An Insulin Dose Fuzzy Control Method Integrating Emotional and Motion Information

Abstract

The present invention discloses a fuzzy control method for insulin dosage integrating emotional and motion information, belonging to the technical field of diabetes insulin treatment, which comprises the following steps: S1, estimating the emotional state and exercise intensity; S2, calculating the current insulin dosage; S3, determining the structure of the fuzzy controller, defining the input-output fuzzy distribution with the emotional state and exercise intensity information as inputs and the insulin dosage adjustment step as the output, and calculating the membership degrees of the inputs with respect to different fuzzy subsets; S4, establishing fuzzy rules; S5, performing fuzzy logic reasoning to obtain the output of the fuzzy system, and multiplying the output by the insulin dosage obtained in step S2 to determine the final insulin dosage. By adopting the above-mentioned fuzzy control method for insulin dosage integrating emotional and motion information, the present invention can achieve a more personalized and dynamic insulin administration strategy, make a more reasonable decision on the insulin dosage, and improve the blood glucose management effect of patients.

Description

Technical Field

[0001] The present invention relates to the technical field of diabetes insulin treatment, and particularly relates to a fuzzy control method for insulin dosage integrating emotional and exercise information. Background Art

[0002] Insulin decision-making is a key link in diabetes blood glucose management, which involves a detailed assessment of the patient's condition, setting of treatment goals, selection of insulin dosage forms, adjustment of dosage, and personalized formulation of the plan. There are various insulin treatment regimens, which are selected according to the specific needs and conditions of the patient. Among them, basal insulin is suitable for type 2 diabetes patients with poor fasting blood glucose control, especially those who have not achieved the standard with oral hypoglycemic drugs. Long-acting insulin analogs such as insulin glargine or insulin degludec are often used to maintain the basal insulin level throughout the day, usually injected before bedtime. Prandial insulin: For postprandial blood glucose control, short-acting or rapid-acting insulin such as insulin aspart, insulin lispro are injected before meals to rapidly reduce the postprandial blood glucose peak.

[0003] Insulin dosage adjustment is a dynamic process that requires regular blood glucose monitoring and adjustment according to the results. The adjustment frequency is usually once every 3 - 5 days, and the adjustment amount each time is 1 - 4 units until the blood glucose reaches the target range. The adjustment basis includes blood glucose levels at different time points such as fasting blood glucose, postprandial blood glucose, bedtime blood glucose, and whether hypoglycemic events occur. If the patient's lifestyle changes, body weight fluctuates, or encounters stress situations, the dosage needs to be adjusted accordingly. In some cases, it may be necessary to switch from one insulin regimen to another to better match the patient's current blood glucose control needs.

[0004] The patient's metabolism and physical activity have a significant impact on the dynamic changes of glucose levels and insulin requirements. Many type 1 diabetes patients experience hypoglycemia during or after exercise, and physiological stress activities also increase the risk of hypoglycemia. Specifically, aerobic exercise can increase glucose uptake by 1.5 to 10 times the resting rate, depending on the duration and intensity of the exercise. Moderate-intensity exercise causes a sharp increase in skeletal muscle glucose demand, which is matched by an increase in liver glucose production. In high-intensity aerobic and anaerobic exercises, liver glucose production may exceed muscle glucose utilization, resulting in a sharp increase in the individual's glucose level. In addition, psychological and emotional changes can activate the hypothalamic-pituitary-adrenal axis and the sympathetic nervous system, leading to the release of stress hormones such as adrenaline, glucagon, and growth hormone, which affect blood glucose metabolism.

[0005] In the daily life of diabetes patients, the randomness of lifestyle leads to the coupling and variability of the above factors' impact on blood glucose metabolism, making it difficult to accurately capture the changes in insulin requirements. Therefore, studying the optimal decision-making of insulin dosage integrating emotional and exercise information is of great significance for improving the patient's blood glucose management performance.

[0006] According to the investigation and understanding, the currently disclosed prior art is as follows:

[0007] The invention patent with the publication number of CN201980042748.9 discloses a system for providing dose recommendations for basal insulin titration. This method adjusts the basal dose by using individualized control goals, historical blood glucose monitoring, and insulin injection information. The invention patent with the publication number of CN201910310314.4 discloses a method and system for insulin formulation selection and dose adjustment, providing an insulin dose adjustment model based on the medical records of type 2 diabetes patients and combined with the Xgboost machine learning model. The invention patent with the publication number of CN109564775A discloses a system and method for optimizing mealtime insulin doses, providing a system and method for adjusting the short-acting dose of the subject's expected meal by using the insulin infusion data recorded in the insulin pen.

[0008] The invention patent with the publication number of CN202010930702.5 discloses an individualized decision-making system for pre-meal insulin doses based on Gaussian processes. This system uses historical blood glucose management information to establish a post-meal blood glucose prediction model, designs a risk-sensitive control framework, and completes the decision-making of pre-meal insulin doses. The invention patent with the publication number of CN202110117772.3 discloses an expert experience-assisted learning and optimization decision-making system for pre-meal insulin doses on the basis of the above patent, which can evaluate the credibility of Gaussian process regression for predicting post-meal blood glucose under few samples and decide whether to assist with expert decision-making. However, none of the above patents consider how to effectively integrate emotional and exercise state information to improve the accuracy of insulin dose decision-making, thereby improving blood glucose management for diabetic patients. Summary of the Invention

[0009] The object of the present invention is to provide a fuzzy control method for insulin doses that integrates emotional and exercise information, which can achieve a more personalized and dynamic insulin administration strategy, make a more reasonable decision on insulin doses, and improve the blood glucose management effect of patients.

[0010] To achieve the above object, the present invention provides a fuzzy control method for insulin doses that integrates emotional and exercise information, including the following steps:

[0011] S1. Estimate the emotional state and exercise intensity through the photoplethysmogram signal and triaxial acceleration signal collected by the bracelet sensor;

[0012] S2. Calculate the current insulin dose;

[0013] S3, determining the structure of the fuzzy controller, defining the input-output fuzzy distribution with the emotional state and exercise intensity information as input and the insulin dosage adjustment step as output, and calculating the membership degree of the input relative to different fuzzy subsets;

[0014] S4, establish fuzzy rules;

[0015] S5. Perform fuzzy logic reasoning to obtain the output of the fuzzy system, and multiply the output by the insulin dose obtained in step S2 to determine the final insulin dose.

[0016] Preferably, step S1 is divided into two parts, using photoplethysmography signals to evaluate emotional state and using triaxial acceleration to evaluate motion state;

[0017] S101, Photoplethysmography signals reflect the beating of the heart by detecting small changes in the volume of blood in the skin;

[0018] Specifically: by analyzing the characteristics of the pulse wave signal, mainly the time domain characteristics, frequency domain characteristics and nonlinear characteristics, the key features for emotion classification are formed; these key features are modeled using the support vector machine algorithm to achieve emotion classification, mainly including stress state and non-stress state. The Gaussian kernel function is selected as the kernel function of the support vector machine. The Gaussian kernel function maps the low-dimensional input space to the high-dimensional feature space, overcoming the limitations of the linear model. The filtered features are input into the classifier model to obtain the estimated emotional state of the human body;

[0019] S102, evaluating the motion state specifically includes:

[0020] First, the three-axis acceleration data is preprocessed;

[0021] Secondly, the data needs to be organized into time windows of fixed length. The data points in each window constitute a sample for training and testing the CNN model. In the design of the CNN model, a network architecture including a one-dimensional convolutional layer, a pooling layer, and a fully connected layer is designed.

[0022] Finally, the CNN model is used to classify exercise intensity, including running, walking, and resting.

[0023] Preferably, step S2 specifically comprises: using historical blood glucose concentrations, meal information and insulin dosages, combined with current blood glucose monitoring and dietary intake information, and calculating the current insulin dosage based on the doctor's clinical strategy or decision algorithm.

[0024] Preferably, the main components of the fuzzy controller in step S3 are described as follows:

[0025] The input is historical emotional state and exercise intensity information. A fuzzification operation is performed on the input. Eight moments of historical emotional state and exercise intensity with an evaluation period \(T = 15\) min are selected. That is, only the emotional state and exercise intensity within 2 hours of the current insulin decision may affect the current insulin dose decision. The historical emotional state \(x1\) and exercise intensity sequence \(x2\) are represented by the following formula:

[0026] x1 = [m0,m1,m2,m3,m4,m5,m6,m7]

[0027] x2 = [n0,n1,n2,n3,n4,n5,n6,n7]

[0028] where, m i represents the emotional state of the \(i\)-th cycle from the current time. If it is a stress state, the value is 1; if it is a non-stress state, the value is 0; n i represents the exercise intensity of the \(i\)-th cycle from the current time. If it is a running state, the value is 2; if it is a walking state, the value is 1; if it is a resting state, the value is 0;

[0029] Set the corresponding weight matrix for the eight historical emotional states and exercise intensities as:

[0030] w = [w0,w1,w2,w3,w4,w5,w6,w7]

[0031] where, w0 and w7 are 1, w1 and w6 are 2, w2 and w5 are 3, w3 and w4 are 4;

[0032] Define the contribution value of the historical emotional state to the current blood glucose as h1 = 5x1*w T , and the contribution value of the historical exercise intensity to the current blood glucose as h2 = 2.5x2*w T .

[0033] Preferably, in step S3, the contribution values are fuzzified. The fuzzy sets are respectively defined as {many stress situations, more stress situations, medium stress situations, fewer stress situations, few stress situations} and {high exercise intensity, relatively high exercise intensity, medium exercise intensity, relatively low exercise intensity, low exercise intensity}, and a triangular membership function is used, with the input being the contribution value;

[0034] The output of the fuzzy controller is the dose adjustment percentage, which is used to adjust the insulin dose calculated in step S2; the percentage is controlled within the range of [0, 150]. The fuzzy set of the insulin dose requirement change considering emotional and exercise information is defined as {much insulin dose, more insulin dose, medium insulin dose, less insulin dose, little insulin dose}, and a triangular membership function is used, with the input being the normalized dose adjustment percentage.

[0035] Preferably, the fuzzy rules in step S4 are as follows:

[0036] S401. The more stressed in the fuzzy subset with the largest membership degree confirmed according to the historical emotional state contribution value, the greater the impact of the emotional state on blood glucose (the faster the blood glucose drops), and less insulin dosage is required;

[0037] S402. The more intense in the fuzzy subset with the largest membership degree confirmed according to the historical exercise intensity contribution value, the greater the impact of the exercise intensity on blood glucose (the faster the blood glucose drops), and less insulin dosage is required.

[0038] Preferably, in step S5, Mamdani method is adopted for fuzzy logic inference. Based on the fuzzy rules in step S4, the fuzzy results output by each rule are aggregated using min-max inference to obtain the final fuzzy set, and then defuzzification is performed to obtain the final output.

[0039] Preferably, the defuzzification is performed by the maximum membership degree method. Denote the normalized insulin dosage adjustment percentage calculated by the maximum membership degree method as y max , then the finally injected insulin dosage is:

[0040]

[0041] wherein, is the insulin dosage calculated in step S2, and λ = 0.67 is the denormalization coefficient.

[0042] Therefore, by adopting the above-mentioned insulin dosage fuzzy control method integrating emotional and exercise information, the present invention can realize a more personalized and dynamic insulin administration strategy, make a more reasonable decision on the insulin dosage, and improve the blood glucose management effect of patients.

[0043] Next, through the drawings and embodiments, the technical solutions of the present invention will be further described in detail. Description of the Drawings

[0044] Figure 1 is the method framework diagram of an embodiment of the insulin dosage fuzzy control method integrating emotional and exercise information of the present invention;

[0045] Figure 2 is the clinical data verification schematic diagram of a certain type 1 diabetic patient provided by the present invention. Detailed Embodiments

[0046] The following further illustrates the technical solutions of the present invention through the drawings and embodiments.

[0047] Unless otherwise defined, the technical terms or scientific terms used in this invention shall have the ordinary meanings as understood by those of ordinary skill in the field to which this invention pertains.

[0048] Example 1

[0049] As Figure 1 shown, the present invention provides an insulin dosage fuzzy control method that integrates emotion and motion information, including an emotion and motion state evaluation module, an insulin dosage decision module, and a decision dosage adjustment module based on fuzzy control. At the same time, the parameters in the present invention are determined based on the clinical experience of doctors and the clinical data of diabetic patients using multiple daily insulin injection therapy in the hospital.

[0050] The emotion and motion state evaluation module mainly estimates the emotion state and motion intensity by using the photoplethysmogram signal and the three-axis acceleration signal collected in the bracelet sensor respectively.

[0051] For emotion state evaluation, first, by analyzing the characteristics of the pulse wave signal, mainly time-domain characteristics, frequency-domain characteristics, and non-linear characteristics, key characteristics for emotion classification are formed. In this embodiment, a total of 32 characteristic forms are selected. For a detailed description, reference can be made to the invention patent with the publication number of CN202410448222.3. Then, these characteristics are modeled using the support vector machine algorithm to achieve emotion classification, mainly including the stress state and the non-stress state. The Gaussian kernel function is selected as the kernel function of the support vector machine. The Gaussian kernel function maps the low-dimensional input space to a high-dimensional feature space, overcoming the limitations of the linear model. Inputting the screened characteristics into the trained classifier model can obtain the estimated human emotion state.

[0052] For motion intensity evaluation, first, data preprocessing is performed on the three-axis acceleration data. The original acceleration data usually contains noise, and high-frequency interference needs to be removed through low-pass filtering. Secondly, the data needs to be organized into time windows of a fixed length. The data points within each window form a sample for training and testing the CNN model. In this embodiment, a time window with a length of 4 seconds is selected, and operations such as resampling and normalization are performed on the data within the window. In the design of the CNN model, considering the spatio-temporal characteristics of the acceleration data, a network architecture including a one-dimensional convolutional layer, a pooling layer, and a fully connected layer is designed, and cross-validation is used to evaluate the generalization ability of the model. In this embodiment, the cross-entropy function is used as the loss function and the Adam optimizer is used as the gradient optimizer. Finally, the CNN model is used to achieve motion intensity classification, mainly including running, walking, and resting.

[0053] The insulin dose decision-making module utilizes historical blood glucose concentrations, meal information, and insulin doses, combines the current blood glucose monitoring and meal information, and calculates the current insulin dose based on the doctor's clinical strategy or decision-making algorithm. This embodiment uses a data-driven decision-making algorithm to complete insulin dose decision-making. For details, refer to the invention patent with the patent number CN202110117772.3.

[0054] The decision dose adjustment module based on fuzzy control mainly completes the design of the fuzzy controller. Its input is the historical emotional state and exercise intensity information, and the output is the percentage of insulin dose adjustment. First, select the historical emotional state and exercise intensity at 8 moments with an evaluation period T = 15 min, that is, only the emotional state and exercise intensity within 2 h of the current insulin decision may affect this insulin dose decision. The historical emotional state x1 and exercise intensity sequence x2 can be expressed by the following formula:

[0055] x1 = [m0, m1, m2, m3, m4, m5, m6, m7],

[0056] x2 = [n0, n1, n2, n3, n4, n5, n6, n7],

[0057] where m i represents the emotional state of the i-th cycle from the current time. If it is a stress state, the value is 1; if it is a non-stress state, the value is 0. n i represents the exercise intensity of the i-th cycle from the current time. If it is a running state, the value is 2; if it is a walking state, the value is 1; if it is a resting state, the value is 0.

[0058] The effects of emotional and exercise states on blood glucose are not immediately visible and there is a significant delay effect. This delay effect reflects the complexity and long-term response characteristics of the human body's physiological regulation mechanism. When an individual experiences emotional stress, the body immediately releases stress hormones such as cortisol and adrenaline, and there is an obvious lag in the effect of these hormones on blood glucose in the body. For current moderate to high-intensity exercise, since it takes a corresponding delay time to act on the blood glucose system, a decrease in blood glucose level cannot be immediately observed. Therefore, for historical emotional states and exercise intensities, the effects on blood glucose level are not positively correlated. According to the empirical formula, set the corresponding weight matrix for 8 historical emotional states and exercise intensities as:

[0059] w = [w0, w1, w2, w3, w4, w5, w6, w7],

[0060] where w0 and w7 are 1, w1 and w6 are 2, w2 and w5 are 3, and w3 and w4 are 4.

[0061] Furthermore, the contribution of historical emotional state to current blood sugar is defined as h1=5x1*w T , the contribution of historical exercise intensity to current blood sugar is h2=2.5x2*w T .

[0062] The influence contribution value is fuzzified as follows. For the emotional state and exercise state, the fuzzy sets are defined as {more stress, more stress, moderate stress, less stress, less stress} and {high exercise intensity, higher exercise intensity, moderate exercise intensity, lower exercise intensity, low exercise intensity}, respectively. The triangular membership function is used to determine the membership of the influence contribution value in each fuzzy set. The specific form is as follows:

[0063] The emotional state membership function is:

[0064] Less stressful situations:

[0065]

[0066] Less stressful situations:

[0067]

[0068] Moderate stress:

[0069]

[0070] More stressful situations:

[0071]

[0072] Many stressful situations:

[0073]

[0074] Where x is the contribution value of the historical emotional state to the current blood sugar, and y1, y2, y3, y4, and y5 are the membership degrees in different fuzzy subsets (few stressful situations, few stressful situations, moderate stressful situations, more stressful situations, and more stressful situations) under the corresponding contribution values of the historical emotional state to the current blood sugar.

[0075] The exercise intensity membership function is:

[0076] Low intensity exercise:

[0077]

[0078] Lower intensity exercise:

[0079]

[0080] Moderate exercise intensity:

[0081]

[0082] Higher exercise intensity:

[0083]

[0084] High exercise intensity:

[0085]

[0086] Where x is the contribution value of historical exercise intensity to the current blood glucose, and y1, y2, y3, y4, y5 are the membership degrees in different fuzzy subsets (low exercise intensity, relatively low exercise intensity, moderate exercise intensity, higher exercise intensity, high exercise intensity) respectively under the condition of the contribution value of historical exercise intensity to the current blood glucose.

[0087] In this embodiment, the output percentage is controlled within the range of [0, 150], and the corresponding fuzzy set is defined as {a lot of insulin dosage, relatively more insulin dosage, moderate insulin dosage, less insulin dosage, little insulin dosage}, and a triangular membership function is used, specifically as follows:

[0088] Little insulin dosage:

[0089]

[0090] Relatively less insulin dosage:

[0091]

[0092] Moderate insulin dosage:

[0093]

[0094] Relatively more insulin dosage:

[0095]

[0096] A lot of insulin dosage:

[0097]

[0098] Where x is the normalized dosage adjustment percentage, and y1, y2, y3, y4, y5 are the membership degrees of the model output results in different fuzzy subsets (little insulin dosage, relatively less insulin dosage, moderate insulin dosage, relatively more insulin dosage, a lot of insulin dosage) respectively.

[0099] According to the input and output of the above fuzzy controller and combined with the empirical knowledge in the field of doctors' blood glucose regulation, the following fuzzy rules are set:

[0100] 1) The more stress in the fuzzy subset with the largest membership degree confirmed according to the historical emotional state contribution value, the greater the impact of the emotional state on blood glucose (the faster the blood glucose drops), and less insulin dosage is required;

[0101] 2) The more intense in the fuzzy subset with the largest membership degree confirmed according to the historical exercise intensity contribution value, the greater the impact of the exercise intensity on blood glucose (the faster the blood glucose drops), and less insulin dosage is required.

[0102] The rule base is a collection of all designed fuzzy rules, which represents the complete knowledge base of the control strategy. According to the above fuzzy rules, for the insulin dosage decision-making of the present invention, a fuzzy control rule base as shown in Table 1 is set up by combining the historical emotional state and exercise intensity.

[0103] Table 1 Fuzzy control system rule table

[0104]

[0105] Fuzzy inference is based on the current input fuzzy set and the rule base to perform fuzzy logic inference. The commonly used inference methods mainly include the Mamdani method and the Tsukamoto method. The Mamdani method combines the outputs of all rules by "maximum membership degree", that is, for each output dimension, the maximum value of the membership degrees corresponding to all rules is taken. The Tsukamoto method is more inclined to mathematical operations and directly operates on the membership functions. The result of this stage is a composite fuzzy set of one or more output variables, representing the comprehensive influence of all activated rules.

[0106] In this embodiment, the Mamdani method is adopted. After knowing the membership degrees of the current historical emotional state and exercise intensity to each fuzzy subset, the fuzzy results output by each rule are aggregated based on the minimum-maximum (Min-Max) inference using the rule base in Table 1 to obtain the membership function of the final output, and the specific form is as follows:

[0107]

[0108] where i = 1, 2,..., N; N is the total number of fuzzy rules in Table 1; is the membership degree that the input current historical emotional state belongs to the emotional state fuzzy subset A in the i-th rule i ; is the membership degree that the input current historical exercise intensity belongs to the exercise intensity fuzzy subset B in the i-th rule i ; μ Ci (x) is the membership degree that the corresponding output insulin dosage relative percentage belongs to the insulin dosage adjustment fuzzy subset C i ;

[0109] Then defuzzification is performed to obtain the final output. In this embodiment, the maximum membership degree method is used for defuzzification, that is, the central value of the fuzzy set with the largest membership degree is selected as the final output, which is the normalized percentage of insulin dose adjustment, denoted as y max , then the final insulin dose injected by the patient is

[0110]

[0111] where is the insulin dose recommended by the insulin dose decision module, and λ = 0.67 is the de-normalization coefficient.

[0112] Finally, using the clinical data of a type 1 diabetic patient treated with MDI in the hospital, the insulin dose decision scenario integrating emotion and motion information and how this embodiment improves the decision performance are elaborated. The specific steps for carrying out retrospective insulin dose decision can be found in the invention patent with the patent number CN202110117772.3

[0113] In addition, during the treatment process of this embodiment, the patient's historical photoplethysmogram signals and triaxial acceleration signals will be synchronously collected through a bracelet, and the insulin dose decision will be completed using the method provided in this embodiment. At the same time, a comparison will be made with the decision method that does not consider emotion and motion information (that is, the method provided in the invention patent with the patent number CN202110117772.3), and the comparison results are as Figure 2 shown.

[0114] From Figure 2 it can be seen that when the patient has no movement and little emotional stress, compared with the decision method that does not consider emotion and motion information, this embodiment will recommend a higher insulin dose according to the current blood glucose level, which can effectively reduce postprandial hyperglycemia.

[0115] When the patient has a high exercise intensity and a lot of emotional stress, compared with the decision method that does not consider emotion and motion information, this embodiment will recommend a smaller insulin dose, which helps to reduce subsequent hypoglycemic events. According to Figure 2 it can be seen that after effectively evaluating the patient's exercise intensity and emotional state, this embodiment will make a more reasonable decision on the insulin dose and improve the patient's blood glucose management effect.

[0116] It should be noted that the content not elaborated in detail in this invention is all prior art and is well-known to those skilled in the art.

[0117] Therefore, by adopting the above-mentioned insulin dose fuzzy control method integrating emotion and motion information, this invention can achieve a more personalized and dynamic insulin administration strategy, make a more reasonable decision on the insulin dose, and improve the patient's blood glucose management effect.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions of the present invention or make equivalent replacements, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An insulin dose fuzzy control method that integrates emotional and motion information, characterized in that It includes the following steps: S1. Estimate the emotional state and exercise intensity based on the photoplethysmogram (PPG) signal and triaxial acceleration signal collected by the bracelet sensor; S2. Calculate the current insulin dose; S3. Determine the structure of the fuzzy controller, define the input-output fuzzy distribution with the emotional state and exercise intensity information as the input and the insulin dose adjustment step size as the output, and calculate the membership degree of the input with respect to different fuzzy subsets; The main components of the fuzzy controller are described as follows: The input is historical emotional state and exercise intensity information. A fuzzification operation is performed on the input, and the historical emotional state and exercise intensity at 8 moments with an evaluation period T = 15 min are selected, and the historical emotional state and exercise intensity sequence are represented by the following formula: ； ； Among them, represents the emotional state in the th cycle from the current time. If it is a stress state, the value is 1; if it is a non-stress state, the value is 0. represents the exercise intensity in the th cycle from the current time. If it is a running state, the value is 2; if it is a walking state, the value is 1; if it is a resting state, the value is 0. Set the corresponding weight matrices for 8 historical emotional states and exercise intensities as: ； Among them, and is 1, and is 2, and is 3, and is 4; Define the contribution value of the historical emotional state to the current blood glucose as , and the contribution value of the historical exercise intensity to the current blood glucose as ; Perform fuzzy processing on the influence contribution value. The fuzzy sets are respectively defined as {many stress situations, more stress situations, medium stress situations, fewer stress situations, few stress situations} and {high exercise intensity, relatively high exercise intensity, medium exercise intensity, relatively low exercise intensity, low exercise intensity}, and use triangular membership functions with the input being the influence contribution value; The output of the fuzzy controller is the dose adjustment percentage, which is used to adjust the insulin dose calculated in step S2; the percentage is controlled within the range of [0, 150]. The fuzzy set of the insulin dose demand change considering emotional and exercise information is defined as {much insulin dose, more insulin dose, medium insulin dose, less insulin dose, little insulin dose}, and use triangular membership functions with the input being the normalized dose adjustment percentage; S4. Establish fuzzy rules; the fuzzy rules are as follows: S401. The more stressed in the fuzzy subset with the largest membership degree confirmed according to the historical emotional state contribution value, the greater the impact of the emotional state on blood glucose, and less insulin dose is required; S402. The more intense in the fuzzy subset with the largest membership degree confirmed according to the historical exercise intensity contribution value, the greater the impact of the exercise intensity on blood glucose, and less insulin dose is required; S5. Conduct fuzzy logic reasoning to obtain the output of the fuzzy system, and multiply the output by the insulin dose obtained in step S2 to determine the final insulin dose.

2. The insulin dose fuzzy control method integrating emotional and motion information according to claim 1, characterized in that, Step S1 is divided into two parts, using the PPG signal for emotional state assessment and using triaxial acceleration for exercise state assessment; S101. The PPG signal reflects the heart beat by detecting minute changes in skin blood volume; Specifically: by analyzing the characteristics of the pulse wave signal, key features for emotion classification are formed; these key features are modeled using the support vector machine algorithm to achieve emotion classification. Select the Gaussian kernel function as the kernel function of the support vector machine, and input the screened features into the classifier model to obtain the estimated emotional state of the human body; S102. The specific method for assessing the exercise state is as follows: First, perform data preprocessing on the triaxial acceleration data; Secondly, the data needs to be organized into time windows of fixed length, and the data points within each window form a sample for training and testing the CNN model; in the design of the CNN model, a network architecture including a one-dimensional convolutional layer, a pooling layer, and a fully connected layer is designed; Finally, use the CNN model to achieve exercise intensity classification, including three categories: running, walking, and resting.

3. The insulin dose fuzzy control method integrating emotional and motion information according to claim 2, characterized in that, Step S2 specifically is: Using historical blood glucose concentration, meal information, and insulin dosage, combining with current blood glucose monitoring and dietary intake information, and calculating the current insulin dosage based on the doctor's clinical strategy or decision-making algorithm.

4. A fuzzy control method for insulin dosage integrating emotional and motion information according to claim 3, characterized in that, In step S5, the fuzzy logic inference adopts the Mamdani method, aggregates the fuzzy results output by each rule based on the minimum-maximum inference using the fuzzy rules in step S4 to obtain the final fuzzy set, and then performs defuzzification to obtain the final output.

5. A fuzzy control method for insulin dosage integrating emotional and motion information according to claim 4, characterized in that Defuzzification is carried out by the maximum membership degree method. Denote the normalized insulin dose adjustment percentage calculated by the maximum membership degree method as , then the final insulin dose to be injected is: ； Among them, is the insulin dose calculated in step S2, is the de-normalized coefficient.

Images