Multi-modal physical and psychological health management scheme generation method and system based on deep learning

Abstract

The invention discloses a multi-modal physical and psychological health management scheme generation method and system based on deep learning, and relates to the technical field of physical and psychological health management, and the method comprises the following steps: collecting basic feature information and physiological index information of a user based on a wearable device, and marking a first similar user; screening a second similar user, and establishing premature beat reference information of the user; establishing a physiological index prediction model according to the physiological index information of the user, and performing physiological index prediction to obtain predicted index information; according to the premature beat reference information and the prediction index information of the user, early warning is carried out on the heart premature beat of the user, and a premature beat management scheme is generated; the method is used for solving the problem that when an existing physical and psychological health management technology carries out health management on the heart premature beat, accurate assessment and early warning cannot be carried out on the heart premature beat of the user according to the difference of physiological characteristics of different users under the condition that the data volume of physiological data of the user about the heart premature beat is small.

Description

Technical Field

[0001] The present invention relates to the technical field of physical and mental health management, and specifically to a method and system for generating a multi-modal physical and mental health management solution based on deep learning. Background Art

[0002] Physical and mental health management technology refers to the general term for a series of technical means that comprehensively use a variety of scientific and technological methods to comprehensively monitor, accurately evaluate and warn, effectively intervene, and continuously track the physical and mental states of individuals or groups, so as to maintain and promote physical and mental health and improve the quality of life.

[0003] When the existing physical and mental health management technology conducts health management on premature beats, it often fails to achieve effective evaluation and accurate warning. Because premature beats are a cardiac rhythm disorder phenomenon, and most people's premature beats do not occur frequently, belonging to sporadic events; a premature beat may occur only once in a long period of time, which makes it difficult to obtain a large amount of physiological data of the user himself when having premature beats during daily monitoring; moreover, everyone's physiological state is unique, and the occurrence of premature beats may be affected by various factors, such as exercise, diet, emotion, sleep, and environment, etc.; relying on the user's own single or small amount of premature beat data cannot capture the special physiological signs when premature beats occur. For example, in the patent application with the publication number CN116933046A, a method and system for generating a multi-modal health management solution based on deep learning are disclosed. Although this solution predicts future health data, when warning of premature beats, due to the lack of physiological data of the user himself when having premature beats, it is impossible to accurately screen out the data corresponding to possible premature beats in the predicted data. Therefore, when the existing physical and mental health management technology conducts health management on premature beats, it cannot accurately evaluate and warn the occurrence of premature beats in users according to the differences in the physiological characteristics of different users when the amount of physiological data of the user himself about premature beats is small. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems in the prior art to some extent. By collecting the basic characteristic information and physiological index information of the user based on a wearable device, and marking the first similar users; screening the second similar users, and establishing the premature beat reference information of the user; establishing a physiological index prediction model, and conducting physiological index prediction to obtain predicted index information; warning the user's premature beats according to the premature beat reference information and predicted index information of the user, and generating a premature beat management plan; to solve the problem that when the existing physical and mental health management technology conducts health management on premature beats, it cannot accurately evaluate and warn the occurrence of premature beats in users according to the differences in the physiological characteristics of different users when the amount of physiological data of the user himself about premature beats is small.

[0005] To achieve the above object, in a first aspect, the present application provides a method for generating a multimodal physical and mental health management solution based on deep learning, including the following steps: Collect the basic characteristic information and physiological index information of the user based on a wearable device, and mark the first similar user according to the basic characteristic information of the user; Screen the second similar user according to the physiological index information of the user, and establish the premature beat reference information of the user according to the physiological index information when the user and the second similar user have premature ventricular contractions; Establish a physiological index prediction model according to the physiological index information of the user, and perform physiological index prediction to obtain predicted index information; Warn the user's premature ventricular contractions according to the premature beat reference information and predicted index information of the user, and generate a premature beat management solution.

[0006] Further, collecting the basic characteristic information and physiological index information of the user based on a wearable device, and marking the first similar user according to the basic characteristic information of the user includes the following sub-steps: Use the wearable device to collect the gender, age and height of the user, collect the weight of the user at a first time interval, and collect the geographical location of the user at a second time interval; record the gender, age, height, weight and geographical location of the user as the basic characteristic information of the user, the first time interval is t1, and the second time interval is t2; And use the wearable device to obtain the heart rate, respiratory rate, body temperature and blood oxygen saturation of the user at a third time interval, and record them as the physiological index information of the user, and the third time interval is t3.

[0007] Further, collecting the basic characteristic information and physiological index information of the user based on a wearable device, and marking the first similar user according to the basic characteristic information of the user further includes the following sub-steps: According to the basic characteristic information of the user, set the age similarity threshold as a1, the height similarity threshold as a2, the weight similarity threshold as a3, and the distance similarity threshold as a4; mark other users with the same gender as the user and simultaneously satisfying that the age difference is less than or equal to a1, the height difference is less than or equal to a2, the weight difference is less than or equal to a3, and the distance difference of the geographical location is less than or equal to a4 as the first similar user of the user.

[0008] Further, screening the second similar user according to the physiological index information of the user, and establishing the premature beat reference information of the user according to the physiological index information when the user and the second similar user have premature ventricular contractions includes the following sub-steps: Set the first time length as R1, obtain the physiological index information of the user within R1 time, and calculate the average heart rate, average respiratory rate, average body temperature, and average blood oxygen saturation per R0 hours within R1 time, which are recorded as the user's similarity reference information; Obtain the similarity reference information of the user's first similar user within R1 time; according to the user's similarity reference information and the corresponding similarity reference information of the first similar user, calculate the heart rate similarity, respiratory rate similarity, body temperature similarity, and blood oxygen saturation similarity of the user and the corresponding first similar user per R0 hours within R1 time through the similarity calculation formula, which are recorded as the first similar data; the similarity calculation formula is as follows: , where D represents similarity, X represents the average value of a certain physiological index information of the user per R0 hours, and Y represents the average value of a certain physiological index information of the corresponding first similar user per R0 hours; Set the heart rate similarity threshold b1, respiratory rate similarity threshold b2, body temperature similarity threshold b3, and blood oxygen saturation similarity threshold b4; For any first similar user of the user, if it satisfies that the heart rate similarity of the user and the first similar user per R0 hours within R1 time is not less than b1, the respiratory rate similarity is not less than b2, the body temperature similarity is not less than b3, and the blood oxygen saturation similarity is not less than b4; then mark the first similar user as the second similar user.

[0009] Furthermore, screen the second similar users according to the user's physiological index information, and establish the premature beat reference information of the user according to the physiological index information of the user and the second similar users when premature beats occur. The specific steps are as follows: For the first similar data of any second similar user of the user, according to the heart rate similarity, respiratory rate similarity, body temperature similarity, and blood oxygen saturation similarity per R0 hours within R1 time in the first similar data, calculate the average value of the heart rate similarity, the average value of the respiratory rate similarity, the average value of the body temperature similarity, and the average value of the blood oxygen saturation similarity within R1 time respectively, and record them as the heart rate similarity V1, respiratory rate similarity V2, body temperature similarity V3, and blood oxygen saturation similarity V4 in sequence, which are marked as the second similar data of the corresponding second similar user; Obtain the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the user when premature beats occur, which are recorded as the first premature beat physiological information, and obtain the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the user's second similar user when premature beats occur, which are recorded as the second premature beat physiological information; record the user's first premature beat physiological information and the corresponding second premature beat physiological information as the user's premature beat reference information.

[0010] Further, a physiological index prediction model is established based on the user's physiological index information, and physiological index prediction is performed. The steps to obtain the predicted index information include the following sub-steps: The physiological index information of the user is normalized separately according to the data type, and all data sizes are scaled to [0, 1] to obtain the first physiological data; The first physiological data is divided into a physiological training set and a physiological test set in a ratio of 8:2; An original prediction model is constructed based on the long short-term memory network. The original prediction model includes an input layer, an LSTM layer, and an output layer. Set the number of neurons in the input layer to c1, the number of neurons in the LSTM layer to c2, and the number of neurons in the output layer to c3. Use the physiological training set to train the original prediction model, and after completion, obtain the first prediction model; Use the physiological test set to test the first prediction model, and use the mean absolute error formula to calculate the mean absolute error of the first prediction model predicting heart rate, respiratory rate, body temperature, and blood oxygen saturation respectively. Denote them as the heart rate prediction error e1, the respiratory rate prediction error e2, the body temperature prediction error e3, and the blood oxygen saturation prediction error e4 in sequence; The mean absolute error formula is as follows: , where E0 represents the mean absolute error, n represents the number of samples input into the model, J represents the obtained true data value, and K represents the predicted data value; Set the first pre-difference threshold f1, the second pre-difference threshold f2, the third pre-difference threshold f3, and the fourth pre-difference threshold f4; Determine whether the first prediction model satisfies e1 ≤ f1 and e2 ≤ f2 and e3 ≤ f3 and e4 ≤ f4. If it is satisfied, determine that the first prediction model is qualified and mark it as the physiological index prediction model. If it is not satisfied, determine that the first prediction model is unqualified, and use the physiological training set to train the first prediction model again until the physiological index prediction model is obtained; And record the heart rate prediction error, respiratory rate prediction error, body temperature prediction error, and blood oxygen saturation prediction error corresponding to the physiological index prediction model, denoted as G1, G2, G1, and G2 in sequence, and mark them as prediction error parameters.

[0011] Further, a physiological index prediction model is established based on the user's physiological index information, and physiological index prediction is performed. The steps to obtain the predicted index information further include the following sub-steps: According to the user's physiological index information, obtain the physiological index information of the latest collected second time length, perform normalization processing, and then input it into the physiological index prediction model; Obtain the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the future third time length, and mark them as the predicted index information. The second time length is R2, and the third time length is R3.

[0012] Further, based on the user's premature beat reference information and prediction index information, a warning is issued for the user's premature beats in the heart, and a premature beat management plan is generated, including the following sub-steps: Combine the heart rate, respiratory rate, body temperature, and blood oxygen saturation at the same time in the prediction index information into a warning feature vector, denoted as M(m1, m2, m3, m4), where m1, m2, m3, and m4 represent the heart rate, respiratory rate, body temperature, and blood oxygen saturation in the prediction index information in sequence; Based on the first premature beat physiological information, combine the heart rate, respiratory rate, body temperature, and blood oxygen saturation of any premature beat of the user into a first reference feature vector, denoted as Q(q1, q2, q3, q4), where q1, q2, q3, and q4 represent the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the first premature beat physiological information in sequence; Based on the second similar data and the second premature beat physiological information, combine the heart rate, respiratory rate, body temperature, and blood oxygen saturation of any premature beat of the user into a second reference feature vector, denoted as P(p1, p2, p3, p4), where p1, p2, p3, and p4 represent the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the second premature beat physiological information in sequence; and combine the second similar data of the second similar user corresponding to P(p1, p2, p3, p4) into a similar feature vector, denoted as PV(v1, v2, v3, v4).

[0013] Further, based on the user's premature beat reference information and prediction index information, a warning is issued for the user's premature beats in the heart, and a premature beat management plan is generated, which also includes the following sub-steps: For any warning feature vector M(m1, m2, m3, m4), any first reference feature vector Q(q1, q2, q3, q4), any second reference feature vector P(p1, p2, p3, p4), and the corresponding similar feature vector PV(v1, v2, v3, v4) of the user; If m1 satisfies any q1 ∈ [m1 - G1, m1 + G1], or satisfies the existence of any [m1 - G1, m1 + G1] ∩ [p1 * v1, p1 / v1], and [(au12 - au11) / (2 * G1)] ≥ k0, then record xm1 = 1, otherwise record xm1 = 0; where [au11, au12] = [m1 - G1, m1 + G1] ∩ [p1 * v1, p1 / v1], and k0 is the set ratio threshold; If m2 satisfies that for any q2 ∈ [m2 - G2, m2 + G2], or satisfies that for any [m2 - G2, m2 + G2] ∩ [p2 * v2, p2 / v2] exists, and [(au22 - au21) / (2 * G2)] ≥ k0, then record xm2 = 1; otherwise, record xm2 = 0; where [au21, au22] = [m2 - G2, m2 + G2] ∩ [p2 * v2, p2 / v2]; If m3 satisfies that for any q3 ∈ [m3 - G3, m3 + G3], or satisfies that for any [m3 - G3, m3 + G3] ∩ [p3 * v3, p3 / v3] exists, and [(au32 - au31) / (2 * G3)] ≥ k0, then record xm3 = 1; otherwise, record xm3 = 0; where [au31, au32] = [m3 - G3, m3 + G3] ∩ [p3 * v3, p3 / v3]; If m4 satisfies that for any q4 ∈ [m4 - G4, m4 + G4], or satisfies that for any [m4 - G4, m4 + G4] ∩ [p4 * v4, p4 / v4] exists, and [(au42 - au41) / (2 * G4)] ≥ k0, then record xm1 = 4; otherwise, record xm4 = 0; where [au41, au42] = [m4 - G4, m4 + G4] ∩ [p4 * v4, p4 / v4]; If for any warning feature vector M(m1, m2, m3, m4) of the user, (xm1 + xm2 + xm3 + xm4) = 3, then obtain the time corresponding to this warning feature vector, send a premature atrial contraction prediction message to the user, and obtain the premature atrial contraction countermeasures from the database and send them to the user; If for any warning feature vector M(m1, m2, m3, m4) of the user, (xm1 + xm2 + xm3 + xm4) = 4, then obtain the time corresponding to this warning feature vector, send a premature atrial contraction warning message to the user, and obtain the premature atrial contraction countermeasures from the database and send them to the user.

[0014] In a second aspect, the present application provides a multi-modal physical and mental health management solution generation system based on deep learning, including a data acquisition module, an index processing module, an information prediction module, and a premature beat management module; The data acquisition module includes a collection unit and a preliminary screening unit. The collection unit collects the basic feature information and physiological index information of the user based on a wearable device, and the preliminary screening unit marks the first similar users according to the basic feature information of the user; The index processing module includes a second screening unit and a processing unit. The second screening unit screens the second similar users according to the physiological index information of the user, and the processing unit establishes the premature beat reference information of the user according to the physiological index information of the user and the second similar users when they have premature atrial contractions; The information prediction module includes a model unit and a prediction unit. The model unit establishes a physiological index prediction model based on the physiological index information of the user, and the prediction unit is used to perform physiological index prediction to obtain prediction index information; The premature beat management module warns of the user's cardiac premature beats according to the premature beat reference information and prediction index information of the user, and generates a premature beat management plan.

[0015] Advantages of the present invention: The present invention collects the basic characteristic information and physiological index information of the user based on a wearable device, and marks the first similar user according to the basic characteristic information of the user; screens the second similar user according to the physiological index information of the user, and establishes the premature beat reference information of the user according to the physiological index information when the user and the second similar user have cardiac premature beats; establishes a physiological index prediction model according to the physiological index information of the user, and performs physiological index prediction to obtain prediction index information; warns of the user's cardiac premature beats according to the premature beat reference information and prediction index information of the user, and generates a premature beat management plan; it can, when the amount of physiological data of the user himself about cardiac premature beats is small, collect the relevant data of users with similar physiological characteristics according to the differences in the physiological characteristics of different users, and accurately evaluate and warn of the user's cardiac premature beats; The present invention marks and screens users with similar physiological characteristics and close geographical locations, and collects the relevant data of similar users. The advantage is that it can integrate the premature beat data of similar groups of people, provide richer reference data for each user, make up for the lack of their personal data, and improve the accuracy of predicting their premature beat risks; geographical location is added when marking and screening users because climate factors have a certain impact on cardiac premature beats. Selecting users with close geographical locations can reduce the difference in physiological effects caused by climate differences, so as to more accurately analyze the relevant characteristics of cardiac premature beats; when warning of cardiac premature beats, the prediction data is corrected according to the similarity of the user and the model error before judgment, which can effectively reduce the deviation caused by individual differences, thereby improving the accuracy of the overall warning. Brief Description of the Drawings

[0016] Figure 1 It is a principle block diagram of the system of the present invention; Figure 2 It is a step flow chart of the method of the present invention; Figure 3 It is an original prediction model structure diagram of the present invention; Figure 4 It is a structural schematic diagram of the electronic device of the present invention. Detailed Embodiments

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] Example 1. Refer to Figure 1 As shown, the present application provides a multi-modal physical and mental health management solution generation system based on deep learning, including a data collection module, an index processing module, an information prediction module, and a premature beat management module; The data collection module includes a collection unit and a preliminary screening unit. The collection unit collects the basic characteristic information and physiological index information of the user based on a wearable device, and the preliminary screening unit marks the first similar user according to the basic characteristic information of the user. The collection unit is configured with a data collection strategy, and the data collection strategy includes: using the wearable device to collect the gender, age, and height of the user, collecting the weight of the user at a first time interval, and collecting the geographical location of the user at a second time interval; recording the gender, age, height, weight, and geographical location of the user as the basic characteristic information of the user. The first time interval is t1, and the second time interval is t2; in this embodiment, the first time interval t1 is 1 month, and the second time interval t2 is 1 day; because the gender of the user is an invariant, and the age is a regular variable, so only one collection is required. The weight changes slowly, so it needs to be collected and updated regularly; while the geographical location can be obtained in the background through the positioning system of the wearable device. And using the wearable device to obtain the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the user at a third time interval, and recording them as the physiological index information of the user. The third time interval is t3; in this embodiment, the third time interval t3 is 3 minutes, and it can be set according to the actual application scenario. The preliminary screening unit is configured with a user preliminary screening strategy, and the user preliminary screening strategy includes: according to the basic characteristic information of the user, setting the age similarity threshold as a1, the height similarity threshold as a2, the weight similarity threshold as a3, and the distance similarity threshold as a4; marking other users who have the same gender as the user and simultaneously satisfy that the age difference is less than or equal to a1, the height difference is less than or equal to a2, the weight difference is less than or equal to a3, and the distance difference of the geographical location is less than or equal to a4 as the first similar users of the user; in this embodiment, a1 = 3 years old, a2 = 5 cm, a3 = 8 kg, a4 = 80 km. For example, user 1, 23 years old, 175 cm, 56 kg; user 2, 25 years old, 178 cm, 60 kg; the two are 60 km apart, then user 1 and user 2 are each other's first similar users. In the specific implementation process, an update period can be set when marking the first similar users; for example, the first similar users of the users are updated every 3 days; the geographical location is added during the marking because the climate conditions are usually similar in areas with close geographical locations, such as temperature, humidity, air pressure, etc.; these climate factors have a certain impact on premature beats of the heart; selecting similar users with close geographical locations can make the research objects in a similar climate environment, reduce the physiological impact differences caused by climate differences, and thus improve the accuracy of subsequent early warnings.

[0019] The index processing module includes a second screening unit and a processing unit. The second screening unit screens the second similar users according to the physiological index information of the users, and the processing unit establishes the premature beat reference information of the users according to the physiological index information of the users and the second similar users when premature beats of the heart occur. The second screening unit is configured with a user second screening strategy. The user second screening strategy includes: setting the first time length as R1, obtaining the physiological index information of the user within R1 time, and calculating the average heart rate, average respiratory rate, average body temperature, and average blood oxygen saturation every R0 hours within R1 time, which is recorded as the similar reference information of the user; in this embodiment, the first time length R1 is 7 days, and R0 = 2, that is, calculating the average heart rate, average respiratory rate, average body temperature, and average blood oxygen saturation every 2 hours within 7 days. Obtain the similar reference information of the first similar users of the user within R1 time; according to the similar reference information of the user and the corresponding similar reference information of the first similar users, calculate the heart rate similarity, respiratory rate similarity, body temperature similarity, and blood oxygen saturation similarity of the user and the corresponding first similar users every R0 hours within R1 time through the similarity calculation formula, which is recorded as the first similar data; the similarity calculation formula is as follows: , where D represents the similarity, X represents the average value of a certain physiological index information of the user every R0 hours, and Y represents the average value of a certain physiological index information of the corresponding first similar user every R0 hours; each user may have multiple first similar users, and calculations need to be performed for each first similar user. Set the heart rate similarity threshold b1, respiratory rate similarity threshold b2, body temperature similarity threshold b3, and blood oxygen saturation similarity threshold b4; in this embodiment, b1 = b2 = b3 = b4 = 0.7; it can be set according to the actual application scenario. If there are fewer first similar users of the user or fewer second similar users screened out, the threshold can be appropriately reduced. For any first similar user of the user, if it satisfies that the heart rate similarity between the user and the first similar user is not less than b1 every R0 hours within R1 time, the respiratory rate similarity is not less than b2, the body temperature similarity is not less than b3, and the blood oxygen saturation similarity is not less than b4; then mark the first similar user as the second similar user; for example, R1 = 7 days, R0 = 2 hours, 7 * 24 / 2 = 84; there will be 84 similarities for each indicator, and it is necessary to ensure that all 84 similarities are not less than the set threshold, and the same is true for the four indicators; The processing unit is configured with an indicator processing strategy, and the indicator processing strategy includes: for the first similar data of any second similar user of the user, according to the heart rate similarity, respiratory rate similarity, body temperature similarity, and blood oxygen saturation similarity in the first similar data every R0 hours within R1 time, respectively calculate the average value of the heart rate similarity, the average value of the respiratory rate similarity, the average value of the body temperature similarity, and the average value of the blood oxygen saturation similarity within R1 time, and record them in sequence as the heart rate similarity V1, the respiratory rate similarity V2, the body temperature similarity V3, and the blood oxygen saturation similarity V4, and mark them as the second similar data of the corresponding second similar user; for example, R1 = 7 days, R0 = 2 hours; there will be 84 similarities for each indicator, then calculate the average value of these 84 similarities as the similarity of this indicator, which is used for error correction during subsequent early warnings; Obtain the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the user when a premature ventricular contraction occurs, and record them as the first premature ventricular contraction physiological information, and obtain the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the second similar user of the user when a premature ventricular contraction occurs, and record them as the second premature ventricular contraction physiological information; record the first premature ventricular contraction physiological information of the user and the corresponding second premature ventricular contraction physiological information as the premature ventricular contraction reference information of the user; The reason for selecting the heart rate, respiratory rate, body temperature, and blood oxygen saturation as the premature ventricular contraction reference information is that they are closely related to heart function, and most wearable devices can obtain the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the user more accurately; In the specific implementation process, the premature ventricular contractions of most people occur sporadically, and individual premature ventricular contraction data is scarce. Through secondary screening, it is possible to find a group of people who are highly similar to the target individual in terms of physiological characteristics, living environment, etc.; this precise matching is not simply based on a single factor, but comprehensively considers multiple dimensions, making the found similar users more targeted and providing more fitting premature ventricular contraction data reference for the target individual.

[0020] The information prediction module includes a model unit and a prediction unit. The model unit establishes a physiological index prediction model based on the physiological index information of the user, and the prediction unit is used for physiological index prediction to obtain prediction index information; The model unit is configured with a model building strategy, which includes: normalizing the physiological index information of the user according to the data type respectively, scaling all data sizes to [0, 1] to obtain the first physiological data; among them, since the maximum value of blood oxygen saturation is less than or equal to 100%, it can be not normalized, but only change the unit. Divide the first physiological data into a physiological training set and a physiological test set according to a ratio of 8:2. Construct an original prediction model based on the long short-term memory network, please refer to Figure 3 As shown, the original prediction model includes an input layer, an LSTM layer, and an output layer. Set the number of neurons in the input layer as c1, the number of neurons in the LSTM layer as c2, and the number of neurons in the output layer as c3. Use the physiological training set to train the original prediction model, and after completion, obtain the first prediction model; in this embodiment, c1 = c3 = 4, and there are only four types of data for input and output, namely heart rate, respiratory rate, body temperature, and blood oxygen saturation; C2 = 128, and c2 can be set according to the actual application scenario. Use the physiological test set to test the first prediction model, and use the mean absolute error formula to calculate the mean absolute error of the first prediction model predicting heart rate, respiratory rate, body temperature, and blood oxygen saturation respectively, and record them in order as the heart rate prediction error e1, the respiratory rate prediction error e2, the body temperature prediction error e3, and the blood oxygen saturation prediction error e4; the mean absolute error formula is as follows: , where E0 represents the mean absolute error, n represents the number of samples input into the model, J represents the obtained real data value, and K represents the predicted data value; the smaller the mean absolute error, the higher the accuracy of the model. Set the first pre-difference threshold f1, the second pre-difference threshold f2, the third pre-difference threshold f3, and the fourth pre-difference threshold f4; judge whether the first prediction model satisfies e1 ≤ f1 and e2 ≤ f2 and e3 ≤ f3 and e4 ≤ f4. If it is satisfied, judge that the first prediction model is qualified and mark it as a physiological index prediction model. If it is not satisfied, judge that the first prediction model is unqualified, and use the physiological training set to train the first prediction model again until a physiological index prediction model is obtained; in this embodiment, e1 = 10, e2 = 2, e3 = 0.5, e4 = 0.12. For example, in a test, f1 = 8, f2 = 2.2, f3 = 0.3, f4 = 0.10, and since f2 > e2, it is judged that the first prediction model is unqualified. And record the heart rate prediction error, respiratory rate prediction error, body temperature prediction error, and blood oxygen saturation prediction error corresponding to the physiological index prediction model, and record them in order as G1, G2, G1, and G2 respectively, and mark them as prediction error parameters. The prediction unit is configured with a physiological prediction strategy, and the physiological prediction strategy includes: according to the physiological index information of the user, obtaining the physiological index information of the latest collected second time length, performing normalization processing, and then inputting it into the physiological index prediction model; obtaining the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the future third time length, marked as prediction index information, the second time length is R2, and the third time length is R3; in this real-time example, the second time length is 24 hours, and the third time length is 1 hour, that is, using the data of the previous 24 hours to predict the data of the next 1 hour; normally, R2 should be much larger than R3 to ensure the accuracy of the prediction; In the specific implementation process, a prediction period can be set to regularly predict the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the user. For example, if the prediction period is set to 1 hour, the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the user are predicted every hour.

[0021] The premature beat management module warns the user's cardiac premature beats according to the premature beat reference information and prediction index information of the user, and generates a premature beat management plan; The premature beat management module is configured with a premature beat management strategy, and the premature beat management strategy includes: combining the heart rate, respiratory rate, body temperature, and blood oxygen saturation at the same time in the prediction index information into a warning feature vector, denoted as M(m1, m2, m3, m4), where m1, m2, m3, and m4 represent the heart rate, respiratory rate, body temperature, and blood oxygen saturation in the prediction index information in sequence; Based on the first premature beat physiological information, combine the heart rate, respiratory rate, body temperature, and blood oxygen saturation of any premature beat of the user into a first reference feature vector, denoted as Q(q1, q2, q3, q4), where q1, q2, q3, and q4 represent the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the first premature beat physiological information in sequence; Based on the second similar data and the second premature beat physiological information, combine the heart rate, respiratory rate, body temperature, and blood oxygen saturation of any premature beat of the user into a second reference feature vector, denoted as P(p1, p2, p3, p4), where p1, p2, p3, and p4 represent the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the second premature beat physiological information in sequence; and combine the second similar data of the second similar user corresponding to P(p1, p2, p3, p4) into a similar feature vector, denoted as PV(v1, v2, v3, v4); each second similar user has a corresponding similar feature vector; For any warning feature vector M (m1, m2, m3, m4) of a user, any first reference feature vector Q (q1, q2, q3, q4), any second reference feature vector P (p1, p2, p3, p4), and the corresponding similarity feature vector PV (v1, v2, v3, v4); If m1 satisfies any q1 ∈ [m1 - G1, m1 + G1], or if any [m1 - G1, m1 + G1] ∩ [p1 * v1, p1 / v1] exists, it means that the feature corresponding to the premature beat of m1 is within the error range; and [(au12 - au11) / (2 * G1)] ≥ k0, then the error range is corrected once to ensure the correctness of the judgment; then record xm1 = 1, otherwise record xm1 = 0; where [au11, au12] = [m1 - G1, m1 + G1] ∩ [p1 * v1, p1 / v1], and k0 is the set ratio threshold; in this embodiment, K0 = 0.5; for example, in a certain case, m1 = 83 beats per minute, G1 = 8, q1 = 92, v1 = 0.88 for a certain second similar user, p1 = 90, so q1 is not within [75, 91], but [m1 - G1, m1 + G1] ∩ [p1 * v1, p1 / v1] = [au11, au12] = [79, 91], and (au12 - au11) / (2 * G1) = (91 - 79) / (2 * 8) = 0.75, 0.75 > K0, so record xm1 = 1; If m2 satisfies any q2 ∈ [m2 - G2, m2 + G2], or if any [m2 - G2, m2 + G2] ∩ [p2 * v2, p2 / v2] exists, and [(au22 - au21) / (2 * G2)] ≥ k0, then record xm2 = 1, otherwise record xm2 = 0; where [au21, au22] = [m2 - G2, m2 + G2] ∩ [p2 * v2, p2 / v2]; If m3 satisfies any q3 ∈ [m3 - G3, m3 + G3], or if any [m3 - G3, m3 + G3] ∩ [p3 * v3, p3 / v3] exists, and [(au32 - au31) / (2 * G3)] ≥ k0, then record xm3 = 1, otherwise record xm3 = 0; where [au31, au32] = [m3 - G3, m3 + G3] ∩ [p3 * v3, p3 / v3]; If m4 satisfies any q4 ∈ [m4 - G4, m4 + G4], or if any [m4 - G4, m4 + G4] ∩ [p4 * v4, p4 / v4] exists, and [(au42 - au41) / (2 * G4)] ≥ k0, then record xm1 = 4, otherwise record xm4 = 0; where [au41, au42] = [m4 - G4, m4 + G4] ∩ [p4 * v4, p4 / v4]; When making a judgment, each warning feature vector needs to be judged against all the first reference feature vectors and the second reference feature vectors; the proportional threshold k0 can be set according to the actual application scenario, K0 If (xm1 + xm2 + xm3 + xm4) = 3 for any warning feature vector M (m1, m2, m3, m4) of the user, then obtain the time corresponding to this warning feature vector, send a premature atrial contraction prediction message to the user, and obtain the coping methods for premature atrial contraction from the database and send them to the user; (xm1 + xm2 + xm3 + xm4) = 3 means that three of the four features are met, indicating a certain risk of premature beats; If (xm1 + xm2 + xm3 + xm4) = 4 for any warning feature vector M (m1, m2, m3, m4) of the user, then obtain the time corresponding to this warning feature vector, send a premature atrial contraction warning message to the user, and obtain the coping methods for premature atrial contraction from the database and send them to the user; (xm1 + xm2 + xm3 + xm4) = 4 means that three of the four features are met, indicating a relatively large risk of premature beats; In the specific implementation process, the coping methods for premature atrial contraction can be obtained through the network, or a relevant database can be established by oneself to obtain the coping methods for premature atrial contraction from the database.

[0022] Example 2, please refer to Figure 2 As shown, the present application provides the generation of a multi-modal physical and mental health management solution based on deep learning, including the following steps: Step S1, collect the basic feature information and physiological index information of the user based on the wearable device, and mark the first similar users according to the basic feature information of the user; Step S1 includes the following sub-steps: Step S101, use the wearable device to collect the gender, age and height of the user, collect the weight of the user at the first time interval, and collect the geographical location of the user at the second time interval; record the gender, age, height, weight and geographical location of the user as the basic feature information of the user, the first time interval is t1, and the second time interval is t2; Step S102, and use the wearable device to obtain the heart rate, respiratory rate, body temperature and blood oxygen saturation of the user at the third time interval, and record them as the physiological index information of the user, the third time interval is t3; Step S103, according to the basic feature information of the user, set the age similarity threshold as a1, the height similarity threshold as a2, the weight similarity threshold as a3, and the distance similarity threshold as a4; Step S104, mark other users who have the same gender as the user and at the same time meet the age difference less than or equal to a1, the height difference less than or equal to a2, the weight difference less than or equal to a3, and the distance difference of the geographical location less than or equal to a4 as the first similar users of the user.

[0023] Step S2: Screen the second similar users according to the physiological index information of the user, and establish the premature beat reference information of the user based on the physiological index information of the user and the second similar users when they have premature ventricular contractions. Step S2 includes the following sub-steps: Step S201: Set the first time length as R1, obtain the physiological index information of the user within R1 time, and calculate the average heart rate, average respiratory rate, average body temperature, and average blood oxygen saturation every R0 hours within R1 time, which are recorded as the similar reference information of the user. Step S202: Obtain the similar reference information of the first similar user of the user within R1 time; according to the similar reference information of the user and the corresponding similar reference information of the first similar user, calculate the heart rate similarity, respiratory rate similarity, body temperature similarity, and blood oxygen saturation similarity of the user and the corresponding first similar user every R0 hours within R1 time through the similarity calculation formula, which are recorded as the first similar data. The similarity calculation formula is as follows: , where D represents similarity, X represents the average value of a certain physiological index information of the user every R0 hours, and Y represents the average value of the corresponding physiological index information of the first similar user every R0 hours. Step S203: Set the heart rate similarity threshold b1, the respiratory rate similarity threshold b2, the body temperature similarity threshold b3, and the blood oxygen saturation similarity threshold b4. Step S204: For any first similar user of the user, if it satisfies that the heart rate similarity of the user and the first similar user every R0 hours within R1 time is not less than b1, the respiratory rate similarity is not less than b2, the body temperature similarity is not less than b3, and the blood oxygen saturation similarity is not less than b4; then mark the first similar user as the second similar user. Step S205: For the first similar data of any second similar user of the user, according to the heart rate similarity, respiratory rate similarity, body temperature similarity, and blood oxygen saturation similarity every R0 hours within R1 time in the first similar data, calculate the average value of the heart rate similarity, the average value of the respiratory rate similarity, the average value of the body temperature similarity, and the average value of the blood oxygen saturation similarity within R1 time respectively, and record them as the heart rate similarity V1, the respiratory rate similarity V2, the body temperature similarity V3, and the blood oxygen saturation similarity V4 in sequence, which are marked as the second similar data of the corresponding second similar user. Step S206: Obtain the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the user when having premature ventricular contractions, which are recorded as the first premature beat physiological information, and obtain the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the second similar user of the user when having premature ventricular contractions, which are recorded as the second premature beat physiological information; record the first premature beat physiological information of the user and the corresponding second premature beat physiological information as the premature beat reference information of the user.

[0024] Step S3: Establish a physiological index prediction model based on the user's physiological index information, and perform physiological index prediction to obtain predicted index information. Step S3 includes the following sub-steps: Step S301: Normalize the user's physiological index information according to the data type respectively, scale all data sizes to [0, 1] to obtain the first physiological data; Step S302: Divide the first physiological data into a physiological training set and a physiological test set according to a ratio of 8:2; Step S303: Construct an original prediction model based on the long short-term memory network. The original prediction model includes an input layer, an LSTM layer, and an output layer. Set the number of neurons in the input layer to c1, the number of neurons in the LSTM layer to c2, and the number of neurons in the output layer to c3; Step S304: Use the physiological training set to train the original prediction model, and obtain the first prediction model after completion; Step S305: Use the physiological test set to test the first prediction model, and use the mean absolute error formula to calculate the mean absolute error of the first prediction model for predicting heart rate, respiratory rate, body temperature, and blood oxygen saturation respectively, and record them in order as the heart rate prediction error e1, the respiratory rate prediction error e2, the body temperature prediction error e3, and the blood oxygen saturation prediction error e4. The mean absolute error formula is as follows: , where E0 represents the mean absolute error, n represents the number of samples input into the model, J represents the obtained true data value, and K represents the predicted data value; Step S306: Set the first pre-difference threshold f1, the second pre-difference threshold f2, the third pre-difference threshold f3, and the fourth pre-difference threshold f4; Determine whether the first prediction model satisfies e1 ≤ f1 and e2 ≤ f2 and e3 ≤ f3 and e4 ≤ f4. If it is satisfied, determine that the first prediction model is qualified and mark it as the physiological index prediction model; Step S307: If not satisfied, determine that the first prediction model is unqualified, and use the physiological training set to train the first prediction model again until the physiological index prediction model is obtained; Step S308: Record the heart rate prediction error, respiratory rate prediction error, body temperature prediction error, and blood oxygen saturation prediction error corresponding to the physiological index prediction model, and record them in order as G1, G2, G1, and G2 respectively, and mark them as prediction error parameters; Step S309: According to the user's physiological index information, obtain the physiological index information of the latest collected second time length, perform normalization processing, and then input it into the physiological index prediction model; Obtain the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the future third time length, and mark them as predicted index information. The second time length is R2, and the third time length is R3.

[0025] Step S4: Based on the user's premature beat reference information and prediction index information, give an early warning of the user's premature beats and generate a premature beat management plan. Step S4 includes the following sub-steps: Step S401: Combine the heart rate, respiratory rate, body temperature, and blood oxygen saturation at the same time in the prediction index information into an early warning feature vector, denoted as M (m1, m2, m3, m4), where m1, m2, m3, and m4 represent the heart rate, respiratory rate, body temperature, and blood oxygen saturation in the prediction index information in sequence; Step S402: Based on the first premature beat physiological information, combine the heart rate, respiratory rate, body temperature, and blood oxygen saturation of any premature beat of the user into a first reference feature vector, denoted as Q (q1, q2, q3, q4), where q1, q2, q3, and q4 represent the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the first premature beat physiological information in sequence; Step S403: Based on the second similar data and the second premature beat physiological information, combine the heart rate, respiratory rate, body temperature, and blood oxygen saturation of any premature beat of the user into a second reference feature vector, denoted as P (p1, p2, p3, p4), where p1, p2, p3, and p4 represent the heart rate, respiratory rate, body temperature, and blood oxygen saturation of the second premature beat physiological information in sequence; and combine the second similar data of the second similar user corresponding to P (p1, p2, p3, p4) into a similar feature vector, denoted as PV (v1, v2, v3, v4); Step S404: For any early warning feature vector M (m1, m2, m3, m4), any first reference feature vector Q (q1, q2, q3, q4), any second reference feature vector P (p1, p2, p3, p4), and the corresponding similar feature vector PV (v1, v2, v3, v4) of the user; Step S405: If m1 satisfies any q1 ∈ [m1 - G1, m1 + G1], or there exists any [m1 - G1, m1 + G1] ∩ [p1 * v1, p1 / v1], and [(au12 - au11) / (2 * G1)] ≥ k0, then record xm1 = 1, otherwise record xm1 = 0; where [au11, au12] = [m1 - G1, m1 + G1] ∩ [p1 * v1, p1 / v1], and k0 is the set ratio threshold; Step S406: If m2 satisfies any q2 ∈ [m2 - G2, m2 + G2], or there exists any [m2 - G2, m2 + G2] ∩ [p2 * v2, p2 / v2], and [(au22 - au21) / (2 * G2)] ≥ k0, then record xm2 = 1, otherwise record xm2 = 0; where [au21, au22] = [m2 - G2, m2 + G2] ∩ [p2 * v2, p2 / v2]; Step S407, if m3 satisfies any q3 ∈ [m3 - G3, m3 + G3], or satisfies that for any [m3 - G3, m3 + G3] ∩ [p3 * v3, p3 / v3] exists and [(au32 - au31) / (2 * G3)] ≥ k0, then record xm3 = 1; otherwise, record xm3 = 0. Here, [au31, au32] = [m3 - G3, m3 + G3] ∩ [p3 * v3, p3 / v3]; Step S408, if m4 satisfies any q4 ∈ [m4 - G4, m4 + G4], or satisfies that for any [m4 - G4, m4 + G4] ∩ [p4 * v4, p4 / v4] exists and [(au42 - au41) / (2 * G4)] ≥ k0, then record xm1 = 4; otherwise, record xm4 = 0. Here, [au41, au42] = [m4 - G4, m4 + G4] ∩ [p4 * v4, p4 / v4]; Step S409, if for any warning feature vector M(m1, m2, m3, m4) of the user, (xm1 + xm2 + xm3 + xm4) = 3, then obtain the time corresponding to this warning feature vector, send a premature atrial contraction prediction message to the user, and obtain the premature atrial contraction countermeasures from the database and send them to the user; Step S410, if for any warning feature vector M(m1, m2, m3, m4) of the user, (xm1 + xm2 + xm3 + xm4) = 4, then obtain the time corresponding to this warning feature vector, send a premature atrial contraction warning message to the user, and obtain the premature atrial contraction countermeasures from the database and send them to the user.

[0026] Example 3, please refer to Figure 4 as shown in Figure 4 illustrates a schematic structural diagram of an electronic device. The electronic device may include: a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete mutual communication through the communication bus. The memory stores computer-readable instructions, and the processor can call the instructions in the memory. When the computer-readable instructions are executed by the processor, it runs the steps in the method for generating a multi-modal physical and mental health management solution based on deep learning to achieve the following functions: collecting the basic feature information and physiological index information of the user based on the wearable device, and marking the first similar user according to the basic feature information of the user; screening the second similar user according to the physiological index information of the user, and establishing the premature atrial contraction reference information of the user according to the physiological index information of the user and the second similar user when premature atrial contraction occurs; establishing a physiological index prediction model according to the physiological index information of the user, and performing physiological index prediction to obtain prediction index information; warning the user's premature atrial contraction according to the premature atrial contraction reference information and the prediction index information of the user, and generating a premature atrial contraction management plan.

[0027] In addition, when the logic instructions in the above-mentioned memory are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0028] Embodiment 4, this application also provides a computer-readable storage medium. This application provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, it runs the steps in the above-mentioned method for generating a multi-modal physical and mental health management solution based on deep learning to achieve the following functions: collecting the basic characteristic information and physiological index information of a user based on a wearable device, and marking the first similar user according to the basic characteristic information of the user; screening the second similar user according to the physiological index information of the user, and establishing the premature beat reference information of the user according to the physiological index information of the user and the second similar user when the user has premature ventricular contractions; establishing a physiological index prediction model according to the physiological index information of the user, and performing physiological index prediction to obtain prediction index information; warning the user's premature ventricular contractions according to the premature beat reference information and the prediction index information of the user, and generating a premature beat management plan.

[0029] Through the description of the above embodiments, the embodiments of the present invention can be provided as a method, a system, or a computer program product. Based on such an understanding, the above technical solution, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disks, optical discs, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments.

[0030] In the embodiments provided in the present application, it should be understood that the disclosed system or method can be implemented in other ways. The above-described embodiments are merely illustrative. For example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation. For another example, multiple modules or units can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some communication interfaces. The indirect coupling or communication connection of systems, modules, and units can be electrical, mechanical, or other forms.

[0031] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A multimodal physical and mental health management program generation method based on deep learning, characterized in that: The steps include: Collecting basic characteristic information and physiological indicator information of the user based on the wearable device, and marking a first similar user according to the basic characteristic information of the user; Filtering a second similar user according to the user's physiological indicator information, and establishing the user's premature beat reference information according to the user's and the second similar user's physiological indicator information when premature beats occur; Establishing a physiological index prediction model according to the user's physiological index information, and performing physiological index prediction to obtain prediction index information; Based on the user's premature beat reference information and prediction indicator information, the user's premature beats are warned and a premature beat management plan is generated.

2. The method for generating a multimodal physical and mental health management program based on deep learning according to claim 1, characterized in that: Collecting basic characteristic information and physiological index information of the user based on the wearable device, and marking the first similar user according to the basic characteristic information of the user includes the following sub-steps: Using the wearable device to collect the user's gender, age, and height, and to collect the user's weight at a first time interval, and to collect the user's geographic location at a second time interval; The user's gender, age, height, weight and geographic location are recorded as the user's basic characteristic information, the first time interval is t1, and the second time interval is t2; The wearable device is used to obtain the user's heart rate, respiratory rate, body temperature and blood oxygen saturation at a third time interval, which are recorded as the user's physiological indicator information. The third time interval is t3.

3. The method for generating a multimodal physical and mental health management program based on deep learning according to claim 2 is characterized in that: Collecting basic characteristic information and physiological index information of the user based on the wearable device, and marking the first similar user according to the basic characteristic information of the user also includes the following sub-steps: According to the user's basic characteristic information, the age similarity threshold is set to a1, the height similarity threshold is set to a2, the weight similarity threshold is set to a3, and the distance similarity threshold is set to a4; other users who have the same gender as the user and who also meet the conditions that the age difference is less than or equal to a1, the height difference is less than or equal to a2, the weight difference is less than or equal to a3, and the geographic location distance difference is less than or equal to a4 are marked as the first similar users of the user.

4. The method for generating a multimodal physical and mental health management program based on deep learning according to claim 3 is characterized in that: The second similar user is selected according to the physiological index information of the user, and the premature beat reference information of the user is established according to the physiological index information of the user and the second similar user when premature beats occur, including the following sub-steps: Set the first time length to R1, obtain the user's physiological index information within the R1 time, and calculate the average heart rate, average respiratory rate, average body temperature, and average blood oxygen saturation for every R0 hours within the R1 time, and record them as the user's similar reference information; Obtain similar reference information of the first similar user of the user within time R1; According to the similar reference information of the user and the similar reference information of the corresponding first similar user, the heart rate similarity, respiratory rate similarity, body temperature similarity and blood oxygen saturation similarity of the user and the corresponding first similar user in each R0 hour of the R1 time are calculated by the similarity calculation formula, and recorded as the first similarity data; the similarity calculation formula is as follows: , where D represents similarity, X represents the average value of a physiological indicator information of a user in R0 hours, and Y represents the average value of a physiological indicator information of the corresponding first similar user in R0 hours; Set the heart rate similarity threshold b1, the respiratory rate similarity threshold b2, the body temperature similarity threshold b3, and the blood oxygen saturation similarity threshold b4; For any first similar user of the user, if the heart rate similarity between the user and the first similar user in every R0 hours within the R1 time is not less than b1, the breathing rate similarity is not less than b2, the body temperature similarity is not less than b3, and the blood oxygen saturation similarity is not less than b4; then the first similar user is marked as the second similar user.

5. The method for generating a multimodal physical and mental health management program based on deep learning according to claim 4 is characterized in that: The second similar user is selected according to the physiological index information of the user, and the premature beat reference information of the user is established according to the physiological index information of the user and the second similar user when premature beats occur, and the following sub-steps are further included: For the first similar data of any second similar user of the user, according to the heart rate similarity, respiratory rate similarity, body temperature similarity and blood oxygen saturation similarity in each R0 hour in the first similar data within the R1 time, the average value of the heart rate similarity, the average value of the respiratory rate similarity, the average value of the body temperature similarity and the average value of the blood oxygen saturation similarity in the R1 time are respectively obtained, and they are recorded in order as heart rate image V1, respiratory rate image V2, body temperature image V3 and blood oxygen saturation image V4, and marked as the second similar data of the corresponding second similar user; The heart rate, respiratory rate, body temperature and blood oxygen saturation of the user when the premature heart beat occurs are obtained, which is recorded as the first premature heart beat physiological information, and the heart rate, respiratory rate, body temperature and blood oxygen saturation of the second similar user of the user when the premature heart beat occurs are obtained, which is recorded as the second premature heart beat physiological information; The user's first premature beat physiological information and the corresponding second premature beat physiological information are recorded as the user's premature beat reference information.

6. The method for generating a multimodal physical and mental health management program based on deep learning according to claim 5, characterized in that: A physiological index prediction model is established based on the user's physiological index information, and physiological index prediction is performed to obtain the predicted index information, including the following sub-steps: Normalizing the user's physiological indicator information according to the data type, scaling all data sizes to [0, 1], and obtaining first physiological data; Dividing the first physiological data into a physiological training set and a physiological test set in a ratio of 8:2; An original prediction model is constructed based on a long short-term memory network. The original prediction model includes an input layer, an LSTM layer, and an output layer. The number of neurons in the input layer is set to c1, the number of neurons in the LSTM layer is set to c2, and the number of neurons in the output layer is set to c3. The original prediction model is trained using a physiological training set, and the first prediction model is obtained after completion. The first prediction model was tested using the physiological test set, and the mean absolute error formula was used to calculate the mean absolute errors of the first prediction model for predicting heart rate, respiratory rate, body temperature, and blood oxygen saturation, which were recorded in order as heart rate pre-difference e1, respiratory rate pre-difference e2, body temperature pre-difference e3, and blood oxygen saturation pre-difference e4; the mean absolute error formula is as follows: , where E0 represents the mean absolute error, n represents the number of samples input into the model, J represents the actual data value obtained, and K represents the predicted data value; Set a first pre-difference threshold value f1, a second pre-difference threshold value f2, a third pre-difference threshold value f3 and a fourth pre-difference threshold value f4; determine whether the first prediction model satisfies e1≤f1 and e2≤f2 and e3≤f3 and e4≤f4; if so, determine that the first prediction model is qualified and marked as a physiological indicator prediction model; if not, determine that the first prediction model is unqualified, and use the physiological training set to train the first prediction model again until a physiological indicator prediction model is obtained; The heart rate difference, respiratory rate difference, body temperature difference and blood oxygen saturation difference corresponding to the physiological index prediction model are recorded and denoted as G1, G2, G1 and G2 in order, and marked as prediction error parameters.

7. The method for generating a multimodal physical and mental health management program based on deep learning according to claim 6, characterized in that: Establishing a physiological index prediction model based on the user's physiological index information and performing physiological index prediction, obtaining the predicted index information also includes the following sub-steps: According to the user's physiological indicator information, the most recently collected physiological indicator information of the second time length is obtained, normalized, and then input into the physiological indicator prediction model; the heart rate, respiratory rate, body temperature and blood oxygen saturation of the future third time length are obtained and marked as predicted indicator information, the second time length is R2, and the third time length is R3.

8. The method for generating a multimodal physical and mental health management program based on deep learning according to claim 7, characterized in that: According to the user's premature beat reference information and prediction index information, early warning of the user's premature beats and generation of a premature beat management plan include the following sub-steps: The heart rate, respiratory rate, body temperature and blood oxygen saturation at the same time in the prediction index information are combined into a warning feature vector, recorded as M (m1, m2, m3, m4), where m1, m2, m3 and m4 represent the heart rate, respiratory rate, body temperature and blood oxygen saturation in the prediction index information in order; Based on the first premature beat physiological information, the heart rate, respiratory rate, body temperature and blood oxygen saturation of any premature beat of the user are combined into a first reference feature vector, recorded as Q (q1, q2, q3, q4), where q1, q2, q3 and q4 represent the heart rate, respiratory rate, body temperature and blood oxygen saturation of the first premature beat physiological information in order; Based on the second similar data and the second premature beat physiological information, the heart rate, respiratory rate, body temperature and blood oxygen saturation of any premature beat of the user are combined into a second reference feature vector, denoted as P (p1, p2, p3, p4), wherein p1, p2, p3 and p4 represent the heart rate, respiratory rate, body temperature and blood oxygen saturation of the second premature beat physiological information, respectively; and the second similar data of the second similar user corresponding to P (p1, p2, p3, p4) are combined into a similar feature vector, denoted as PV (v1, v2, v3, v4).

9. The method for generating a multimodal physical and mental health management program based on deep learning according to claim 8, characterized in that: According to the user's premature beat reference information and prediction index information, early warning of the user's premature beats and generating a premature beat management plan also include the following sub-steps: For any warning feature vector M (m1, m2, m3, m4), any first reference feature vector Q (q1, q2, q3, q4), any second reference feature vector P (p1, p2, p3, p4) and the corresponding similar feature vector PV (v1, v2, v3, v4) of the user; If m1 satisfies any q1∈[m1-G1, m1+G1], or satisfies any [m1-G1, m1+G1]∩[p1*v1, p1 / v1], and [(au12-au11) / (2*G1)]≥k0, then xm1=1, otherwise xm1=0; where [au11, au12]=[m1-G1, m1+G1]∩[p1*v1, p1 / v1], k0 is the set ratio threshold; If m2 satisfies any q2∈[m2-G2, m2+G2], or satisfies any [m2-G2, m2+G2]∩[p2*v2, p2 / v2], and [(au22-au21) / (2*G2)]≥k0, then xm2=1, otherwise xm2=0; where [au21, au22]=[m2-G2, m2+G2]∩[p2*v2, p2 / v2]; If m3 satisfies any q3∈[m3-G3, m3+G3], or satisfies any [m3-G3, m3+G3]∩[p3*v3, p3 / v3], and [(au32-au31) / (2*G3)]≥k0, then xm3=1, otherwise xm3=0; where [au31, au32]=[m3-G3, m3+G3]∩[p3*v3, p3 / v3]; If m4 satisfies any q4∈[m4-G4, m4+G4], or satisfies any [m4-G4, m4+G4]∩[p4*v4, p4 / v4], and [(au42-au41) / (2*G4)]≥k0, then xm1=4, otherwise xm4=0; where [au41, au42]=[m4-G4, m4+G4]∩[p4*v4, p4 / v4]; If (xm1+xm2+xm3+xm4)=3 corresponding to any warning feature vector M (m1, m2, m3, m4) of the user, then the time corresponding to the warning feature vector is obtained, premature heart beat prediction information is sent to the user, and the premature heart beat coping method is obtained from the database and sent to the user; If (xm1+xm2+xm3+xm4)=4 corresponding to any warning feature vector M (m1, m2, m3, m4) of the user, the time corresponding to the warning feature vector is obtained, a premature heart beat warning message is issued to the user, and a method for dealing with premature heart beats is obtained from the database and sent to the user.

10. A multimodal physical and mental health management program generation system based on deep learning, used to implement the multimodal physical and mental health management program generation method based on deep learning according to any one of claims 1 to 9, characterized in that: It includes data collection module, index processing module, information prediction module and premature beat management module; The data collection module includes a collection unit and a preliminary screening unit. The collection unit collects basic feature information and physiological index information of the user based on the wearable device, and the preliminary screening unit marks the first similar user according to the basic feature information of the user; The index processing module includes two screening units and a processing unit. The two screening units screen the second similar user according to the physiological index information of the user. The processing unit establishes the user's premature beat reference information according to the physiological index information of the user and the second similar user when premature beats occur. The information prediction module includes a model unit and a prediction unit. The model unit establishes a physiological index prediction model according to the user's physiological index information. The prediction unit is used to predict the physiological index and obtain the prediction index information. The premature beat management module warns the user of premature beats based on the user's premature beat reference information and prediction index information, and generates a premature beat management plan.

Images