CN115633957A - Blood glucose prediction method and system based on high-order and fraction low-order statistics - Google Patents
Blood glucose prediction method and system based on high-order and fraction low-order statistics Download PDFInfo
- Publication number
- CN115633957A CN115633957A CN202211274466.1A CN202211274466A CN115633957A CN 115633957 A CN115633957 A CN 115633957A CN 202211274466 A CN202211274466 A CN 202211274466A CN 115633957 A CN115633957 A CN 115633957A
- Authority
- CN
- China
- Prior art keywords
- data
- blood sugar
- order
- imf
- blood glucose
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention provides a blood sugar prediction method and a blood sugar prediction system based on high-order and fractional low-order statistics, which relate to the technical field of blood sugar prediction and solve the problems that in the current noninvasive blood sugar prediction method, the detail information of a near-infrared blood sugar signal cannot be extracted and the accuracy of blood sugar data prediction is low.
Description
Technical Field
The invention relates to the technical field of blood sugar prediction, in particular to a blood sugar prediction method and a blood sugar prediction system based on high-order and fraction low-order statistics.
Background
Diabetes refers to a metabolic disorder syndrome mainly manifested by fasting hyperglycemia or postprandial hyperglycemia caused by absolute or relative insufficiency of insulin secretion, and currently, diabetes cannot be cured in clinical medicine, and diabetics can only effectively manage and control the diabetes by taking artificial insulin, exercising, careful diet intake and frequent blood sugar monitoring.
Blood glucose monitoring is mainly divided into invasive monitoring and non-invasive monitoring, wherein the invasive monitoring causes pain and trauma of a user when a blood sample is collected, and test paper of the blood sample is needed, so that the cost is high; the non-invasive monitoring mainly adopts a near infrared spectrum to carry out non-invasive blood sugar detection, carries out blood sugar prediction on signals acquired by near infrared light, does not need blood sampling, and is convenient and accurate to detect. The prior art discloses a noninvasive blood glucose prediction method, which comprises the steps of obtaining a maximum near-infrared signal from a near-infrared signal in blood glucose data, extracting a mean value, a variance, a slope and a peak value of the maximum near-infrared signal, constructing a characteristic signal, constructing a mapping matrix according to the characteristic signal and a blood glucose value of the blood glucose data, inputting a signal to be detected into the mapping matrix to generate a mapping matrix to be detected, and optimizing the mapping matrix to be detected by using a convolutional neural network.
Disclosure of Invention
In order to solve the problems that the detailed information of a near-infrared blood glucose signal cannot be extracted and the accuracy of blood glucose data prediction is low in the current noninvasive blood glucose prediction method, the invention provides a blood glucose prediction method and system based on high-order and fractional low-order statistics, the detailed information of the near-infrared blood glucose signal can be extracted, and the accuracy of blood glucose data prediction is effectively improved.
In order to achieve the technical effects, the technical scheme of the invention is as follows:
a blood glucose prediction method based on high-order and fractional low-order statistics comprises the following steps:
s1, obtaining blood glucose data of a subject, wherein the blood glucose data comprises near-infrared blood glucose index data of the subject and actual blood glucose values of a human body synchronously corresponding to the near-infrared blood glucose index data, and setting the obtained near-infrared blood glucose index data as a noisy signal;
s2, decomposing the signal with noise to obtain a de-noising signal;
s3, respectively extracting characteristic values of the high-order statistic and the fraction low-order statistic of the de-noising signal, and constructing a characteristic matrix containing a plurality of characteristic values;
s4, dividing the characteristic values in the characteristic matrix and the actual blood sugar values corresponding to the characteristic values in the characteristic matrix into a training set and a verification set;
s5, constructing a blood sugar prediction model, training the blood sugar prediction model by using a training set, and evaluating the blood sugar prediction model in the training process by using a verification set to obtain the trained blood sugar prediction model;
s6, collecting blood sugar data to be detected, inputting the blood sugar data to be detected into the trained blood sugar prediction model, and outputting a blood sugar prediction result of the blood sugar data to be detected.
According to the technical scheme, near-infrared blood glucose index data serving as a noisy signal is obtained firstly, the noisy signal is decomposed to obtain a de-noised signal, effective pre-de-noising processing of the noisy signal is guaranteed at the aim, a reliable signal data source is provided, then a high-order statistic and a score low-order statistic characteristic value of the de-noised signal are extracted, more detailed information of a non-Gaussian near-infrared blood glucose signal is extracted, so that the characteristics of the non-Gaussian near-infrared blood glucose signal are fully reflected, a blood glucose prediction model is further constructed, the blood glucose prediction model is trained by using a training set in a data set consisting of the extracted characteristic value and an obtained actual blood glucose value, finally the collected blood glucose data to be detected is detected by using the trained blood glucose prediction model, a blood glucose prediction result of the blood glucose data to be detected is obtained, and the blood glucose data prediction accuracy is effectively improved.
Preferably, in step S1, the near-infrared blood glucose index data is represented as [ x ] 1 ，x 2 ，…，x n ]Sequence, n representing the total number of collected near-infrared blood glucose indicator data that was non-invasively acquired using a near-infrared LED sensor, the actual blood glucose value being of interestInvasive access was achieved with a glucometer.
Preferably, in step S2, let the noisy signal be x (n), and the specific step of decomposing the noisy signal x (n) is:
s21, carrying out EMD on the noisy signal x (n) to obtain k IMF data components and 1 remainder data component, wherein k is more than or equal to 2;
s22, M IMF data components are selected from the k IMF data components, k is larger than M, SSA decomposition is carried out on each IMF data component in the M IMF data components to obtain an SVD component corresponding to each IMF data component in the M IMF data components, and a subset { X corresponding to the SVD component is constructed according to the obtained SVD component I }；
S23, corresponding subset { X ] of each SVD component I Reconstructing to obtain M groups of reconstruction data;
s24, accumulating M groups of reconstruction data and remainder data components, and outputting a de-noising signal x i 。
Preferably, before performing SSA decomposition on each of the M IMF data components, each of the M IMF data components is divided, and the specific dividing step is: first, the noise energy E of each of M IMF data components is collected nk Then, calculating the denoising energy of each IMF data component in the IMF data components, wherein the specific calculation expression is as follows:
E‘ xk ＝E xk -E nk
wherein E is k Representing the energy, E ', of each of the IMF data components' xk Representing a denoising energy for each of the IMF data components;
calculating E' xk And E k Ratio R k The specific calculation expression is as follows:
according to the ratio R k Each of the M IMF data components is partitioned.
Preferably, the characteristic value of the higher order statistic of the denoised signal includes a skewness of the third order statistic and a kurtosis of the fourth order statistic, and the specific calculation expression of the skewness of the third order statistic is as follows:
wherein SK i Representing the skewness of the third-order statistic corresponding to the ith preprocessing data, n representing the length of the near-infrared blood sugar original data, mu i3 Represents the third-order central moment, s, corresponding to the ith preprocessing data i Represents the sample standard deviation;
μ i3 the specific calculation expression of (2) is:
wherein the content of the first and second substances,
representing a denoised signal x i The average value of (a) of (b),
the specific calculation expression of (2) is:
s i the specific calculation expression of (2) is:
the specific calculation expression of the fourth order statistic kurtosis is as follows:
wherein, K i Represents the kurtosis, mu, of the fourth order statistic corresponding to the ith pre-processed data i2 Represents the second central moment, mu, corresponding to the ith pre-processed data i4 And representing the fourth-order central moment corresponding to the ith preprocessing data.
Preferably, the eigenvalues of the fractional low order statistic include an eigenvalue α and a dispersion coefficient γ.
Preferably, a logarithm method is used for extracting a characteristic index alpha and a dispersion coefficient gamma representing the low order statistic of the fraction of the de-noised signal, and the specific extraction steps of the characteristic index alpha and the dispersion coefficient gamma are as follows:
s31, setting x i Is a group of x with characteristic index alpha and symmetrical coefficient beta both being 0 stably distributed random variables and stably distributed for alpha i Deaveraging process
Obtaining a finite negative moment E { | x of an alpha-stable random variable with a position parameter a =0 i | p H, a negative order moment E { | x i | p The following conditions are satisfied:
wherein p represents a fractional order; c (p, α) represents a function of α, p, and a random variable x i Regardless, the specific computational expression for C (p, α) is:
wherein Γ (·) represents a gamma-ma function;
for E { | x i | p Performing equivalent transformation, wherein a specific expression is as follows:
let Y = log | x i I, to
And (3) carrying out conversion, wherein a specific conversion expression is as follows:
wherein, E { E } pY Denotes the first eigenfunction generation of Y, E { E } pY The specific calculation expression of is:
E(Y k ) The specific calculation expression of (2) is:
s32, calculating the mean value E (Y) and the variance Var (Y) of Y to obtain values of E (Y) and Var (Y);
s33, carrying out equivalent transformation on the variance Var (Y), wherein the specific expression is as follows:
substituting the Var (Y) value obtained in the step S32 into the formula in the step S33 to solve to obtain a value of the characteristic index alpha;
s34, performing equivalent transformation on the average value E (Y), wherein the specific expression is as follows:
wherein, C e Represents the Euler constant;
and substituting the alpha value obtained in the step S33 into the formula in the step S34 to solve to obtain the value of the dispersion coefficient gamma.
Preferably, in step S5, the construction method of the blood sugar prediction model is a random forest algorithm.
The present application further provides a system for blood glucose prediction based on high order and fractional low order statistics, the system comprising:
the data acquisition module is used for acquiring blood sugar data of a subject, wherein the blood sugar data comprises near-infrared blood sugar index data of the subject and an actual blood sugar value of a human body synchronously corresponding to the near-infrared blood sugar index data, and the acquired near-infrared blood sugar index data is set as a signal with noise;
the data decomposition module is used for decomposing the signal with noise to obtain a de-noising signal;
the characteristic value extraction module is used for respectively extracting the characteristic values of the high-order statistic and the fractional low-order statistic of the de-noising signal and constructing a characteristic matrix containing a plurality of characteristic values;
the dividing module is used for dividing the characteristic values in the characteristic matrix and the actual blood sugar values corresponding to the characteristic values in the characteristic matrix into a training set and a verification set;
the blood sugar prediction model building and processing module is used for building a blood sugar prediction model, training the blood sugar prediction model by utilizing a training set, and evaluating the blood sugar prediction model in the training process by utilizing a verification set to obtain the trained blood sugar prediction model;
and the prediction module is used for acquiring the blood sugar data to be detected, inputting the blood sugar data to be detected into the trained blood sugar prediction model and outputting the blood sugar prediction result of the blood sugar data to be detected.
Preferably, in the eigenvalue extraction module, the higher order statistic includes a skewness of a third order statistic and a kurtosis of a fourth order statistic, and the fractional lower order statistic includes a characteristic index α and a dispersion coefficient γ.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
the invention provides a blood sugar prediction method and a blood sugar prediction system based on high-order and fractional low-order statistics, in the technical scheme, near-infrared blood sugar index data serving as noisy signals are firstly obtained, the noisy signals are decomposed to obtain de-noised signals, the noisy signals can be effectively pre-de-noised for the purpose of providing a reliable signal data source, then the high-order statistics and the fractional low-order statistics of the de-noised signals are extracted, more detailed information of non-Gaussian near-infrared blood sugar signals is extracted, so that the characteristics of the non-Gaussian near-infrared blood sugar signals are fully reflected, a blood sugar prediction model is further constructed, the blood sugar prediction model is trained by utilizing a training set in a data set consisting of the extracted feature values and the obtained actual blood sugar values, finally, the collected blood sugar data to be detected is detected by utilizing the trained blood sugar prediction model, the blood sugar prediction result of the blood sugar data to be detected is obtained, and the accuracy of the blood sugar data prediction is effectively improved.
Drawings
FIG. 1 is a flow chart illustrating a method for predicting blood glucose based on high-order and fractional-low-order statistics as proposed in an embodiment of the present invention;
FIG. 2 is a schematic diagram of a denoising process proposed in an embodiment of the present invention;
FIG. 3 is a block diagram of a blood glucose prediction system based on higher order and fractional lower order statistics as contemplated by an embodiment of the present invention.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
for better illustration of the embodiment, some parts in the drawings may be omitted, enlarged or reduced, and do not represent actual dimensions, and the description of directions of the parts such as "up" and "down" is not limited to the patent;
it will be understood by those skilled in the art that certain well-known descriptions of the figures may be omitted;
the positional relationships depicted in the drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
the technical solution of the present invention is further described with reference to the drawings and the embodiments.
Example 1
As shown in fig. 1, the present embodiment provides a method for predicting blood glucose based on high-order and score low-order statistics, which includes the following steps:
s1, obtaining blood glucose data of a subject, wherein the blood glucose data comprises near-infrared blood glucose index data of the subject and an actual blood glucose value of the subject synchronously corresponding to the near-infrared blood glucose index data, and setting the obtained near-infrared blood glucose index data as a noisy signal;
in step S1, the near-infrared blood glucose index data is represented as [ x ] 1 ，x 2 ，…，x n ]And n represents the total number of the collected near-infrared blood glucose index data, the near-infrared blood glucose index data is obtained noninvasively by using a near-infrared LED sensor, the actual blood glucose value is obtained by using a glucometer, namely, a finger of a subject is poked to obtain a blood sample, and the blood glucose meter is used for detecting the blood sample to obtain the actual blood glucose value of the subject.
S2, decomposing the signal with noise to obtain a de-noising signal;
in step S2, let the noisy signal be x (n), see fig. 2, the specific steps of decomposing the noisy signal x (n) are:
s21, carrying out EMD on the noisy signal x (n) to obtain k IMF data components IMF k And 1 remainder data component R N Wherein k is more than or equal to 2;
s22. From k IMF data components IMF k M IMF data components are selected, k is larger than M, SSA decomposition is carried out on each IMF data component in the M IMF data components, an SVD component corresponding to each IMF data component in the M IMF data components is obtained, and a subset { XI } corresponding to the SVD component is constructed according to the obtained SVD component;
before performing SSA decomposition on each of the M IMF data components in S22, dividing each of the M IMF data components, specifically including: first, the noise energy E of each of M IMF data components is collected nk Since the first IMF data component is located in the highest frequency band, occupies the widest frequency band and contains the largest noise, the energy E of the noise contained in the first IMF data component is usually collected from the second IMF data component nk (ii) a Each IMF in the IMF data components is then computedThe specific calculation expression of the denoising energy of the data component is as follows:
E‘ xk ＝E xk -E nk
wherein E is k Representing the energy, E ', of each of the IMF data components' xk Representing a denoising energy for each of the IMF data components;
calculating E' xk And E k Ratio R k The specific calculation expression is:
according to the ratio R k Each of the M IMF data components is divided.
S23, corresponding subset { X ] of each SVD component I Reconstructing to obtain M groups of reconstruction data G K ；
S24, reconstructing M groups of data G K With remainder data component R n Adding up and outputting de-noised signal x i 。
S3, respectively extracting characteristic values of the high-order statistic and the fraction low-order statistic of the de-noising signal, and constructing a characteristic matrix containing a plurality of characteristic values;
s4, dividing the characteristic values in the characteristic matrix and the actual blood sugar values corresponding to the characteristic values in the characteristic matrix into a training set and a verification set;
s5, constructing a blood sugar prediction model, training the blood sugar prediction model by using a training set, and evaluating the blood sugar prediction model in the training process by using a verification set to obtain the trained blood sugar prediction model;
s6, collecting blood sugar data to be detected, inputting the blood sugar data to be detected into the trained blood sugar prediction model, and outputting a blood sugar prediction result of the blood sugar data to be detected.
Example 2
Referring to fig. 1, in step S3, the feature value of the high-order statistic of the denoised signal includes a skewness of the third-order statistic and a kurtosis of the fourth-order statistic, and a specific calculation expression of the skewness of the third-order statistic is as follows:
wherein SK i Representing the skewness of the third-order statistic corresponding to the ith preprocessing data, n representing the length of the near-infrared blood sugar original data, mu i3 Represents the third-order central moment, s, corresponding to the ith preprocessed data i Represents the sample standard deviation;
μ i3 the specific calculation expression of (2) is:
wherein the content of the first and second substances,
representing a denoised signal x i The average value of (a) of (b),
the specific calculation expression of (2) is:
s i the specific calculation expression of (2) is:
the specific calculation expression of the four-order statistic kurtosis is as follows:
wherein, K j Represents the kurtosis, mu, of the fourth order statistic corresponding to the ith pre-processed data i2 Represents the second corresponding to the ith preprocessing dataCentral moment of order, mu i4 And representing the fourth-order central moment corresponding to the ith preprocessing data.
In step S3, the feature value of the fractional low order statistic includes a feature index α and a dispersion coefficient γ, the feature index α and the dispersion coefficient γ representing the fractional low order statistic are extracted from the denoised signal by using a logarithm method, and the specific extraction steps of the feature index α and the dispersion coefficient γ are as follows:
s31, setting x i Is a group of random variables whose characteristic index alpha and symmetry coefficient beta are both 0 and stably distributed, and for alpha, x i Deaveraging process
Obtaining a finite negative moment E { | x of an alpha-stable random variable with a position parameter a =0 i | p H, a negative order moment E { | x i | p The following conditions are satisfied:
wherein p represents a fractional order; c (p, α) represents a function of α, p, and a random variable x i Regardless, the specific computational expression of C (p, α) is:
wherein Γ (·) represents a gamma function;
for E { | x i | p Performing equivalent transformation, wherein a specific expression is as follows:
let Y = log | x i | pair
And (3) carrying out conversion, wherein the specific conversion expression is as follows:
wherein E { ep Y Denotes the first eigenfunction generation of Y, E { E } pY The specific calculation expression of is:
E(Y k ) The specific calculation expression of (2) is:
s32, calculating the mean value E (Y) and the variance Var (Y) of Y to obtain values of E (Y) and Var (Y);
s33, carrying out equivalent transformation on the variance Var (Y), wherein the specific expression is as follows:
substituting the Var (Y) value obtained in the step S32 into the formula in the step S33 to solve to obtain a value of the characteristic index alpha;
s34, carrying out equivalent transformation on the mean value E (Y), wherein the specific expression is as follows:
wherein, C e Denotes the Euler constant, C e ＝0.577212566；
And substituting the alpha value obtained in the step S33 into the formula in the step S34 for solving to obtain the value of the dispersion coefficient gamma.
In step S5, the method for constructing the blood glucose prediction model is a random forest algorithm.
Example 3
Referring to fig. 3, the present embodiment proposes a blood glucose prediction system based on high-order and score low-order statistics, the system comprising:
the data acquisition module 11 is configured to acquire blood glucose data of a subject, where the blood glucose data includes near-infrared blood glucose indicator data of the subject and an actual blood glucose value of a human body acquired synchronously and correspondingly to the near-infrared blood glucose indicator data, and the acquired near-infrared blood glucose indicator data is set as a noisy signal;
the data decomposition module 12 is configured to decompose the noisy signal to obtain a denoised signal;
the eigenvalue extraction module 13 is configured to extract eigenvalues of the high-order statistic and the fractional low-order statistic of the denoised signal, and construct an eigenvalue matrix including a plurality of eigenvalues;
in the eigenvalue extraction module, the high-order statistic comprises skewness of a third-order statistic and kurtosis of a fourth-order statistic, and the fractional low-order statistic comprises a characteristic index alpha and a dispersion coefficient gamma.
A dividing module 14, configured to divide the eigenvalue in the feature matrix and the actual blood glucose value corresponding to the eigenvalue in the feature matrix into a training set and a verification set;
the blood sugar prediction model construction processing module 15 is used for constructing a blood sugar prediction model, training the blood sugar prediction model by using a training set, and evaluating the blood sugar prediction model in the training process by using a verification set to obtain the trained blood sugar prediction model;
and the prediction module 16 is used for acquiring the blood sugar data to be detected, inputting the blood sugar data to be detected into the trained blood sugar prediction model, and outputting a blood sugar prediction result of the blood sugar data to be detected.
In this embodiment, first, near-infrared blood glucose index data serving as a noisy signal is obtained, the noisy signal is decomposed to obtain a denoised signal, effective pre-denoising processing of the noisy signal is guaranteed at the time of the purpose, a reliable signal data source is provided, then, a high-order statistic and a feature value of a score low-order statistic of the denoised signal are extracted, more detailed information of a non-gaussian near-infrared blood glucose signal is extracted, so that the features of the non-gaussian near-infrared blood glucose signal are fully reflected, a blood glucose prediction model is further constructed, the blood glucose prediction model is trained by using a training set in a data set consisting of the extracted feature value and an obtained actual blood glucose value, finally, the collected blood glucose data to be detected is detected by using the trained blood glucose prediction model, a blood glucose prediction result of the blood glucose data to be detected is obtained, and the accuracy of blood glucose data prediction is effectively improved.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (10)
1. A blood glucose prediction method based on high-order and fractional-low-order statistics is characterized by comprising the following steps:
s1, obtaining blood glucose data of a subject, wherein the blood glucose data comprises near-infrared blood glucose index data of the subject and an actual blood glucose value of the subject synchronously corresponding to the near-infrared blood glucose index data, and setting the obtained near-infrared blood glucose index data as a noisy signal;
s2, decomposing the signal with noise to obtain a de-noising signal;
s3, respectively extracting characteristic values of the high order statistic and the fraction low order statistic of the de-noised signal, and constructing a characteristic matrix containing a plurality of characteristic values;
s4, dividing the characteristic values in the characteristic matrix and the actual blood sugar values corresponding to the characteristic values in the characteristic matrix into a training set and a verification set;
s5, constructing a blood sugar prediction model, training the blood sugar prediction model by using a training set, and evaluating the blood sugar prediction model in the training process by using a verification set to obtain the trained blood sugar prediction model;
s6, collecting blood sugar data to be detected, inputting the blood sugar data to be detected into the trained blood sugar prediction model, and outputting a blood sugar prediction result of the blood sugar data to be detected.
2. The method of claim 1 wherein in step S1, the near-infrared glycemic index data is expressed as [ x [ ] 1 ，x 2 ，…，x n ]And n represents the total number of the collected near-infrared blood sugar index data, the near-infrared blood sugar index data is obtained noninvasively by using a near-infrared LED sensor, and the actual blood sugar value is obtained invasively by using a blood sugar meter.
3. The method of claim 2, wherein in step S2, the noisy signal is set to x (n), and the specific step of decomposing the noisy signal x (n) is:
s21, carrying out EMD on the noisy signal x (n) to obtain k IMF data components and 1 remainder data component, wherein k is more than or equal to 2;
s22, M IMF data components are selected from the k IMF data components, k is larger than M, SSA decomposition is carried out on each IMF data component in the M IMF data components to obtain an SVD component corresponding to each IMF data component in the M IMF data components, and a subset { X corresponding to the SVD component is constructed according to the obtained SVD component I }；
S23, corresponding subset { X ] of each SVD component I Reconstructing to obtain M groups of reconstruction data;
s24, accumulating M groups of reconstruction data and remainder data components, and outputting a de-noising signal x i 。
4. The method of claim 3, wherein each of the M IMF data components is partitioned before SSA decomposition of each of the M IMF data components, the partitioning comprising: first, the noise energy E of each of M IMF data components is collected nk Then calculating IMF data componentsThe specific calculation expression of the denoising energy of each IMF data component in (1) is:
E‘ xk ＝E xk -E nk
wherein E is k Representing the energy, E ', of each of the IMF data components' xk Representing a denoising energy for each of the IMF data components;
calculating E' xk And E k Ratio R k The specific calculation expression is:
according to the ratio R k Each of the M IMF data components is divided.
5. The method of claim 3, wherein the eigenvalues of the higher order statistic of the de-noised signal include a skewness of the third order statistic and a kurtosis of the fourth order statistic, and wherein the skewness of the third order statistic is specified by the formula:
wherein SK i Representing the skewness of the third-order statistic corresponding to the ith preprocessing data, n representing the length of the near-infrared blood sugar original data, mu i3 Represents the third-order central moment, s, corresponding to the ith preprocessing data i Represents the sample standard deviation;
μ i3 the specific calculation expression of (2) is:
wherein the content of the first and second substances,
representing a denoised signal x i The average value of (a) of (b),
the specific calculation expression of (2) is:
s i the specific calculation expression of (2) is:
the specific calculation expression of the fourth order statistic kurtosis is as follows:
wherein, K i Represents the kurtosis, mu, of the fourth order statistic corresponding to the ith pre-processed data i2 Represents the second central moment, mu, corresponding to the ith pre-processed data i4 And representing the fourth-order central moment corresponding to the ith preprocessing data.
6. The method of claim 5 wherein the characteristic values of the fractionally low order statistic include a characteristic index α and a dispersion coefficient γ.
7. The method for predicting blood glucose based on higher-order and fractional-lower-order statistics of claim 6, wherein the denoising signal is subjected to the extraction of a feature index α and a dispersion coefficient γ representing the fractional-lower-order statistics by a logarithm method, and the specific extraction steps of the feature index α and the dispersion coefficient γ are as follows:
s31, setting x i Is a set of characteristic indexes alphaAnd x with symmetrical coefficient beta of 0 stable distribution i Deaveraging process
Obtaining a finite negative moment E { | x of an alpha-stable random variable with a position parameter a =0 i | p H, a negative order moment E { | x i | p The following conditions are satisfied:
wherein p represents a fractional order; c (p, α) represents a function of α, p, and a random variable x i Regardless, the specific computational expression of C (p, α) is:
wherein Γ (·) represents a gamma function;
for E { | x i | p Carrying out equivalent transformation, wherein a specific expression is as follows:
let Y = log | x i I, to
And (3) carrying out conversion, wherein a specific conversion expression is as follows:
wherein, E { E } pY Denotes the first eigenfunction generation of Y, E { E } pY The specific calculation expression of the method is as follows:
E(Y k ) The specific calculation expression of (2) is:
s32, calculating the mean value E (Y) and the variance Var (Y) of Y to obtain values of E (Y) and Var (Y);
s33, carrying out equivalent transformation on the variance Var (Y), wherein the specific expression is as follows:
substituting the Var (Y) value obtained in the step S32 into the formula in the step S33 to solve to obtain a value of the characteristic index alpha;
s34, performing equivalent transformation on the average value E (Y), wherein the specific expression is as follows:
wherein, C e Represents an Euler constant;
and substituting the alpha value obtained in the step S33 into the formula in the step S34 to solve to obtain the value of the dispersion coefficient gamma.
8. The method for predicting blood glucose based on higher-order and lower-order statistics of score as claimed in claim 1, wherein in step S5, the method for constructing the blood glucose prediction model is a random forest algorithm.
9. A system for blood glucose prediction based on higher order and fractional lower order statistics, the system comprising:
the data acquisition module is used for acquiring blood sugar data of a subject, wherein the blood sugar data comprises near-infrared blood sugar index data of the subject and actual blood sugar value of a human body synchronously and correspondingly acquired with the near-infrared blood sugar index data, and the acquired near-infrared blood sugar index data is set as a signal with noise;
the data decomposition module is used for decomposing the signal with noise to obtain a de-noising signal;
the characteristic value extraction module is used for respectively extracting the characteristic values of the high-order statistic and the fractional low-order statistic of the de-noising signal and constructing a characteristic matrix containing a plurality of characteristic values;
the dividing module is used for dividing the characteristic values in the characteristic matrix and the actual blood sugar values corresponding to the characteristic values in the characteristic matrix into a training set and a verification set;
the blood sugar prediction model construction processing module is used for constructing a blood sugar prediction model, training the blood sugar prediction model by using a training set, and evaluating the blood sugar prediction model in the training process by using a verification set to obtain the trained blood sugar prediction model;
and the prediction module is used for acquiring the blood sugar data to be detected, inputting the blood sugar data to be detected into the trained blood sugar prediction model and outputting the blood sugar prediction result of the blood sugar data to be detected.
10. The system of claim 9, wherein the higher order statistics comprise a skewness of a third order statistic and a kurtosis of a fourth order statistic, and the fractional lower order statistics comprise a feature index α and a dispersion coefficient γ.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211274466.1A CN115633957A (en) | 2022-10-18 | 2022-10-18 | Blood glucose prediction method and system based on high-order and fraction low-order statistics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211274466.1A CN115633957A (en) | 2022-10-18 | 2022-10-18 | Blood glucose prediction method and system based on high-order and fraction low-order statistics |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115633957A true CN115633957A (en) | 2023-01-24 |
Family
ID=84943788
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211274466.1A Pending CN115633957A (en) | 2022-10-18 | 2022-10-18 | Blood glucose prediction method and system based on high-order and fraction low-order statistics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115633957A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115944293A (en) * | 2023-03-15 | 2023-04-11 | 汶上县人民医院 | Neural network-based hemoglobin level prediction system for kidney dialysis |
-
2022
- 2022-10-18 CN CN202211274466.1A patent/CN115633957A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115944293A (en) * | 2023-03-15 | 2023-04-11 | 汶上县人民医院 | Neural network-based hemoglobin level prediction system for kidney dialysis |
| CN115944293B (en) * | 2023-03-15 | 2023-05-16 | 汶上县人民医院 | Neural network-based hemoglobin level prediction system for kidney dialysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Arrhythmia classification algorithm based on multi-head self-attention mechanism | |
| Wei et al. | The research of sleep staging based on single-lead electrocardiogram and deep neural network | |
| CN111310570B (en) | Electroencephalogram signal emotion recognition method and system based on VMD and WPD | |
| Wójcikowski et al. | Photoplethysmographic time-domain heart rate measurement algorithm for resource-constrained wearable devices and its implementation | |
| EP3698708A1 (en) | Ecg information processing method and ecg workstation | |
| Tsai et al. | Diabetes care in motion: Blood glucose estimation using wearable devices | |
| CN113509186B (en) | ECG classification system and method based on deep convolutional neural network | |
| Roy et al. | On-device reliability assessment and prediction of missing photoplethysmographic data using deep neural networks | |
| CN115633957A (en) | Blood glucose prediction method and system based on high-order and fraction low-order statistics | |
| CN114420301B (en) | Method, system and storage medium for predicting blood glucose based on segmented domain RF modeling | |
| CN114548158A (en) | Data processing method for blood sugar prediction | |
| Desquins et al. | A survey of photoplethysmography and imaging photoplethysmography quality assessment methods | |
| CN116250834A (en) | Noninvasive personalized dynamic blood glucose trend monitoring and early warning method and device | |
| Philip et al. | Continous monitoring of blood glucose using photophlythesmograph signal | |
| CN115089179A (en) | Psychological emotion insights analysis method and system | |
| Wang et al. | Ppg signal reconstruction using deep convolutional generative adversarial network | |
| CN112120711B (en) | Noninvasive diabetes prediction system and method based on photoplethysmography pulse waves | |
| Abbasi et al. | Multiple contaminant biosignal quality analysis for electrocardiography | |
| US20200330040A1 (en) | Method and system for detecting noise in vital sign signal | |
| CN114403866B (en) | Noninvasive blood sugar prediction device based on near-infrared light wavelength conversion | |
| CN116451110A (en) | Blood glucose prediction model construction method based on signal energy characteristics and pulse period | |
| WO2023095171A1 (en) | Non-invasive sensor fusion and ai technology for human vital measurement | |
| Gecili et al. | Functional data analysis and prediction tools for continuous glucose-monitoring studies | |
| JP2023116614A (en) | Method and system for determining cardiovascular parameters | |
| Zhang et al. | Non-invasive blood glucose estimation using near-infrared spectroscopy based on SVR |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |