms that the problem user group is not a monolithic block, but there are "outliers", and the decision is: The condition is not met, and the system is not immediately stopped. Go to the final decision step.

[0064] S234 obtains the suspected deviation proportion and determines the recognition deviation probability of the brain-computer staging model in combination with the comprehensive similarity coefficients between different suspected deviation users, and determines whether the sleep staging result can be used for environmental temperature adjustment processing based on the recognition deviation probability.

[0065] It can be understood that when the recognition deviation probability is greater than the preset deviation probability threshold, it is determined that the sleep staging result cannot be used for environmental temperature adjustment processing.

[0066] In summary of the above steps, the final comprehensive risk probability assessment is performed, and a final "identification bias probability" needs to be calculated, where the identification bias probability = the suspected bias proportion × (1 - the average comprehensive similarity coefficient), the average comprehensive similarity coefficient: calculate the average value of the comprehensive similarity coefficient between all 50 users, which is 0.75, the identification bias probability = 5% × (1 - 0.75) = 5% × 0.25 = 1.25% Judgment: the identification bias probability (1.25%) > the preset bias probability threshold (assuming 1%), which is a comprehensive trade-off of cost / risk / benefit. It combines "problem scope" (proportion) and "problem severity and repair difficulty" (1 - average similarity) into a final indicator. Low probability means: 1) few affected users; 2) the problem is relatively concentrated and easy to locate and repair. Therefore, the benefits brought by the system automatic control are still greater than its risks.

[0067] Identification bias probability: a quantitative indicator of the overall risk of a comprehensive evaluation model producing unreliable staging results in the current user group. Final decision: since the identification bias probability (1.25%) exceeds the safety threshold (1%), the system finally determines that: sleep staging results cannot be used to adjust the ambient temperature.

[0068] It should be noted that when sleep staging results cannot be used to adjust the ambient temperature, only the sleep staging model needs to be updated according to the preset time period.

[0069] Specifically, the ambient temperature interval is divided into equal intervals based on 5 degrees Celsius.

[0070] Further, the response delay data is determined according to the time taken by the temperature control device to reach the target temperature when adjusting the temperature, for example, the time taken to adjust 1 degree Celsius.

[0071] Specifically, as shown in Figure 3 The method for determining the temperature adjustment requirement user in the suspected bias user is: Objective: From the "suspected bias user", select those users who most need and are most suitable for receiving fine temperature adjustment services based on sleep staging.

[0072] Core reverse thinking: a good performance temperature control system: can quickly respond to instructions, so it can "calmly" perform fine temperature adjustment, small amplitude, and pre-judgment. For example, the system can predict that the user will enter deep sleep, and can slowly reduce 0.5℃ 30 minutes in advance. Because the hardware responds quickly, this strategy can be perfectly implemented, improving the experience and energy efficiency.

[0073] Poorly performing temperature control system: slow response, fine-tuning strategies fail due to hardware latency. For such systems, the most effective strategy is "strong and fast adjustment". Once adjustment is needed, use maximum power to reach the target temperature as quickly as possible, without distinguishing sleep stages. Because the hardware itself is the bottleneck, it is pointless to have a more sophisticated software strategy.

[0074] Therefore, "temperature regulation demand user" specifically refers to those "hardware performance is good enough to deploy fine software strategy".

[0075] S31 determines the average value of the time taken by the suspected deviation user to reach the target temperature in different ambient temperature intervals based on the response delay data of the temperature control device in different ambient temperature intervals, and takes it as the adjustment duration mean value; Quantify hardware performance - calculate adjustment duration mean value, system analyzes historical data of temperature control equipment in user's home, calculates the average value of time required to execute a unit temperature change (such as adjusting 1℃) under different ambient temperature intervals.

[0076] Establish performance baseline: this is the objective basis for evaluating hardware capability, which converts performance from "fast / slow" perceptual cognition to quantifiable data, and identifies working condition differences: the performance of temperature control equipment is greatly affected by ambient temperature (such as air conditioner slow in extremely hot weather). Subinterval statistics can more accurately reflect its true ability in different seasons and different weather.

[0077] Ambient temperature interval: divide external climate or initial indoor temperature into multiple ranges, such as (25-30℃] (high temperature in summer), [10-15℃] (low temperature in winter). Used to distinguish different working conditions of the device, response delay data: refers to the actual time taken by the temperature control device to complete a unit temperature change (such as increasing or decreasing 1℃) from receiving the instruction.

[0078] Adjustment duration mean value: the arithmetic mean of all historical response delay data in a specific ambient temperature interval. It represents the typical performance of the device in that working condition.

[0079] S32 determines the adjustment delay temperature interval in the ambient temperature interval based on the adjustment duration mean value in different ambient temperature intervals; In the above steps, the adjustment duration mean value of each ambient temperature interval is compared with a unified performance standard, and the pass line is set: a clear and quantitative standard is needed to judge performance "good" and "poor", and the problem interval is located: find out the specific working conditions in which the performance of the device does not meet the standard, which is the key basis for subsequent decision-making.

[0080] Pre-set adjustment duration threshold: the maximum time tolerance limit that the system administrator considers as "fast enough" for the temperature control device to respond. For example, set to 10 minutes, means that it is considered "slow" if it takes more than 10 minutes to adjust 1°C. Adjustment delay temperature interval: refers to those environmental temperature intervals whose adjustment duration mean is greater than the pre-set adjustment duration threshold. These are the performance short board intervals of the device.

[0081] S33 determines whether the suspected deviation user is a temperature adjustment demand user according to the adjustment delay temperature interval data.

[0082] It can be understood that the adjustment delay temperature interval is an environmental temperature interval whose adjustment duration mean is greater than the pre-set adjustment duration threshold.

[0083] Case 1: when the number of adjustment delay temperature intervals is less than the pre-set temperature interval number threshold, it means that the temperature control device of the suspected deviation user has a fast adjustment rate in different environmental temperature intervals, so it is necessary to adopt different adjustment rates in different sleep staging intervals, and therefore the suspected deviation user is determined as a temperature adjustment demand user.

[0084] Case 1: the performance is generally good, which is an ideal object for fine adjustment. Judgment: the number of adjustment delay temperature intervals < the pre-set temperature interval number threshold? If the device only performs poorly in a small number of extreme working conditions, but performs well in most normal working conditions, it means that the overall performance of the device is reliable and has the hardware basis to implement fine strategies.

[0085] Pre-set temperature interval number threshold: the critical value for judging whether the performance short board of the device is "too much". For example, there are a total of 10 temperature intervals, and the threshold is set to 3, which means that only when 3 or more intervals have poor performance, the overall performance of the device is considered to be poor.

[0086] Embodiment and decision: Mr. Li's device has 2 "adjustment delay temperature intervals" in 10 intervals. 2 < 3 (threshold), decision: Mr. Li is a temperature adjustment demand user, and his device responds quickly most of the time, which is worthy of deploying fine sleep staging temperature control strategies.

[0087] Case 2: when the number of adjustment delay temperature intervals is not less than the pre-set temperature interval number threshold, based on the adjustment duration mean in different adjustment delay temperature intervals, it is determined whether there is an adjustment delay temperature interval whose adjustment duration mean is greater than the pre-set duration threshold (greater than the pre-set adjustment duration threshold), if yes, it is determined that the suspected deviation user is not a temperature adjustment demand user, if no, it goes to the next step; Case 2: Check if there is a serious performance defect, trigger: when the number of adjusted delay temperature intervals ≥ preset temperature interval number threshold, enter this case, judge: whether there is an interval with an adjustment duration mean value > preset duration threshold? (The preset duration threshold is a more stringent value than the preset adjustment duration threshold) This is a "veto" mechanism. Used to identify devices that have serious faults or performance bottlenecks. The preset duration threshold: a higher delay threshold for determining whether a device is in a serious performance failure state. For example, set to 20 minutes.

[0088] Example and decision: (Assume Mr. Li has 4 poor performance intervals, enter case 2). Check that the [10-15℃] interval requires 25 minutes, which is greater than 20 minutes. Decision: Determine that Mr. Li is not a temperature regulation demand user. Logic: The device has serious defects and fine adjustment is out of the question.

[0089] Case 3: Determine the delay coefficient based on the proportion of the number of adjusted delay temperature intervals within the environmental temperature interval, judge whether the delay coefficient is less than the preset delay coefficient threshold, if yes, determine that the suspected deviation user belongs to the temperature regulation demand user, if not, determine that the suspected deviation user does not belong to the temperature regulation demand user.

[0090] Case 3: Quantify the range of insufficient performance (for the final evaluation of devices with medium performance), when the judgment of case 2 is "no" (i.e. no serious defects, but many intervals with poor performance), enter this case, judge: delay coefficient < preset delay coefficient threshold, make the final trade-off for devices with "medium to low" performance. Even if there are many intervals with poor performance, if the proportion does not exceed a limit, the system can still adopt a "mixed strategy" (use fine strategy in intervals with good performance, and use strong strategy in intervals with poor performance).

[0091] Delay coefficient: Number of adjusted delay temperature intervals / Total number of environmental temperature intervals. It reflects the proportion of the range of insufficient performance of the device.

[0092] Preset delay coefficient threshold: Critical value for judging whether the range of insufficient performance is "too large". For example, 0.3, i.e. 30% of the working conditions have poor performance.

[0093] Example and decision: (Assume Mr. Li has 4 poor performance intervals and no serious defects). Total intervals are 10, delay coefficient = 4 / 10 = 0.4, 0.4 is not less than 0.3, decision: Determine that Mr. Li is not a temperature regulation demand user.

[0094] Logic: The device is slow to respond in more than 40% of the working conditions, and does not have the general conditions for deploying fine strategies.

[0095] Specifically, the method for determining the update management strategy of the brain-machine staging model is: First, define the two update strategies, which is the key to understanding the entire process: preset update strategy (aggressive strategy): as long as the sleep staging results of any suspected deviation user under the same brain-machine feature combination are inconsistent, the model update is triggered immediately.

[0096] Features: high sensitivity, low tolerance. Aimed at quickly fixing any identified minor inconsistencies to ensure the model is in optimal condition. Suitable for high-demand, high-risk scenarios, but may result in frequent updates.

[0097] Second preset update strategy (conservative strategy): only when there are multiple suspected deviation users with inconsistent sleep staging results under the same staging deviation combination, the model update is triggered. Features: emphasis on universality, tolerance of individual phenomena. Aimed at fixing systematic model defects that affect multiple users, avoiding unnecessary updates for individual user's special cases. More stable, less resource consumption.

[0098] Decision goal: intelligently choose to use the "aggressive strategy" or the "conservative strategy" based on the current system status.

[0099] S41 determines the number of adjustments in different environmental temperature intervals based on the adjustment data of the temperature adjustment demand users in different environmental temperature intervals; In the above step, the historical adjustment times of all temperature adjustment demand users in different environmental temperature intervals are counted to understand the service usage intensity of users. Frequent adjustment users are more sensitive to service quality fluctuations. Adjustment times: refers to the frequency of temperature adjustment operations performed by the system on behalf of users, reflecting the degree of user dependence on the function.

[0100] S42 determines the proportion of temperature adjustment demand users in suspected deviation users based on the overlap between the two, and takes it as the demand user proportion; In the above step, the demand overlap is calculated: the system found that among the 50 suspected deviation users, 40 were also temperature adjustment demand users, and the demand user proportion = 40 / 50 = 80%. Evaluate the overlap between model problems and core service users. If the problem is concentrated on high-value users, the risk is highest.

[0101] Demand user proportion: the proportion of temperature adjustment demand users in suspected deviation users, reflecting the severity of model instability.

[0102] S43 determines the update management strategy of the brain-machine staging model according to the demand user quantity proportion, the adjustment times of the temperature adjustment demand user in different environment temperature intervals, and the delay coefficient.

[0103] Further, it is judged whether the demand user quantity proportion is greater than a preset demand user quantity proportion threshold value. If yes, the update management strategy of the brain-machine staging model is determined as a preset update strategy. If no, the next step is entered. First judgment: demand user quantity proportion (80%) > preset demand user quantity proportion threshold value (assumed to be 60%), decision: condition is met! Immediately determine the update management strategy as "preset update strategy" (aggressive strategy).

[0104] At this time, more than 80% of the problem users are core service users, which means that the defects of the model directly impact the core value of the system. The fastest response (aggressive strategy) must be taken to repair the model to ensure the most important user experience.

[0105] S431 judges whether the number of adjustment demand users is greater than a preset demand user quantity threshold value. If yes, the next step is entered. If no, the update management strategy of the brain-machine staging model is determined as a second preset update strategy. Suppose another case: there are only 20 temperature adjustment demand users, and the demand user quantity proportion = 20 / 50 = 40%, 40% < 60% (threshold value), enter S431, and evaluate the absolute scale of demand users Action: judge the absolute number of adjustment demand users. Even if the proportion is not high, if the absolute number is large, it is still a group that cannot be ignored and needs further evaluation.

[0106] Preset demand user quantity threshold value: judge whether the demand user group reaches the critical value of the scale that needs to be focused on. Judgment: the number of adjustment demand users (20) is not greater than the preset demand user quantity threshold value (assumed to be 30). Decision: condition is not met. Determine the update management strategy as "second preset update strategy" (conservative strategy).

[0107] At this time, not only the coincidence degree is low, but also the absolute number is small. This indicates that the instability of the model has a narrow impact, and there are also few users who are highly dependent on the temperature control service. The system determines that the priority of this problem is low, so a conservative strategy is adopted to avoid unnecessary update cost.

[0108] S432 determines the number of adjustments of different temperature adjustment demand users in different environmental temperature intervals, determines the total number of adjustments of different temperature adjustment demand users in a single sleep process, determines whether the total number of adjustments of different temperature adjustment demand users in different sleep processes is greater than the preset adjustment number threshold, if yes, determines that the update management strategy of the brain-machine staging model is a preset update strategy, if no, proceeds to the next step; Assume the third case: there are 40 temperature adjustment demand users, but the demand user number ratio is 50% (less than the 60% threshold), and the number of adjustment demand users (40) > the preset demand user number threshold (30)? Yes, proceed to S432 to evaluate user activity Action: analyze the total number of adjustments of each of the 40 users in a single sleep process, and determine whether they are all greater than the preset threshold, identify high-activity users in the user group, and the active users have the strongest perception of service interruption.

[0109] Judgment: it is found that 35 users have a very high total number of adjustments, greater than the preset adjustment number threshold (e.g., the total number of adjustments in a single sleep process is greater than 10), but 5 are not active. Decision: Since it is not "all greater than", the condition is not met, proceed to S433.

[0110] S433 determines the adjustment demand factor of different temperature adjustment demand users according to the total number of adjustments of different temperature adjustment demand users and the delay coefficient, determines whether there is a temperature adjustment demand user whose adjustment demand factor is greater than the preset demand factor threshold, if yes, proceeds to the next step, if no, determines that the update management strategy of the brain-machine staging model is a second preset update strategy; In the above steps, the individual demand intensity is evaluated, and the action is to calculate the adjustment demand factor of each user. The formula is assumed to be: adjustment demand factor = average of total number of adjustments in different sleep processes * (1 - delay coefficient) * 0.1, for more detailed user value evaluation. A user who frequently uses (high total number of adjustments) and has good hardware performance (small delay coefficient) is a VIP user who cannot tolerate service degradation.

[0111] Noun explanation: adjustment demand factor: a composite index that combines user activity and hardware performance, used to identify high-value core users.

[0112] Judgment: the system calculates that among the 40 users, 10 users have a very high adjustment demand factor, exceeding the preset demand factor threshold of 0.5, i.e., decision: the condition is met (there are high demand factor users), proceed to S434.

[0113] S434 determines the update demand coefficient of the brain-machine staging model based on the adjustment demand factor of different temperature adjustment demand users and the proportion of the number of demand users, and determines the update management strategy of the brain-machine staging model according to the update demand coefficient.

[0114] The final comprehensive decision is to calculate the update demand coefficient, and the formula is: update demand coefficient = demand user number proportion × (number of high demand factor users / total number of adjustment demand users).

[0115] The example calculation is: update demand coefficient = 50% × (10 / 40) = 0.5 × 0.25 = 0.125, which is the final risk-reward trade-off. Combine "problem range" (proportion) and "core user affected degree" (high value user proportion) into a final score.

[0116] Update demand coefficient: a comprehensive urgency index that quantifies whether the current system state needs an urgent model update, the final judgment: update demand coefficient (0.125) > preset demand coefficient threshold (assuming 0.1), final decision: determine the update management strategy as "preset update strategy" Decision logic: the absolute number of demand users is large (40 people), but the proportion of real "high demand VIP users" is high (10 / 40=25%), and the overall demand user proportion is also high. The update demand coefficient (0.125) calculated by the comprehensive calculation reaches the threshold (0.1) of urgent update. Therefore, the system finally decides to take the preset update strategy.

[0117] Further, the adjustment demand factor is related to the total number of adjustments and the delay coefficient, wherein the more the total number of adjustments, the smaller the delay coefficient, and the larger the adjustment demand factor.

[0118] Further, the update demand coefficient of the brain-machine staging model is related to the adjustment demand factor of different temperature adjustment demand users and the proportion of the number of demand users, wherein the larger the adjustment demand factor, the larger the proportion of the number of demand users, and the higher the update demand coefficient.

[0119] Further, the update management strategy of the brain-machine staging model is determined according to the update demand coefficient, specifically including: When the update demand coefficient is greater than the preset demand coefficient threshold, the update management strategy of the brain-machine staging model is determined as the preset update strategy, and when the update demand coefficient is not greater than the preset demand coefficient threshold, the update management strategy of the brain-machine staging model is determined as the second preset update strategy.

[0120] It should be noted that the preset updating strategy is that when the sleep staging results of any one suspected deviation user under the same brain-computer feature combination are inconsistent, it is determined that the brain-computer staging model needs to be updated. The second preset updating strategy is that when the sleep staging results of multiple suspected deviation users under the same staging deviation combination are inconsistent, it is determined that the brain-computer staging model needs to be updated.

[0121] In another possible embodiment, in addition to the preset updating strategy or the second preset updating strategy for updating the brain-computer staging model, the brain-computer staging model also needs to be updated regularly according to a preset time period.

[0122] This embodies the robust operation and maintenance idea of combining "event triggering" and "time triggering". Event triggering (current process): for the discovered model performance degradation signal, perform purposeful emergency update. Time triggering (regular update): whether there is a problem or not, retrain the model with accumulated new data according to the plan (such as once a month). This can prevent potential degradation of the model and enable it to continuously learn new patterns of the user group.

[0123] This process is a macro-to-micro, layer-by-layer decision funnel that ensures the data-driven and economic nature of model update decisions. The system will not overreact to individual problems, nor will it do nothing when risks accumulate. Instead, it will make the most balanced and efficient management decisions, like an experienced operations engineer, by considering various factors.

[0124] Embodiment 2 In a second aspect, the present application provides a computer system, comprising a memory and a processor connected in communication, and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to perform the above-mentioned sleep temperature regulation method based on a non-invasive brain-computer interface.

[0125] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment mainly describes the difference from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiment.

[0126] The above-described embodiments of the application have special structure and can achieve the desired results. Other embodiments can have different structures and achieve the same results. The purpose of the above-described embodiments is to illustrate the principles of the application and not to limit the scope of the application. The scope of the application is defined by the claims and their equivalents. Other embodiments are within the scope of the claims.

[0127] The above description is merely illustrative of the embodiments of the present application and is not intended to limit the scope of the present application. Various modifications can be made by those skilled in the art based upon the teachings disclosed herein. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present application shall fall within the scope of the claims of the present application.

Claims

1. A sleep temperature regulation method based on a non-invasive brain-computer interface, characterized in that, Specifically comprising: Based on the analysis result of the agricultural product, determine the matching condition of the marketing selling point of the agricultural product and the user portrait of the existing marketing account, and based on the matching condition, if there is no matching deviation agricultural product in the agricultural product, proceed to the next step; Based on the matching condition, determine the available marketing selling point in the agricultural product, based on the constituent data of the marketing selling point, and combine the matching condition of the available marketing selling point and different user portraits, determine the marketing association type of the marketing account in the agricultural product; According to the marketing association type of the marketing account in different agricultural products, determine the data processing method of the update processing of the user portrait of the marketing account, and based on the update result of the user portrait of different marketing accounts, determine the generation processing method of the marketing selling point of the agricultural product.

2. The sleep temperature regulation method based on a non-invasive brain-computer interface according to claim 1, wherein, The analysis result of the agricultural product includes the marketing selling point of the agricultural product. 3.The sleep temperature regulation method based on non-invasive brain-computer interface of claim 1, wherein, The matching condition of the marketing selling point of the agricultural product and the user portrait of the existing marketing account is determined according to whether the customer of the user portrait matches the marketing selling point. 4.The sleep temperature regulation method based on non-invasive brain-computer interface of claim 1, wherein, Whether the customer of the user portrait is interested in the marketing selling point is determined according to the purchase quantity proportion of the customer of the user portrait in the goods of the marketing selling point. 5.The sleep temperature regulation method based on non-invasive brain-computer interface of claim 1, wherein, Determine that there is no matching deviation agricultural product in the agricultural product, specifically comprising: Determine the marketing account matched with the marketing selling point of the agricultural product based on the matching condition of the user portrait of the existing different marketing account and the agricultural product, and take it as the matching marketing account; According to the matching marketing account of different agricultural products, determine whether there is a matching deviation agricultural product in the agricultural product.

6. The sleep temperature regulation method based on non-invasive brain-computer interface according to claim 5, wherein, If there is a matching marketing account for different agricultural products, it is determined that there is no matching deviation agricultural product in the agricultural product.

7. The sleep temperature regulation method based on non-invasive brain-computer interface of claim 1, wherein, The available marketing selling point in the agricultural product is the marketing selling point that has a matching marketing account, that is, if the user of the user portrait of the marketing account is interested in the marketing selling point of the agricultural product, it is determined that the marketing selling point of the agricultural product is the available marketing selling point in the agricultural product.

8. The sleep temperature regulation method based on non-invasive brain-computer interface of claim 1, wherein, The determination method of the generation processing method of the marketing selling point of the agricultural product is: Based on the matching marketing account of the agricultural product, determine the matching marketing account of the agricultural product and the available marketing selling point matched with the matching marketing account; According to the update result of the user portrait of different marketing accounts, determine the marketing account matched with the marketing selling point of the agricultural product after updating; Based on the matching marketing account of the agricultural product and the available marketing selling point matched with the matching marketing account, and combined with the marketing account matched with the marketing selling point of the agricultural product after updating, determine the generation processing method of the marketing selling point of the agricultural product. 9.The sleep temperature regulation method based on non-invasive brain-computer interface of claim 8, wherein, Based on the matching marketing account of the agricultural product and the available marketing selling point matched with the matching marketing account, and combined with the marketing account matched with the marketing selling point of the agricultural product after updating, determine the generation processing method of the marketing selling point of the agricultural product, specifically comprising: The number of available marketing selling points of the agricultural product is acquired, and when the number of available marketing selling points of the agricultural product is greater than a preset marketing selling point number threshold, it is determined that the generation processing method of the marketing selling point of the agricultural product is that whether the updated marketing account is matched with the marketing selling point outside the available marketing selling point or not, the identification processing of the marketing selling point outside the available marketing selling point is not needed.

10. A computer system comprising: The memory and the processor connected in communication, and the computer program stored on the memory and capable of running on the processor, characterized in that the processor executes the computer program to perform the sleep temperature regulation method based on the non-invasive brain-computer interface according to any one of claims 1-9.

Images