A rapid stroke identification method combining TOPSIS and artificial intelligence

Abstract

The present invention relates to a method for rapid stroke identification combining TOPSIS and artificial intelligence. First, the existence of unilateral muscle strength decline is judged by the degree of horizontal raising of the user's arms, and the muscle strength difference is graded by the angle between the arms; then the degree of slurred speech of the user is analyzed by artificial intelligence speech recognition and combined with the error word rate calculation method; then the position coordinates of the facial feature points are detected by the artificial intelligence facial feature point recognition model, the degree of facial droop is calculated by the TOPSIS method, and finally the risk degree of stroke is rapidly identified by the TOPSIS method by comprehensively grading the muscle strength difference, the degree of slurred speech and the degree of facial droop. The present invention can enable patients and their families to effectively and quickly identify stroke symptoms at home, and automatically shunt stroke patients to corresponding stroke centers for treatment according to different severity of stroke, thereby improving the emergency rescue efficiency and quality of emergency medical rescue services.

Description

Technical Field

[0001] The present invention relates to a method for rapid stroke recognition on a mobile terminal, and in particular to a method for rapid stroke recognition combining TOPSIS and artificial intelligence. Background Art

[0002] According to the "China Stroke Prevention and Treatment Report", stroke is the leading cause of life expectancy loss in my country. Relevant data show that among the people who die from illness in my country, at least 1 in every 5 people dies from stroke. Stroke is not only the leading cause of death among Chinese adults, but also the leading cause of disability, which brings heavy blows and burdens to individuals, families and society. However, there are still two bottlenecks in the treatment of stroke that need to be solved urgently: difficulty in pre-hospital identification and long pre-hospital transfer time. These two problems are related to the pre-hospital emergency time (OTT) of acute ischemic stroke. Studies have shown that every 1 minute reduction in OTT can increase the average healthy life time by 1.8 days, and every 15 minutes reduction can increase the healthy life time by an additional month, and reduce the in-hospital mortality rate by 4%. Timely identification and transfer of patients to stroke centers with stroke treatment capabilities can increase the intravenous thrombolysis rate and increase the proportion of patients successfully treated. Rapid identification of stroke is the first step in starting the stroke emergency survival chain, which affects the emergency efficiency and quality of emergency medical rescue services. Studies have shown that the public's correct recognition rate of stroke is only 3-16%. At the same time, the entry threshold for pre-hospital emergency personnel in my country's medical emergency centers is low, and professional training in stroke is insufficient. Only about 33% of stroke patients can be correctly identified by pre-hospital emergency personnel. Therefore, improving the self-identification rate of stroke patients will inevitably improve the efficiency of pre-hospital emergency treatment of stroke, reduce pre-hospital delays, and increase the proportion of ischemic stroke patients receiving intravenous thrombolytic therapy.

[0003] One of the main reasons for the delay of stroke patients in my country is that patients and their families lack effective recognition of stroke symptoms and fail to call the emergency number "120" in time. At present, the commonly used tools for self-identification of stroke in the world are face, arm, speech, time (FAST) and Cincinnati prehospital stroke scale (CPSS). Chinese researchers have developed a stroke identification method that is easy for the public to understand, "Stroke 120" (see an asymmetrical face, check whether the two arms are unilaterally weak, and listen to whether the speech is clear) based on Chinese cultural characteristics. In order to further improve the pre-hospital stroke identification rate, an automatic stroke identification system based on artificial intelligence algorithms has been developed internationally, which has improved the ability of automatic classification and identification of stroke and automatically diverted stroke patients to the corresponding stroke center for treatment according to the severity of stroke. However, most of the existing machine learning-based work requires a large number of cases to improve the accuracy of the model, and the model based on existing cases may have overfitting problems, resulting in the generated machine learning model being unsuitable for rapid identification of new stroke risks. The present invention adopts a method combining TOPSIS with artificial intelligence to quickly identify stroke and issue warnings based on the FAST criteria and the arm raising data, voice data and facial data that can be collected by the smart terminal. On the one hand, it reduces the training samples of the model and only needs to extract typical stroke risk level cases as references. On the other hand, it avoids relying on highly subjective manual experience for stroke identification, thereby improving the accuracy and reliability of stroke identification. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects of the prior art and provide a method for rapid stroke identification combining TOPSIS and artificial intelligence, so that patients and their families can quickly identify stroke risks at home through mobile terminals, thereby improving the first aid efficiency and quality of emergency medical rescue services.

[0005] The technical solution to achieve the purpose of the present invention is: a method for rapid stroke identification combining TOPSIS and artificial intelligence, which collects the patient's muscle strength, voice and current facial photo on a mobile terminal, analyzes the patient's muscle strength grade, slurred speech and facial droop, and identifies the current stroke risk level, specifically including the following steps:

[0006] Step 1: Determine whether there is unilateral muscle strength loss by the degree to which the user raises both arms horizontally, and grade the muscle strength difference by the angle between the arms;

[0007] Step 2: Analyze the user's articulation level through artificial intelligence speech recognition and combined with the error word rate calculation method;

[0008] Step 3: Detect the position coordinates of facial feature points based on the artificial intelligence facial feature point recognition model, and calculate the degree of facial sagging by the TOPSIS method;

[0009] Step 4: Use the TOPSIS method to comprehensively consider the muscle strength difference grade, slurred speech level, and facial droop to quickly identify the risk of stroke.

[0010] In step 1 described in the above technical solution, by raising both arms horizontally, observing whether both arms can be raised to the same height and whether one arm is weak or drooping, it is determined whether the unilateral muscle strength has decreased, and the degree of the bilateral muscle strength difference can be determined by comparing the angles of the two arms.

[0011] The specific steps of the above technical solution are as follows:

[0012] (1) The patient stands with both hands raised to a height that can be reached;

[0013] (2) Compare the angle differences between the two arms and divide them into three levels: 0° to 30° as level I, 30° to 60° as level II, and 60° to 90° as level III;

[0014] (3) The user inputs the level into the mobile terminal.

[0015] In step 2 of the above technical solution, the patient is asked to read aloud the given text, and the text is recognized as corresponding text through a speech recognition interface, and then the error word rate is calculated to quantify the degree of slurred speech of the patient.

[0016] The specific steps of the above technical solution are as follows:

[0017] 1) The mobile terminal randomly selects a paragraph of text with 15-30 words from the sentence library and prompts the user to read it aloud;

[0018] 2) The mobile terminal converts the read voice into text through the voice recognition interface;

[0019] 3) Calculate the word error rate R and quantify the level of inarticulation according to the word error rate, where 0 to 25% is level I, 25% to 50% is level II, 50% to 75% is level III, and more than 75% is level IV.

[0020] The error rate R is calculated by formula (1):

[0021]

[0022] Among them, I is the number of characters added to the converted text relative to the original text, D is the number of characters reduced from the converted text relative to the original text, and C is the number of characters corresponding to the original text.

[0023] In step 3 of the above technical solution: first, facial features are identified by a facial feature point recognition method, including the left corner of the left eye E1, the right corner of the left eye E2, the left corner of the right eye E3, the right corner of the right eye E4, the root of the nose N, the left corner of the mouth M1, the right corner of the mouth M2, and the lower lip feature point M3; then the inclination between the feature points is calculated, and the absolute value of the difference in inclination of the symmetrical feature points is used as an indicator, and the degree of sagging of a typical face is used as a reference to calculate the degree of sagging of the user's current face.

[0024] The specific steps of the above technical solution are as follows:

[0025] a. Calculate the absolute value g1 of the inclination difference between the facial feature points E1-N and E4-N of the user's current facial sagging level I, II, III, and IV, respectively, as shown in formula (2):

[0026]

[0027] Among them, y represents the ordinate of the feature point, x represents the abscissa of the feature point, and abs represents the absolute value.

[0028] b. Calculate the absolute value g2 of the inclination difference between the facial feature points E2-N and E3-N of the user's current facial sagging level I, II, III, and IV, respectively, as shown in formula (3):

[0029]

[0030] c. Calculate the absolute value g3 of the inclination difference between the facial feature points M1-M3 and M2-M3 of the facial sagging degree I, II, III, and IV respectively, as shown in formula (4):

[0031]

[0032] d. The absolute values ​​of the inclination differences of the facial feature points of the user and the facial sagging degrees I, II, III, and IV calculated in steps a, b, and c are combined into a 5*3 matrix G, where each row is the inclination difference g1, g2, and g3 of the facial feature points corresponding to the facial sagging degree, as shown in the following matrix:

[0033]

[0034] e. Normalize G to get the matrix R = [r ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and the specific normalization method is shown in formula (5):

[0035]

[0036] f. Given a weight matrix Calculate the weight normalization matrix T = R·W1, T = [z ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and T is the dot product of R and W1.

[0037] g. The smaller the absolute value of the difference in the inclination of facial feature points, the more symmetrical the face is, so the ideal solution matrix is and the negative ideal solution matrix It can be calculated by formula (6) and (7) respectively:

[0038]

[0039] h. Use Euclidean distance as a measure to calculate any feasible solution z ij To the ideal solution With negative ideal solution Distance as well as They can be calculated according to formulas (8) and (9) respectively:

[0040]

[0041] i. Calculate the relative closeness of each solution The current degree of facial sagging is classified as C i The closest degree of facial sagging.

[0042] The specific details of step 4 in the above technical solution are as follows:

[0043] A. The stroke risk levels I, II, III, and IV calculated in steps 1, 2, and 3 as well as the muscle strength difference level, slurred speech level, and facial droop level of the current user are combined into a 5*3 matrix H, where the first four rows are the muscle strength difference level, slurred speech level, and facial droop level corresponding to different stroke risk levels, respectively. The fifth row is the muscle strength difference level h1, slurred speech level h2, and facial droop level h3 of the current user. The value range of the muscle strength difference level is [1, 2, 3], representing the risk level I, II, and III, respectively. The value range of the slurred speech level is [1, 2, 3, 4], representing the risk level I, II, III, and IV, respectively. The value range of the facial droop level is [1, 2, 3, 4], representing the risk level I, II, III, and IV, respectively, as shown in the following matrix:

[0044]

[0045] B. Normalize the matrix H to get the matrix R′=[r ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and the specific normalization method is shown in formula (5):

[0046]

[0047] C. The weights of poor muscle strength, slurred speech, and facial droop are 5:3:2 for stroke risk, so the weight matrix is Calculate the weight normalization matrix T′=R′·W2, T′=[z ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and T′ is the dot product of R and W2;

[0048] D. The smaller the absolute value of the difference in the inclination of facial feature points, the more symmetrical the face is, so the ideal solution matrix is and the negative ideal solution matrix Calculate using formula (6) and (7) respectively

[0049]

[0050] E. Use Euclidean distance as a measure to calculate any feasible solution z ij To the ideal solution With negative ideal solution Distance as well as They can be calculated according to formulas (8) and (9) respectively:

[0051]

[0052] F. Calculate the relative closeness of each solution Classify the current stroke risk level as O i The closest stroke risk level.

[0053] After adopting the above technical solution, the present invention has the following positive effects:

[0054] (1) When the present invention is in use, if the identified stroke risk level reaches the standard for medical treatment, the intelligent terminal automatically dials the 120 emergency number, thereby improving the emergency efficiency and quality of emergency medical rescue services.

[0055] (2) The present invention is mainly used for rapid stroke identification on mobile terminals, reducing manual errors in stroke identification by non-neurologists, reducing time delays in the transfer of patients between different centers due to incorrect judgment of stroke severity, and reducing the bottleneck problem of pre-hospital emergency care for stroke, such as patients being delayed in calling for help due to the inability to automatically identify a stroke.

[0056] (3) The present invention avoids the use of traditional machine learning methods of training models with large amounts of data through the TOPSIS analysis method, which can avoid labeling large amounts of data and also avoid the problem of overfitting of machine learning models causing the model to be unsuitable for new patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments and in conjunction with the accompanying drawings, wherein

[0058] Figure 1 Shown is a framework diagram of the present invention;

[0059] Figure 2 Shown is a flow chart of the present invention;

[0060] Figure 3 The figure shows the facial droop degree determination method based on the TOPSIS method in the present invention;

[0061] Figure 4 Shown are facial feature points in the present invention. DETAILED DESCRIPTION

[0062] According to the FAST criteria, the patient's muscle strength, voice and current facial photo are collected on the mobile terminal, the patient's muscle strength grade, slurred speech and facial droop are analyzed, and the patient's current stroke risk level is determined by the TOPSIS method and typical cases. The specific steps include the following:

[0063] Step 1: Determine whether there is unilateral muscle weakness by the degree to which the user raises both arms, and grade the muscle strength difference by the angle between the arms. In the FAST standard, unilateral muscle weakness is determined by raising both arms to observe whether both arms can be raised to the same height and whether one arm is weak or drooping. Since stroke mainly causes unilateral muscle weakness, there must be one arm that can be raised. If there is a risk of stroke, it will be difficult to raise the other arm. The degree of bilateral muscle strength difference can be determined by comparing the angles of the two arms. The present invention gives operation prompts through a mobile terminal, and the specific steps are as follows:

[0064] 1. The patient stands with both hands raised to the height they can reach.

[0065] 2. Compare the angle difference between the two arms and divide it into three levels: 0° to 30° is level I, 30°

[0066] To 60° is Class II, and 60° to 90° is Class III.

[0067] 3. The user enters the level into the mobile terminal.

[0068] Step 2: Analyze the degree of slurred speech of the user through artificial intelligence speech recognition and combined with the method of calculating the word error rate. The slurred speech of patients at risk of stroke is mainly manifested in repeated pronunciation, dragging of sounds and broken words. The present invention allows the patient to read aloud the given text, and recognizes it into corresponding text through the speech recognition interface, and then calculates the word error rate to quantify the degree of slurred speech of the patient. The present invention prompts through the mobile terminal contact operation, and the specific steps are as follows:

[0069] 1. The mobile terminal randomly selects a paragraph of text with 15-30 words from the sentence library and prompts the user to read it aloud;

[0070] 2. The mobile terminal converts the spoken voice into text through the voice recognition interface;

[0071] 3. Calculate the word error rate and quantify the level of inarticulation based on the word error rate, where 0 to 25% is level I, 25% to 50% is level II, 50% to 75% is level III, and above 75% is level IV (which may exceed 100%).

[0072] The calculation method of the error word rate R is given by formula (1):

[0073]

[0074] Among them, I is the number of characters added to the converted text relative to the original text, D is the number of characters reduced from the converted text relative to the original text, and C is the number of characters corresponding to the original text.

[0075] Step 3: Detect the position coordinates of facial feature points based on the artificial intelligence facial feature point recognition model, and calculate the degree of facial sagging through the TOPSIS method. First, identify facial features through the facial feature point recognition method, including the left corner of the left eye E1, the right corner of the left eye E2, the left corner of the right eye E3, the right corner of the right eye E4, the root of the nose N, the left corner of the mouth M1, the right corner of the mouth M2, and the lower lip feature point M3. First, calculate the inclination between the feature points, take the absolute value of the difference in inclination of the symmetrical feature points as an indicator, and the degree of typical facial sagging as a solution to calculate the degree of current facial sagging of the user. The specific steps are as follows:

[0076] 1. Calculate the absolute value of the inclination difference between the facial droop levels I, II, III, and IV and the user's current facial feature points E1-N and E4-N, as shown in formula (2):

[0077]

[0078] 2. Calculate the absolute value of the difference in the inclination between the facial feature points E2-N and E3-N of the user's current facial sagging level I, II, III, and IV, respectively, as shown in formula (3):

[0079]

[0080] 3. Calculate the absolute value of the inclination difference between the facial droop levels I, II, III, and IV and the user's current facial feature points M1-M3 and M2-M3, as shown in formula (4):

[0081]

[0082] 4. The absolute values ​​of the inclination differences of the facial features of the user and the facial sagging degrees I, II, III, and IV calculated in steps 1, 2, and 3 are combined into a 5*3 matrix G. Each row is the inclination difference g1, g2, and g3 of the facial features corresponding to the degree of facial sagging, as shown in the following matrix:

[0083]

[0084] 5. Normalize G to get the matrix R = [r ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and the specific normalization method is shown in formula (5):

[0085]

[0086] 6. Given the weight matrix Calculate the weight normalization matrix T = R·W1, T = [z ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and T is the dot product of R and W1.

[0087] 7. The smaller the absolute value of the inclination difference of facial feature points, the more symmetrical the face is, so the ideal solution matrix is and the negative ideal solution matrix It can be calculated by formula (6) and (7) respectively

[0088]

[0089] 8. Using Euclidean distance as a measure, calculate any feasible solution z ij To the ideal solution With negative ideal solution Distance as well as They can be calculated according to formulas (8) and (9) respectively:

[0090]

[0091] 9. Calculate the relative closeness of each solution The current degree of facial sagging is classified as C i The closest degree of facial sagging.

[0092] Step 4: Use the TOPSIS method to comprehensively consider the muscle strength difference, slurred speech, and facial droop to quickly identify the risk of stroke. The specific steps are as follows:

[0093] 1. Combine the stroke risk levels I, II, III, and IV calculated in steps 1, 2, and 3, as well as the muscle strength difference level, slurred speech level, and facial ptosis level of the current user into a 5*3 matrix H. The first four rows are the muscle strength difference level, slurred speech level, and facial ptosis level corresponding to different stroke risk levels, respectively. The fifth row is the muscle strength difference level h1, slurred speech level h2, and facial ptosis level h3 of the current user. The value range of the muscle strength difference level is [1,2,3], representing the risk level I, II, and III, respectively. The value range of the slurred speech level is [1,2,3,4], representing the risk level I, II, III, and IV, respectively. The value range of the facial ptosis level is [1,2,3,4], representing the risk level I, II, III, and IV, respectively. As shown in the following matrix:

[0094]

[0095] 2. Normalize G to get the matrix R′=[r ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and the specific normalization method is shown in formula (5):

[0096]

[0097] 3. According to the theoretical practice results, the weight value of stroke risk determined by poor muscle strength, slurred speech and facial droop is 5:3:2, so the weight matrix is ​​given as Calculate the weight normalization matrix T′=R′·W2, T′=[z ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and T′ is the dot product of R′ and W2.

[0098] 4. The smaller the absolute value of the difference in the inclination of facial feature points, the more symmetrical the face is, so the ideal solution matrix is and the negative ideal solution matrix It can be calculated by formula (6) and (7) respectively

[0099]

[0100] 5. Using Euclidean distance as a measure, calculate any feasible solution z ij To the ideal solution With negative ideal solution Distance as well as They can be calculated according to formulas (8) and (9) respectively:

[0101]

[0102] 6. Calculate the relative closeness of each solution Classify the current stroke risk level as O i The closest stroke risk level.

[0103] Step 5: If the identified stroke risk level reaches the standard for medical treatment, the smart terminal automatically dials the 120 emergency number to improve the efficiency and quality of emergency medical rescue services.

[0104] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for rapid stroke identification combining TOPSIS and artificial intelligence, characterized in that: The patient's muscle strength, voice and current facial photo are collected on the mobile terminal, the patient's muscle strength grade, slurred speech and facial droop are analyzed, and the current stroke risk level is identified. The specific steps include the following: Step 1: Determine whether there is unilateral muscle strength loss by the degree to which the user raises both arms horizontally, and grade the muscle strength difference by the angle between the arms; Step 2: Analyze the user's articulation level through artificial intelligence speech recognition and combined with the error word rate calculation method; Step 3: Detect the position coordinates of facial feature points based on the artificial intelligence facial feature point recognition model, and calculate the degree of facial sagging by the TOPSIS method; Step 4: Rapidly identify the risk of stroke by using the TOPSIS method based on the muscle strength difference grade, slurred speech level, and facial droop; Wherein, the step 4 is specifically as follows: A. The stroke risk levels I, II, III, and IV calculated in steps 1, 2, and 3 as well as the muscle strength difference level, slurred speech level, and facial droop level of the current user are combined into a 5*3 matrix H, where the first four rows are the muscle strength difference level, slurred speech level, and facial droop level corresponding to different stroke risk levels, respectively. The fifth row is the muscle strength difference level h1, slurred speech level h2, and facial droop level h3 of the current user. The value range of the muscle strength difference level is [1, 2, 3], representing the risk level I, II, and III, respectively. The value range of the slurred speech level is [1, 2, 3, 4], representing the risk level I, II, III, and IV, respectively. The value range of the facial droop level is [1, 2, 3, 4], representing the risk level I, II, III, and IV, respectively, as shown in the following matrix: B. Normalize the matrix H to get the matrix R′=[r ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and the specific normalization method is shown in formula (5): C. The weights of poor muscle strength, slurred speech, and facial droop are 5:3:2 for stroke risk, so the weight matrix is Calculate the weight normalization matrix T′=R′·W2, T′=[z ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and T′ is the dot product of R′ and W2; D. The smaller the absolute value of the difference in the inclination of facial feature points, the more symmetrical the face is, so the ideal solution matrix is and the negative ideal solution matrix Calculate using formula (6) and (7) respectively E. Use Euclidean distance as a measure to calculate any feasible solution z ij To the ideal solution With negative ideal solution Distance as well as Calculate according to formula (8) and (9) respectively: F. Calculate the relative closeness of each solution Classify the current stroke risk level as O i The closest stroke risk level.

2. The method for rapid stroke identification combining TOPSIS and artificial intelligence according to claim 1, characterized in that: In step 1, by raising both arms horizontally, observing whether both arms can be raised to the same height and whether one arm is weak or drooping, the decrease in unilateral muscle strength can be determined, and the degree of difference in bilateral muscle strength can be determined by comparing the angles of the two arms.

3. The method for rapid stroke identification combining TOPSIS and artificial intelligence according to claim 2, characterized in that: The specific steps are as follows: (1) The patient stands with both hands raised to a height that can be reached; (2) Compare the angle differences between the two arms and divide them into three levels: 0° to 30° as level I, 30° to 60° as level II, and 60° to 90° as level III; (3) The user inputs the level into the mobile terminal.

4. The method for rapid stroke identification combining TOPSIS and artificial intelligence according to claim 1, characterized in that: In step 2, the patient is asked to read aloud the given text, and the text is recognized as corresponding text through a speech recognition interface, and then the error word rate is calculated to quantify the degree of slurred speech of the patient.

5. The method for rapid stroke identification combining TOPSIS and artificial intelligence according to claim 4, characterized in that: The specific steps are as follows: 1) The mobile terminal randomly selects a paragraph of text with 15-30 words from the sentence library and prompts the user to read it aloud; 2) The mobile terminal converts the read voice into text through the voice recognition interface; 3) Calculate the error rate R, and quantify the level of inarticulation according to the error rate, where 0 to 25% is level I, 25% to 50% is level II, 50% to 75% is level III, and more than 75% is level IV; The error rate R is calculated by formula (1): Among them, I is the number of characters added to the converted text relative to the original text, D is the number of characters reduced from the converted text relative to the original text, and C is the number of characters corresponding to the original text.

6. The method for rapid stroke identification combining TOPSIS and artificial intelligence according to claim 1, characterized in that: In the step 3: first, the facial features are identified by a facial feature point recognition method, including the left corner of the left eye E1, the right corner of the left eye E2, the left corner of the right eye E3, the right corner of the right eye E4, the root of the nose N, the left corner of the mouth M1, the right corner of the mouth M2, and the lower lip feature point M3; then the inclinations between the feature points are calculated, and the absolute value of the difference in inclinations of the symmetrical feature points is used as an indicator, and the degree of sagging of a typical face is used as a reference to calculate the degree of sagging of the user's current face.

7. The method for rapid stroke identification combining TOPSIS and artificial intelligence according to claim 6, characterized in that: The specific steps are as follows: a. Calculate the absolute value g1 of the inclination difference between the facial feature points E1-N and E4-N of the user's current facial sagging level I, II, III, and IV, respectively, as shown in formula (2): Among them, y represents the ordinate of the feature point, x represents the abscissa of the feature point, and abs represents the absolute value; b. Calculate the absolute value g2 of the inclination difference between the facial feature points E2-N and E3-N of the user's current facial sagging level I, II, III, and IV, respectively, as shown in formula (3): c. Calculate the absolute value g3 of the inclination difference between the facial feature points M1-M3 and M2-M3 of the facial sagging degree I, II, III, and IV respectively, as shown in formula (4): d. The absolute values ​​of the inclination differences of the facial feature points of the user and the facial sagging degrees I, II, III, and IV calculated in steps a, b, and c are combined into a 5*3 matrix G, where each row is the inclination difference g1, g2, and g3 of the facial feature points corresponding to the facial sagging degree, as shown in the following matrix: e. Normalize G to get the matrix R = [r ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and the specific normalization method is shown in formula (5): f. Given a weight matrix Calculate the weight normalization matrix T = R·W1, T = [z ij ], where i = 1, 2, ..., 5, j = 1, 2, 3, and T is the dot product of R and W1; g. The smaller the absolute value of the difference in the inclination of facial feature points, the more symmetrical the face is, so the ideal solution matrix is and the negative ideal solution matrix Calculate using formula (6) and (7) respectively: h. Use Euclidean distance as a measure to calculate any feasible solution z ij To the ideal solution With negative ideal solution Distance as well as Calculate according to formula (8) and (9) respectively: i. Calculate the relative closeness of each solution The current degree of facial sagging is classified as C i The closest degree of facial sagging.

Images