Pulse wave curve fitting method based on improved ABC algorithm

Abstract

The present invention provides a pulse wave curve fitting method based on an improved ABC algorithm. In the pulse wave curve fitting method, since the honey source is randomly normally distributed after initialization, the diversity of the honey source can be improved, the possibility of local convergence can be reduced, and the bees that cross the boundary can be evenly placed within a limited range, thereby improving the convergence efficiency of the algorithm. Also, since the employed bees search for the employed honey source through the bee search method, more opportunities for the employed bees to jump out of the restrictions in a single dimension are provided, and a higher possibility is provided for leading to the global optimal solution. In addition, since the bees are converted into scout bees according to the chaotic traversal characteristics, the scout bees find the scout honey source according to the mapping algorithm, thereby improving the efficiency of the global search for the optimal honey source. The optimal simulation curve is drawn according to the optimal honey source, laying a foundation for the subsequent accurate classification of the vector machine and the establishment of the disease prediction model, and finally realizing the accurate analysis of human physiological information.

Description

Technical Field

[0001] The invention relates to a pulse wave curve fitting method based on an improved ABC algorithm. Background Art

[0002] The pulse wave is formed by the pulsation of the heart propagating along the arteries and blood flow to the periphery, which can reflect the state of the cardiovascular system and blood fluid dynamics. Accurately obtaining the pulse characteristics in the pulse wave has become a key link in accurately analyzing the physiological information of the human body based on the pulse wave.

[0003] Since there is a lot of noise in the pulse wave signal collected by wearable devices, it is difficult to distinguish the main wave, tidal wave and dicrotic wave from each other; and the overlap of multiple pulse waves will cause some sampling points in the pulse wave to disappear. All these make the acquired pulse wave signal inaccurate, making it impossible to extract feature parameters from the measured pulse wave, thus affecting the subsequent machine learning classifier's use of accurate pulse waves to establish a disease prediction model.

[0004] In response to the above problems, some people have proposed to model the pulse wave through curve fitting to describe the signal trend of the pulse wave. Among them, the Gaussian model can better characterize the cardiovascular function-related parameters related to pulse decomposition. However, limited by the inherent shape of the curve and the number of functions, it is difficult for the Gaussian mixture function to fit the peak and details of the pulse wave well at the same time. There are also problems such as few high-order functions, many parameters to be optimized, and many iterations. In addition to the Gaussian model, group algorithms are also suitable for curve fitting, such as genetic algorithms, particle swarm algorithms, ant colony algorithms, and bee colony algorithms (ABC algorithms). Among them, the ABC algorithm has attracted attention due to its relatively simple concept, few variables, and easy modification. The traditional ABC algorithm has problems such as poor search strategy, slow convergence speed, and low precision. It cannot fit a better pulse wave curve and cannot accurately extract pulse wave features for further classification. Summary of the invention

[0005] To solve the above problems, the present invention provides a pulse wave fitting method that can extract characteristic parameters from the measured pulse wave by fitting a simulation curve formed by superimposing multiple Gaussian functions and the measured pulse wave. The present invention adopts the following technical solutions:

[0006] The present invention provides a pulse wave curve fitting method based on an improved ABC algorithm, which is used for fitting a simulation curve formed by mixing multiple Gaussian functions with a measured pulse wave to form a fitting curve, and is characterized in that it comprises the following steps: step S1, treating each parameter to be optimized in the simulation curve as a vector element of a feasible solution and setting it as a nectar source, initializing the nectar source to obtain an initial nectar source and assigning an employed bee to each initial nectar source; step S2, calculating the fitness value of the initial nectar source and performing random normal distribution near the initial nectar source corresponding to the optimal fitness value to obtain a random nectar source; step S3, employing The hired bees search around the random honey source according to the predetermined bee search method and obtain the hired honey source; step S4, the hired bees recruit follower bees, and the follower bees perform selection operations among the hired honey source and the random honey source; step S5, generate scout bees according to the chaotic array mapping, and the scout bees obtain the scout honey source according to the mapping algorithm; step S6, record the best nectar source among all hired honey sources and the scout honey source as the optimal honey source; step S7, repeat steps S2 to S6 until the number of repetitions reaches the predetermined maximum number of cycles, obtain the only optimal honey source, and draw the optimal simulation curve according to the optimal honey source.

[0007] The pulse wave curve fitting method based on the improved ABC algorithm provided by the present invention may also have the following technical features: Step S2-1, the initial nectar source is evenly divided into a first initial nectar source set X k And the second initial nectar source set O k Two parts, for the first initial nectar source set X k And the second initial nectar source set O k Randomly distribute and obtain the i-th initial nectar source X in the first initial nectar source set i,k Location and the second initial nectar source concentration S n -i initial nectar sources Position, as shown in formula (1):

[0008]

[0009] Where k∈{1,2,…,K}, K is the maximum number of parameters to be optimized in the simulation curve, S n is the number of employed bees, the first initial nectar source set X k is a vector containing the number of parameters to be optimized of the simulation curve, the second initial honey source set O k Is the first initial nectar source set X k The opposite solution obtained by half splitting within the boundary; Step S2-2, evaluate the first initial honey source set X k Each initial nectar source in and the second initial nectar source set O kThe initial nectar sources in the data are selected to obtain a fitness value representing the quality of each initial nectar source; step S2-3, selecting a plurality of relatively good nectar sources from all the initial nectar sources according to the fitness value and taking them as good nectar sources; step S2-4, performing random normal distribution around the good nectar sources to obtain random nectar sources.

[0010] The pulse wave curve fitting method based on the improved ABC algorithm provided by the present invention may also have such a technical feature, wherein, when four relatively good nectar sources are selected as good nectar sources in step S2-3, the four good nectar sources are marked as X respectively. b and O b , X s and O s , around X b and O b , X s and O s Perform a random normal distribution to obtain the random nectar source Y:

[0011]

[0012] Where D 1,2,3,4 Refers to the four subsets generated by random normal distribution, D min With D max is a fixed parameter, and D is a combination of the above four subsets.

[0013] The pulse wave curve fitting method based on the improved ABC algorithm provided by the present invention may also have such a technical feature, wherein the predetermined bee search method is a bee search method combining cross-dimensional search and adaptive search, and the cross-dimensional update of the employed honey source Y' can be realized through formula (3): i,k :

[0014] Y' i,k =Y t,k +φ i,k (Y i,k -Y j,k ) (3)

[0015] Where i, t, j represents the i-th, t-th and j-th employed bees in the bee colony, i, t, j∈{1,2,...,S n} is randomly created, and j≠i≠t, φ i,k is a random number uniformly distributed between [-1,1], and the employed honey source Y' i,k is through random nectar source Y i,k and Y t.k Combining them, we can use formula (4) to update the hiring honey source Y' across neighborhoods i,k :

[0016]

[0017] Where k, l represent different dimensions, k, l∈{1,2,…,K}, and k≠l,φ i,l is a random number between {-1, 1}, α, β are fixed values, f i , f j is the fitness value of the i-th and j-th employed honey sources in the bee colony.

[0018] The pulse wave curve fitting method based on the improved ABC algorithm provided by the present invention may also have such a technical feature, wherein step S5 also includes the following sub-steps: step S5-1, generating scout bees according to chaotic array mapping, and the scout bees obtain scout honey sources according to the mapping algorithm; step S5-2, judging whether there are abandoned hired honey sources, and entering step S5-3 when judging that there are, and entering step S6 when judging that there are not; step S5-3, the hired bees are converted into scout bees, and randomly search for scout honey sources in the solution space.

[0019] The pulse wave curve fitting method based on the improved ABC algorithm provided by the present invention may also have such a technical feature, wherein step S5-1 includes the following sub-steps: step T1, when the employed bee becomes a scout bee, a K-dimensional initial vector y is created 0 =[y 1,1 ,y 1,2 ,...,y 1,K ], step T2, generate the chaotic array y according to formula (5) z+1,k :

[0020] y z+1,k =μy z,k (1-y z,k ) (5)

[0021] Where z∈{1,2,...,Z}, z is the number of dimensions, and μ is the factor causing confusion; Step T3, according to y 0 And formula (5) calculates y z,k Step T4, according to formula (6) to obtain the honey source Y' z,k :

[0022] Y' z,k =Y i,k +r i,k (2y z,k -1) (6)

[0023] Where Y i,k is a random nectar source, Y' z,k Y i,k Centered, r i,k The scout honey source obtained by radius mapping.

[0024] Function and Effect of the Invention

[0025] According to the pulse wave curve fitting method based on the improved ABC algorithm of the present invention, since each parameter to be optimized in the simulation curve formed by superimposing multiple Gaussian functions is set as a honey source, and the honey source is initialized and randomly normally distributed, it is possible to improve the diversity of honey sources, reduce the possibility of local convergence caused by random creation of honey sources, and effectively solve the problem of bees crossing the boundary at the edge due to honey source updates, and evenly place the bees that cross the boundary within a limited range, thereby improving the convergence efficiency of the algorithm. Also, since the employed bees search for the employed honey source around the random honey source through the bee search method, it provides the employed bees with more opportunities to jump out of the restrictions in a single dimension, and provides a higher possibility for leading to the global optimal solution. In addition, since the employed bees are converted into scout bees according to the chaotic ergodic characteristics, and the scout bees find the scout honey source according to the mapping algorithm, the efficiency of the global search for the optimal honey source is improved. Based on the optimal nectar source finally obtained, a simulation curve is drawn to better fit the peak value and details of the pulse wave, and then more accurate and useful characteristic parameters of the measured pulse wave are extracted from the simulation curve, which also lays the foundation for the subsequent precise classification of vector machines and the establishment of disease prediction models, and ultimately realizes accurate analysis of human physiological information. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 Flow chart of a pulse wave curve fitting method based on an improved ABC algorithm according to an embodiment of the present invention;

[0027] Figure 2 This is an example diagram of a single cycle of a pulse wave actually measured at the wrist of a healthy person according to an embodiment of the present invention;

[0028] Figure 3 A comparison diagram of the pulse wave fitting effect of a single cycle according to an embodiment of the present invention; and

[0029] Figure 4 4 is a comparison diagram of the pulse wave fitting effects of multiple cycles according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the technical means, creative features, objectives and effects of the present invention easy to understand, the pulse wave curve fitting method based on the improved ABC algorithm of the present invention is specifically described below in conjunction with embodiments and drawings.

[0031] <Example>

[0032] Figure 1 It is a flow chart of a pulse wave curve fitting method based on an improved ABC algorithm according to an embodiment of the present invention.

[0033] like Figure 1As shown, the pulse wave curve fitting method based on the improved ABC algorithm includes the following steps:

[0034] Step S1, treating each parameter to be optimized in the simulation curve as a vector element of a feasible solution and setting it as a nectar source, initializing the nectar source to obtain an initial nectar source and assigning an employed bee to each initial nectar source.

[0035] Among them, each initial nectar source corresponds to a hired bee.

[0036] Step S2, calculating the fitness value of the initial nectar source and performing random normal distribution near the initial nectar source corresponding to the optimal fitness value to obtain a random nectar source.

[0037] In step S2 of this embodiment, the following sub-steps are also included:

[0038] Step S2-1, divide the initial nectar source into the first initial nectar source set X k And the second initial nectar source set O k Two parts, for the first initial nectar source set X k And the second initial nectar source set O k Randomly distribute and obtain the i-th initial nectar source X in the first initial nectar source set i,k Location and the second initial nectar source concentration S n -i initial nectar sources Position, as shown in formula (1):

[0039]

[0040] Where k∈{1,2,…,K}, K is the maximum number of parameters to be optimized in the simulation curve, S n is the number of employed bees, the first initial nectar source set X k is a vector containing the number of parameters to be optimized of the simulation curve, the second initial honey source set O k Is the first initial nectar source set X k The opposite solution obtained by half-splitting within the boundary.

[0041] Among them, X k represents a series of feasible solutions for the kth parameter to be optimized, X i,k It is represented as the i-th feasible solution of the k-th parameter to be optimized, that is, X k The location of an initial nectar source in a set of initial nectar sources.

[0042] In this embodiment, the initial nectar source is divided into a first initial nectar source set X according to the symmetry of the random normal distribution. k And the second initial nectar source set O k Two parts, O k Yes X kThe opposite solution obtained by partitioning within the boundary. k Contains half of the feasible solutions for the kth parameter to be optimized, and O k Contains the other half and X k The feasible solutions that are opposite to each feasible solution in .

[0043] Step S2-2, evaluate the first initial nectar source set X k Each initial nectar source in and the second initial nectar source set O k For each initial nectar source in , we get the fitness value representing the quality of each initial nectar source.

[0044] In this embodiment, each initial nectar source is evaluated according to formula (a).

[0045]

[0046] In the formula, n refers to the horizontal coordinate of the Gaussian function, that is, the number of curve points, m refers to the number of Gaussian functions, and a k 、b k and c k are the amplitude, center position and standard deviation of the kth Gaussian function, d m and e m are the slope and constant of the linear piecewise function, m represents the number of pulse cycles, p 1 ,p 2 ,...,p m+1 (m=1, 2, ...) is the boundary of each complete pulse cycle, usually located at the valley point of the multi-cycle wave, and F(n) represents the measured pulse wave.

[0047] Step S2-3, selecting a plurality of relatively good initial nectar sources from all initial nectar sources according to the fitness values ​​and using them as good nectar sources.

[0048] In this embodiment, four relatively good initial nectar sources are selected as good nectar sources, and the four good nectar sources are marked as X respectively. b and O b , X s and O s The number of each good nectar source is S n one quarter of .

[0049] Step S2-4, performing random normal distribution around good nectar sources to obtain random nectar sources.

[0050] In this embodiment, around X b and O b , X s and O s Perform a random normal distribution to obtain the random nectar source Y:

[0051]

[0052] Where D 1,2,3,4 Refers to the four subsets generated by random normal distribution, D min With D max is a fixed parameter, D is a combination of the above four subsets, which can be X b and O b , X s and O s Choose the best nectar source, D min With D max is a fixed parameter. 1,2,3,4 Normal distribution is performed around the bees to improve diversity and reduce the possibility of local convergence caused by random creation of bees.

[0053] In this embodiment, D min Set to 80%, D max Set to 120%.

[0054] Step S3, the employed bees search around the random honey source according to a predetermined bee search method and obtain the employed honey source.

[0055] Among them, the predetermined bee search method is a bee search method that combines cross-dimensional search and adaptive search.

[0056] In this embodiment, formula (3) can be used to achieve cross-dimensional updating of the hired honey source Y' i,k :

[0057] Y' i,k =Y t,k +φ i,k (Y i,k -Y j,k ) (3)

[0058] Where i, t, j represents the i-th, t-th and j-th employed bees in the bee colony, i, t, j∈{1,2,...,S n} is randomly created, and j≠i≠t, φ i,k A random number between -1 and 1, hiring honey source Y' i,k is through random nectar source Y i,k and Y t.k Instead of using only Y in the traditional ABC algorithm i,k Calculated.

[0059] According to formula (3), the employed honey source can be updated in other dimensional search spaces, which also provides bees with more opportunities to jump out of the restrictions in a single dimension and provides a higher possibility of leading to the global optimal solution.

[0060] Formula (4) can be used to update the hiring honey source across neighborhoods:

[0061]

[0062] Where k, l represent different dimensions, k, l∈{1,2,…,K}, and k≠l,φ i,l is a random number between {-1, 1}, α, β are fixed values, f i , f j is the fitness value of the i-th and j-th employed honey sources in the bee colony. When the fitness value is positive, the optimization problem is transformed into a minimum solution search problem.

[0063] In this embodiment, the adaptive search performs quantitative calculation of the employed honey source based on the fitness value within the iteration time to achieve cross-neighborhood update of the employed honey source.

[0064] Formula (3) and formula (4) can be used to implement a bee search method that combines cross-dimensional search and adaptive search, and complete the cross-dimensional and cross-neighborhood update of the hired honey source.

[0065] In this embodiment, after the employed bee finds the employed honey source, it uses the conventional greedy strategy to select the employed honey source and the random honey source according to the fitness value of the employed honey source. If the employed bee selects the employed honey source, it proceeds to step S4; if the employed bee retains the random honey source, it continues to search for the employed honey source around the random honey source.

[0066] Step S4, the employed bees recruit follower bees, and the follower bees perform selection operations among the employed honey sources and the random honey sources.

[0067] In this embodiment, employed bees recruit follower bees according to a conventional roulette method. Follower bees find employed honey sources according to information transmitted by employed bees, and select a better one between employed honey sources and random honey sources through a conventional greedy strategy.

[0068] Step S5, generating scout bees according to chaotic array mapping, and the scout bees obtain scout honey sources according to the mapping algorithm.

[0069] Wherein, step S5 also includes the following sub-steps:

[0070] Step S5-1, generating a scout bee according to chaotic array mapping, and the scout bee obtains a scout honey source according to a mapping algorithm.

[0071] Wherein, step S5-1 also includes the following sub-steps:

[0072] Step T1: When the maximum number of iterations pre-set during nectar source initialization is reached, the employed bees become scout bees and create a K-dimensional initial vector y 0 =[y 1,1,y 1,2 ,...,y 1,K ].

[0073] Step T2, generate the chaotic array y according to formula (5) z+1,k :

[0074] y z+1,k =μy z,k (1-y z,k ) (5)

[0075] Where z∈{1,2,...,Z}, z is the number of dimensions, and μ is the factor that causes confusion.

[0076] Step T3, according to y 0 And formula (5) calculates y z,k .

[0077] Step T4, according to formula (6), the honey source Y' is obtained z,k :

[0078] Y' z,k =Y i,k +r i,k (2y z,k -1) (6)

[0079] Where Y i,k is a random nectar source, Y' z,k Y i,k Centered, r i,k The scout honey source obtained by radius mapping.

[0080] Step S5-2, determine whether there is an abandoned hired honey source, if it is determined that there is, proceed to step S5-3, if it is determined that there is not, proceed to step S6.

[0081] Step S5-3, the employed bees are transformed into scout bees, and randomly search for honey sources in the solution space.

[0082] Step S6, recording the best honey source among all hired honey sources and scout honey sources as the optimal honey source.

[0083] Step S7, repeating steps S2 to S6 until the number of repetitions reaches a predetermined maximum number of cycles, obtaining a unique optimal nectar source, and drawing an optimal simulation curve based on the optimal nectar source.

[0084] In this embodiment, the optimal solution of all parameters to be optimized will be outputted in each cycle. When the optimal solution of all parameters to be optimized outputted in the next cycle is better than the optimal solution outputted in the previous cycle, the previous optimal solution will be automatically replaced. After a predetermined maximum number of cycles, only one set of optimal solutions of all parameters to be optimized is retained, and the optimal simulation curve is drawn based on this set of optimal solutions.

[0085] Figure 2 This is an example diagram of a single cycle of a pulse wave actually measured at the wrist of a healthy person according to an embodiment of the present invention.

[0086] In this embodiment, the measured pulse waves of 50 healthy candidates are used as the test data set, and 4 of them are randomly selected for fitting quality inspection, respectively. Figure 2 As shown in (a), (b), (c) and (d).

[0087] in, Figure 2 The horizontal axis represents the number of sampling points, and the vertical axis represents the amplitude of the measured pulse wave. Figure 2 "Sample 01" in (a), "Sample 02" in (b), "Sample 03" in (c), and "Sample 04" in (d) are the numbers of the measured pulse waves of the four healthy candidates.

[0088] In this embodiment, the measured pulse wave numbered "sample 01" is selected as the basis for the subsequent single-cycle pulse wave fitting effect analysis.

[0089] Figure 3 This is a comparison diagram of the single-cycle pulse wave fitting effect of an embodiment of the present invention.

[0090] like Figure 3 As shown in (a), (c), and (e), the black solid line is a simulation curve (i.e., experimental data) formed by mixing multiple Gaussian functions, the black dotted line is a simulation curve of the traditional ABC algorithm (i.e., ABC), and the gray dotted line is the optimal simulation curve (i.e., MABC) drawn by the pulse wave curve fitting method based on the improved ABC algorithm of the present invention. Figure 3 In (a), (b) and (c), 3, 4 and 5 Gaussian functions are used for mixing respectively.

[0091] from Figure 3 It can be seen from (a), (c) and (e) in that the optimal simulation curve drawn by the present invention can better fit the peak value and details of the measured pulse wave compared with the simulation curve of the traditional ABC algorithm.

[0092] like Figure 3 As shown in (b), (d), and (f), the solid line is the minimum root mean square (RMS) of the traditional ABC algorithm (i.e., ABC), and the dotted line is the minimum root mean square of the fitting method of the present invention (i.e., MABC). It can be clearly seen that the fitting method of the present invention converges faster than the traditional ABC algorithm and the RMS is less than or equal to the RMS of the traditional ABC algorithm. Therefore, the fitting method of the present invention not only has good detail fitting quality, fast convergence speed, but also high fitting accuracy.

[0093] Figure 44 is a comparison diagram of the pulse wave fitting effects of multiple cycles according to an embodiment of the present invention.

[0094] like Figure 4 As shown in FIG. 1 , the solid line is a simulation curve (i.e., experimental data) formed by mixing five Gaussian functions, and the dotted line is a simulation curve (i.e., fitting curve) of the pulse wave curve fitting method based on the improved ABC algorithm of the present invention. The horizontal axis is the number of sampling points, and the vertical axis is the amplitude. It can be seen that the four test data (such as Figure 2 The present invention can fit a better simulation curve in multiple cycles of each test data in (a), (b), (c), and (d), so the fitting method of the present invention can better draw a simulation curve in a high-dimensional space.

[0095] Example Function and Effect

[0096] According to the pulse wave curve fitting method based on the improved ABC algorithm provided by the above embodiment, since each parameter to be optimized in the simulation curve formed by mixing multiple Gaussian functions is set as a nectar source, and the nectar source is initialized and randomly normally distributed, it is possible to improve the diversity of nectar sources, reduce the possibility of local convergence caused by random creation of nectar sources, and effectively solve the problem of bees crossing the boundary at the edge due to nectar source update, and evenly place the bees crossing the boundary within the limited range, thereby improving the convergence efficiency of the algorithm. Also, since the employed bees search for the employed honey source around the random nectar source by the bee search method, it provides the employed bees with more opportunities to jump out of the restrictions in a single dimension, and provides a higher possibility for leading to the global optimal solution. In addition, since the employed bees are converted into scout bees according to the chaotic traversal characteristics, and the scout bees find the scout honey source according to the mapping algorithm, the efficiency of the global search for the optimal nectar source is improved. According to the optimal nectar source finally obtained, the simulation curve is adjusted to better fit the peak and details of the pulse wave, and the foundation is also laid for the subsequent accurate classification of the vector machine and the establishment of the disease prediction model, and finally the accurate analysis of human physiological information is realized.

[0097] In addition, in the embodiment, since the random nectar source is divided into two parts and a plurality of good nectar sources are selected from the two parts of the random nectar sources and then randomly normally distributed around the plurality of good nectar sources, diversity can be improved and the possibility of local convergence caused by random creation of nectar sources can be reduced.

[0098] In addition, in the embodiment, since the bees that have crossed the boundary are reset within the limited range, the efficiency of algorithm convergence can be improved.

[0099] In addition, in the embodiment, since a cross-dimensional search method is used, the employed bees can be updated in combination with other dimensions, thereby increasing information sharing among the bees.

[0100] In addition, in the embodiment, since the scout bees are generated by chaotic array mapping and the scout bees search for scout honey sources according to the mapping algorithm, the strategy of random generation of scout bees in the traditional ABC algorithm is changed, and the efficiency of global search for the optimal honey source is improved.

[0101] The above embodiments are only used to illustrate specific implementation modes of the present invention, and the present invention is not limited to the description scope of the above embodiments.

[0102] In the above embodiment, the number of good nectar sources is 4. The present invention can also select 2, 6 or other numbers of initial nectar sources for random normal distribution according to the fitness value.

[0103] In the above embodiments, the number of Gaussian functions selected is 3, 4 and 5 respectively. The present invention can also select 6, 7 or more Gaussian functions to mix and obtain the simulation curve.

Claims

1. A pulse wave curve fitting method based on an improved ABC algorithm, which is used to fit a simulated curve formed by superimposing multiple Gaussian functions with a measured pulse wave to extract characteristic parameters from an original measured pulse wave, characterized in that: The following steps are included: Step S1, treating each parameter to be optimized in the simulation curve as a vector element of a feasible solution and setting it as a nectar source, initializing the nectar source to obtain an initial nectar source and assigning an employed bee to each of the initial nectar sources; Step S2, calculating the fitness value of the initial nectar source and performing random normal distribution near the initial nectar source corresponding to the optimal fitness value to obtain a random nectar source; Step S3, the employed bees search around the random honey source according to a predetermined bee search method and obtain an employed honey source; Step S4, the employed bees recruit follower bees, and the follower bees perform selection operations between the employed honey source and the random honey source; Step S5, generating a scout bee according to the chaotic array mapping, and the scout bee obtains a scout honey source according to the mapping algorithm; Step S6, recording the best honey source among all the employed honey sources and the scout honey sources as the optimal honey source; Step S7, repeating steps S2 to S6 until the number of repetitions reaches a predetermined maximum number of cycles, obtaining the only optimal nectar source, and drawing an optimal simulation curve based on the optimal nectar source.

2. The pulse wave curve fitting method based on the improved ABC algorithm according to claim 1, characterized in that: in, The step S2 includes the following sub-steps: Step S2-1, the initial nectar source is evenly divided into a first initial nectar source set X k And the second initial nectar source set O k Two parts, for the first initial honey source set X k And the second initial nectar source set O k Randomly distribute to obtain the i-th initial nectar source X in the first initial nectar source set i,k location and the first S in the second initial nectar source n -i initial nectar sources Position, as shown in formula (1): Where k∈{1,2,…,K}, K is the maximum number of parameters to be optimized for the simulation curve, S n is the number of employed bees, the first initial nectar source set X k is a vector containing the number of parameters to be optimized of the simulation curve, and the second initial honey source set O k is the first initial honey source set X k The opposite solution obtained by half-splitting within the boundary; Step S2-2, evaluating the first initial honey source set X k Each initial nectar source in and the second initial nectar source set O k For each initial nectar source in , a fitness value representing the quality of each initial nectar source is obtained; Step S2-3, selecting a plurality of relatively good initial nectar sources from all the initial nectar sources according to the fitness value and taking them as good nectar sources; Step S2-4, performing random normal distribution around the good nectar source to obtain a random nectar source.

3. The pulse wave curve fitting method based on the improved ABC algorithm according to claim 2, characterized in that: in, When four relatively good initial nectar sources are selected as the good nectar sources in step S2-3, the four good nectar sources are marked as X respectively. b and O b , X s and O s , Around X b and O b , X s and O s The random nectar source Y is obtained by performing a random normal distribution: Where D 1,2,3,4 Refers to the four subsets generated by random normal distribution, D min With D max is a fixed parameter, and D is a combination of the above four subsets.

4. The pulse wave curve fitting method based on the improved ABC algorithm according to claim 2, characterized in that: in, The predetermined honey bee search method is a honey bee search method that combines cross-dimensional search and adaptive search. The cross-dimensional update of the hired honey source Y' is achieved through formula (3) i,k : AND' i,k =And t,k +φ i,k (AND i,k -AND j,k ) (3) Where i, t, j represents the i-th, t-th and j-th employed bees in the bee colony, i, t, j∈{1,2,...,S n } is randomly created, and j≠i≠t, φ i,k is a random number uniformly distributed between [-1,1], the hired honey source Y' i,k is the random nectar source Y i,k and Y t.k Combined to get, The cross-neighborhood update of the employed honey source Y' is realized by formula (4) i,k : Where k, l represent different dimensions, k, l∈{1,2,…,K}, and k≠l,φ i,l is a random number between {-1, 1}, α, β are fixed values, f i , f j is the fitness value of the i-th and j-th employed honey sources in the bee colony.

5. The pulse wave curve fitting method based on the improved ABC algorithm according to claim 1, characterized in that: in, The step S5 further comprises the following sub-steps: Step S5-1, generating the scout bee according to chaotic array mapping, and the scout bee obtains the scout honey source according to the mapping algorithm; Step S5-2, determining whether there is any abandoned hired honey source, if it is determined that there is, proceeding to step S5-3, if it is determined that there is no, proceeding to step S6; Step S5-3, the employed bees are transformed into scout bees, and randomly search for the scout honey source in the solution space.

6. The pulse wave curve fitting method based on the improved ABC algorithm according to claim 5, characterized in that: in, The step S5-1 includes the following sub-steps: Step T1, when the employed bee becomes the scout bee, create a K-dimensional initial vector y0=[y 1,1 ,y 1,2 ,...,y 1,K ], Step T2, generate the chaotic array y according to formula (5) z+1,k : y z+1,k =μy z,k (1-y z,k ) (5) Where z∈{1,2,...,Z}, z is the number of dimensions, and μ is the factor causing confusion; Step T3, calculate y according to y0 and formula (5) z,k ; Step T4, according to formula (6), the detected honey source Y' is obtained z,k : Y’ z,k =Y i,k +r i,k (2y z,k -1) (6) Where Y i,k is the random nectar source, Y' z,k Y i,k Centered, r i,k The scouting honey source is obtained by radius mapping.

Images