CN105550502A - Data processing method based on health monitoring - Google Patents

Abstract

The invention discloses a data processing method based on health monitoring. According to the method, firstly, measurement accuracy is ensured through a multi-touch multi-sensing data collection method; secondly, multiple groups of physiological data are screened according to the Grubbs criterion so as to remove collected gross error data; and finally collected data are fused through combining a batch estimation algorithm and a least squares algorithm so as to reduce influences of different measurement precisions of different sensors on results. According to the method of the invention, the physiological information of a user can be obtained accurately so as to provide valid basis for real time detection of a health monitoring system.

Description

A kind of data processing method based on health monitoring

Technical field

The invention belongs to portable medical technical field, particularly relate to a kind of data processing method based on health monitoring.

Background technology

Wearable technology is the brand new ideas grown up recent years, and it is widely used in the aspect such as clinical monitoring, family health care, emergent rescue, special population monitoring, Psychological Assessment, athletic training.And wearable technology is exactly the clothing of daily physiologic information detection technique and the daily wearing of people, ornaments are merged mutually in the application of medical domain, thus is not affecting user's daily life and under ensureing comfortable prerequisite, the characteristic parameter of user detected.But, in the physiological parameter detected, due to various environmental factor, many rough error parameters can be mixed.And in image data process, the angle of measurement is single, can not ensure the accuracy of data.

In addition, in some wearable devices, adopt single data fusion method, the physiologic information gathered is processed, when sensor experiences failure, the data that gather of Timeliness coverage can not there is error, finally cause the physiologic information that accurately can not reflect user.

Summary of the invention

In order to overcome above-mentioned the deficiencies in the prior art, the present invention proposes a kind of data processing method based on health monitoring.First the method utilizes the method for the many sensings of multiconductor to ensure that the accuracy of measurement, then screens the physiological data gathered, and merges, and final acquisition one group accurately can reflect the data of user's physiologic information.On the one hand, the multiconductor many sensing measurements mode in the method, learning process has good theoretical performance and ensures, accuracy of measurement is high; On the other hand, it is the reference scheme that wearable watch-dog provides a kind of data processing, has good application prospect.

In order to solve the problems of the technologies described above, the technical solution used in the present invention is:

Based on a data processing method for health monitoring, comprise the steps:

Step 1: the M class physiological data being gathered person to be measured by sensor, uses L _mindividual m+1 class physiological data sensor, gather the m+1 class physiological data of person to be measured, its detailed process is:

Step 1-1, gather for the 1st time of physiological data, each sensor is not measured in the same time respectively the N number of of same contact, obtains N group physiological data, often organizes data and is expressed as:

$A_{1k}^m\[Q\]=\[a_{1k}^{m0},a_{1k}^{m1},\mathrm{...}{a}_{1k}^{mi},\mathrm{...},a_{1k}^{m(Q-1)}\];$

Step 1-2, gather for the 2nd time of physiological data, each sensor is respectively at L _mthe N number of of individual contact does not measure in the same time, obtains N group physiological data, often organizes data and is expressed as:

$A_{2k}^m\[Q\]=\[a_{2k}^{m0},a_{2k}^{m1},\mathrm{...},a_{2k}^{mi},\mathrm{...},a_{2k}^{m(Q-1)}\];$

Wherein, M is natural number, according to actual conditions value; M is the label of physiology data type, 0≤m≤M-1; L _mrepresent the quantity of m+1 class physiological data sensor; N is the group number of each image data, and N is L _mintegral multiple; K is the label often organizing data, 0≤k≤N-1; Q is the total number of measured value often organizing data; J represents the label of data acquisition order, j ∈ { 1,2}; I is the label of middle element, 0≤i≤Q-1, for in the i-th+1 element;

Step 2, to m+1 class physiological data element, utilize Grobus criterion carry out first time screening, concrete steps are as follows:

Step 2-1, calculates middle element arithmetic mean $\stackrel{\‾}{a_{jk}^m}=\frac{1}{Q}\underset{i=0}{\overset{Q-1}{\Σ}}{a}_{jk}^{mi},$ And then calculate its standard deviation

${\mathrm{\σ}}_{jk}^m=\sqrt{\frac{\underset{i=0}{\overset{Q-1}{\Σ}}{(a_{jk}^{mi}-\stackrel{\‾}{a_{jk}^m})}^2}{Q-1}};$

Step 2-2, by Ge Luobusi tables of critical values, chooses a critical value d (Q, θ), and the size of d (Q, θ) is determined by the value of Q and significance θ; Wherein significance θ gets 0.05;

Step 2-3, the g (i) that distributed by Ge Luobusi carries out comparing for Q time with d (Q, θ), wherein, if g (i)>=d (Q, θ), be in gross error, reject; Result after rejecting is designated as wherein, Q' is after middle element screens, total number of remaining element, Q'≤Q;

Step 3, to the data after first time screening element, carry out first time data fusion, adopt the method in batches estimated, calculate the fusion value often organizing data concrete formula is as follows:

$R_{jk}^m=\frac{{\mathrm{\σ}}_e^2\stackrel{\‾}{f}+{\mathrm{\σ}}_f^2\stackrel{\‾}{e}}{{\mathrm{\σ}}_e^2+{\mathrm{\σ}}_f^2}$

Wherein, for middle element superscript the mean value of element set for middle element superscript the mean value of element set, for middle element superscript the variance of element set, for middle element superscript the variance of element set; expression is not more than maximum integer;

Step 4, for different j, k, each group data fusion value difference, represents all fusion values set ψ: wherein, v is the label of element in described set ψ, 0≤v≤2N-1; Again utilize Grobus criterion, carry out programmed screening to element in set ψ, reject wherein gross error, the result after screening is designated as set for described set total number of middle element, 0≤S ^m≤ 2N; for described set middle element numerals, described set in individual element with represent;

Step 5, adopts least square method of weighting, to described set element carry out second time data fusion, be specially:

By described set element according to formula:

${\tilde{R}}^m=\underset{\stackrel{\‾}{v}=0}{\overset{S^m-1}{\Σ}}{w}_{\stackrel{\‾}{v}}{r}_{\stackrel{\‾}{v}}$

Merge, wherein, for last fusion value, for weights coefficient, computing formula be:

$w_{\stackrel{\‾}{v}}=\frac{1}{{\mathrm{\σ}}_{\stackrel{\‾}{v}}\underset{\stackrel{\‾}{v}=0}{\overset{S^m-1}{\Σ}}\frac{1}{{\mathrm{\σ}}_{\stackrel{\‾}{v}}}}$

And weights coefficient meets: wherein, for the standard deviation of element in corresponding array; Last fusion value as the end value after the process of m+1 class physiological data.

Beneficial effect: a kind of data processing method based on health monitoring that the present invention proposes, first the method utilizes the collecting method of the many sensings of multiconductor, ensure that the accuracy of measurement.Secondly, to the many groups physiological data collected, screen according to Grobus criterion, to remove the rough error data collected.Finally, by being combined algorithm for estimating and least-squares algorithm in batches, the data collected being merged, reducing different sensors because of measuring accuracy difference, on the impact that result causes.The method is conducive to the physiologic information obtaining user accurately, provides effective foundation for health monitoring system detects in real time.

Accompanying drawing explanation

Fig. 1 is that a kind of data processing method based on health monitoring performs schematic flow sheet.

Fig. 2 is the partial crit value table of Grobus criterion.

Fig. 3 is LS weighted fusion algorithm algorithm model figure.

Embodiment

In order to a kind of data processing method based on health monitoring that more detailed description the present invention proposes, by reference to the accompanying drawings, illustrate as follows:

Fig. 1 shows a kind of data processing method based on health monitoring and performs schematic flow sheet.This method mainly comprises three contents: one is the collecting method utilizing the many sensings of multiconductor, gathers user's physiologic information; Two is the many groups physiological datas to collecting, and carries out twice screening in front and back; Three is be combined algorithm for estimating and least-squares algorithm in batches to merge the data after screening.

1. gathered the M class physiological data of person to be measured by sensor, M gets 2, and physiological data can be specially temperature data and blood pressure data.Use infrared sensor and thermistor (temperature) sensor to gather the temperature data of person to be measured, use resistive pressure sensor and capacitance pressure transducer, to gather the blood pressure data of person to be measured, its detailed process is:

1st step: first time single contact collection, each sensor is not measured in the same time respectively four of same contact, obtains 4 groups of physiological datas, often organizes data and be expressed as:

2nd step: second time multi-contact data collection, each sensor is not measured in the same time at four of four contacts respectively, obtains 4 groups of physiological datas, often organizes data and is expressed as:

M is the label of physiology data type, because only take temperature and blood pressure parameter, so 0≤m≤1; Namely the first kind is temperature data, and Equations of The Second Kind is blood pressure data.K is the label often organizing data, 0≤k≤3; J represents the label of data acquisition order, j ∈ { 1,2}; I is the label of middle element, 0≤i≤3, in the i-th+1 element.

It should be noted that, this enforcement gathers at every turn and only gathers 4 groups of data, often organize data and contain 4 measured values, also more multi-group data can be gathered according to actual conditions, gather from the physiological data of more angle to user, be that the accuracy of data will be higher because the data gathered are more, the health monitoring of user could accurately.

2. pair many groups collected physiological data, carries out first time screening:

1st step: differentiation and the rejecting of the data separate collected " Grobus criterion " being carried out to gross error; Concrete grammar is:

Calculate middle element arithmetic mean and then calculating standard deviation:

${\mathrm{\σ}}_{jk}^m=\sqrt{\frac{\underset{i=0}{\overset{3}{\Σ}}{(a_{jk}^{mi}-\stackrel{\‾}{a_{jk}^m})}^2}{3}}---\left(1\right)$

2nd step: by the tables of critical values of Fig. 2 Grobus criterion, choose a critical value d (Q, α), its size is determined by the value of Q and α; Wherein significance α gets 0.05; Table look-up known d (4,0.05)=1.46, according to Ge Luobusi distribution g (i):

$g\left(i\right)=\frac{|a_{jk}^{mi}-\stackrel{\‾}{a_{jk}^m}|}{{\mathrm{\σ}}_{jk}^m}---\left(2\right)$

G (i) and d (4,0.05) are carried out 4 times compare, if g (i)>=d (4,0.05), be in gross error, reject; Result after screening is designated as wherein, Q' is after middle element screens, total number of remaining element.

3. adopt algorithm for estimating in batches to carry out first time fusion to each group of data, detailed process is as follows:

Adopt the method in batches estimated, to often organizing data respectively according to formula:

$R_{jk}^m=\frac{{\mathrm{\σ}}_e^2\stackrel{\‾}{f}+{\mathrm{\σ}}_f^2\stackrel{\‾}{e}}{{\mathrm{\σ}}_e^2+{\mathrm{\σ}}_f^2}---\left(3\right)$

Calculate the fusion value of data, wherein, for fusion value, for middle element superscript the mean value of element set, for middle element superscript the mean value of element set, for middle element superscript the variance of element set, for middle element superscript the variance of element set;

4: for different j, k, each group data fusion value difference, represents all fusion values set ψ: wherein, v is the label of element in described set ψ, 0≤v≤7; Again utilize Grobus criterion, carry out programmed screening to element in set ψ, reject wherein gross error, the result after screening is designated as set s ^mfor described set total number of middle element, 0≤S ^m≤ 8; for described set middle element numerals, described set in individual element with represent;

It should be noted that, why carry out differentiation and the rejecting of twice gross error, because primary process is just carried out for single-sensor, if one or several sensor degradation wherein, the data obtained is to sensor it doesn't matter itself gross error, and can fusion results be drawn, secondary data processing just can weed out having the sensing data of damage.

5: on the basis of previous step, adopt least square method of weighting, to set middle element carries out second time data fusion, and detailed process is:

By described set element according to formula:

${\tilde{R}}^m=\underset{\stackrel{\‾}{v}=0}{\overset{S^m-1}{\Σ}}{w}_{\stackrel{\‾}{v}}{r}_{\stackrel{\‾}{v}}---\left(4\right)$

Merge, wherein, for last fusion value, for weights coefficient, computing formula be::

$w_{\stackrel{\‾}{v}}=\frac{1}{{\mathrm{\σ}}_{\stackrel{\‾}{v}}\underset{\stackrel{\‾}{v}=0}{\overset{S^m-1}{\Σ}}\frac{1}{{\mathrm{\σ}}_{\stackrel{\‾}{v}}}}---\left(5\right)$

And weights coefficient meets: for the standard deviation of element in corresponding array; Last fusion value as the end value after the process of m+1 class physiological data, this example is the last processing costs of body temperature, blood pressure.

It should be noted that, LS weighted fusion algorithm algorithm is adopted to carry out secondary fusion to data, because merge between the sensor measurement data estimating to be applicable to equally accurate in batches, owing to using many sensing measurements data, the precision of sensor is not exclusively the same, in order to make the result of fusion more excellent, so use minimum weight blending algorithm to carry out secondary fusion to data.Finally, the data of representative of consumer physiological parameter are drawn.

Comprehensive foregoing teachings, the execution flow process that can obtain this method is:

The first step, utilizes the collecting method of the many sensings of multiconductor, gathers user's physiologic information.

Second step, the many groups collected physiological data, carries out first time data screening according to Grobus criterion.

3rd step, to the often group data after screening, utilizes algorithm for estimating in batches to carry out first time data fusion.

4th step, to the result after fusion, according to the grouping of measurement parameter classification, carries out secondary data screening in group.

5th step, to the data after postsearch screening, adopts least square method to carry out the fusion of data secondary, the final measured value of last result representative of consumer physiological parameter.

Claims (1)

1. based on a data processing method for health monitoring, it is characterized in that, comprise the steps:

Step 1: the M class physiological data being gathered person to be measured by sensor, uses L _mindividual m+1 class physiological data sensor, gather the m+1 class physiological data of person to be measured, its detailed process is:

Step 1-1, gather for the 1st time of physiological data, each sensor is not measured in the same time respectively the N number of of same contact, obtains N group physiological data, often organizes data and is expressed as:

Step 1-2, gather for the 2nd time of physiological data, each sensor is respectively at L _mthe N number of of individual contact does not measure in the same time, obtains N group physiological data, often organizes data and is expressed as:

Wherein, M is natural number, according to actual conditions value; M is the label of physiology data type, 0≤m≤M-1; L _mrepresent the quantity of m+1 class physiological data sensor; N is the group number of each image data, and N is L _mintegral multiple; K is the label often organizing data, 0≤k≤N-1; Q is the total number of measured value often organizing data; J represents the label of data acquisition order, j ∈ { 1,2}; I is the label of middle element, 0≤i≤Q-1, for in the i-th+1 element;

Step 2, to m+1 class physiological data element, utilize Grobus criterion carry out first time screening, concrete steps are as follows:

Step 2-1, calculates middle element arithmetic mean and then calculate its standard deviation

Step 2-2, by Ge Luobusi tables of critical values, chooses a critical value d (Q, θ), and the size of d (Q, θ) is determined by the value of Q and significance θ; Wherein significance θ gets 0.05;

Step 2-3, the g (i) that distributed by Ge Luobusi carries out comparing for Q time with d (Q, θ), wherein, if g (i)>=d (Q, θ), be in gross error, reject; Result after rejecting is designated as wherein, Q' is after middle element screens, total number of remaining element, Q'≤Q;

Step 3, to the data after first time screening element, carry out first time data fusion, adopt the method in batches estimated, calculate the fusion value often organizing data concrete formula is as follows:

Wherein, for middle element superscript the mean value of element set, for middle element superscript the mean value of element set, for middle element superscript the variance of element set, for middle element superscript the variance of element set; expression is not more than maximum integer;

Step 4, for different j, k, each group data fusion value difference, represents all fusion values set ψ: wherein, v is the label of element in described set ψ, 0≤v≤2N-1; Again utilize Grobus criterion, carry out programmed screening to element in set ψ, reject wherein gross error, the result after screening is designated as set s ^mfor described set total number of middle element, 0≤S ^m≤ 2N; for described set middle element numerals, ; Described set in individual element with represent;

Step 5, adopts least square method of weighting, to described set element carry out second time data fusion, be specially:

By described set element according to formula:

Merge, wherein, for last fusion value, for weights coefficient, computing formula be:

And weights coefficient meets: wherein, for the standard deviation of element in corresponding array; Last fusion value as the end value after the process of m+1 class physiological data.