CN103239239B - Fixed-amplitude dynamic spectrum data extraction method - Google Patents

Abstract

The invention discloses a fixed-amplitude dynamic spectrum data extraction method. The method includes following steps: collecting full-wave-band photoelectric volume pulse waves and performing logarithmic transformation, and choosing a preset wavelength and a preset equal amplitude in a full-wave band; obtaining a rising edge area and a falling edge area in full-wave-band logarithmic pulse waves, taking the equal amplitude as a standard to obtain amplitudes in a same sampling time period in each rising edge and fitting, and using all amplitudes obtained after fitting to form a fixed-amplitude real-time dynamic spectrum; and eliminating the fixed-amplitude real-time dynamic spectrum containing gross errors, subjecting the remaining fixed-amplitude real-time dynamic spectrum to superposition averaging to finally obtain a dynamic spectrum and output the dynamic spectrum. By the fixed-amplitude dynamic spectrum data extraction method, signal to noise ratio of the fixed-amplitude dynamic spectrum is increased; by determining the equal amplitudes of different to-be-tested positions, effective information of the dynamic spectrum is unchanged while influences of differences of long optical distances, and the signal to noise ratios of the dynamic spectrum at the different to-be-tested positions are balanced, so that an effective method is provided for improving accuracy of a dynamic spectrum noninvasive blood composition detection and calibration model.

Description

A kind of dynamic spectral data extracting method of tentering value

Technical field

The present invention relates to field of spectral analysis technology, particularly a kind of dynamic spectral data extracting method of tentering value.

Background technology

The noinvasive of blood constituent detects and has very important significance, and not only can alleviate patient's misery, the risk of infection, be also simultaneously instruct clinically prevent, the key of diagnosis and treatment of chronic diseases (for example: diabetes, hyperlipidemia and anemia etc.).The dynamic optical spectrometry of the langbobier law based on revising, utilizes the photoelectricity pulse wave under the multiple wavelength that synchronously obtain, and extracts the only absorbance spectrum of reflection pulsation arterial blood, the impact of effectively having removed its hetero-organizations such as skin, fat and skeleton.With other spectral measurement method comparisons, it is convenient to realize, and can overcome the impact of individual variation and measuring condition.But in the photoelectricity volume pulsation wave gathering, to account for the proportion of DC component less for AC compounent, and be accompanied by multiple interference and noise in actual samples signal, the dynamic spectrum that on pulse wave, any two points obtains is difficult to meet certainty of measurement requirement.

In prior art, conventionally adopt Frequency domain extracting method (patent of invention " method of noninvasive measurement of blood spectra and composition " publication number: CN101507607, open day: on August 19th, 2009) and single along extraction method (patent of invention " a kind of based on single method for processing dynamic spectral data along extraction method " publication number: CN101912256A, open day: on December 15th, 2010).Two kinds of methods all utilize statistical average method to improve dynamic spectrum signal to noise ratio, but for the impact of the factors such as the shake in gatherer process, contact pressure change, Frequency domain extracting method can not real time reaction and is avoided these unusual waveforms and singular value impact, list can affect by rejecting abnormalities pulse wave along rule, inhibition ability to noise is stronger, extracts dynamic spectrum more accurately.Above-mentioned two kinds of methods, although improve the signal to noise ratio of dynamic spectrum from different angles, cannot balance for the difference of dynamic spectrum signal to noise ratio between Different Individual.

In addition, Different Individual, different parts have different tremulous pulse filling degrees, therefore corresponding different optical lengths, and the spectral calibration model that this species diversity can detect blood constituent noinvasive produces adverse influence.In prior art, conventional polynary scatter correction method (MSC), normalization or net signal analytical method are proofreaied and correct the impact of optical length near infrared spectrum.

Inventor finds realizing in process of the present invention, and prior art not proposes for suppressing the impact of optical length, when optical length information is proofreaied and correct, has also correspondingly lost part effective information.

Summary of the invention

The dynamic spectral data extracting method that the invention provides a kind of tentering value, this method solves the problem of dynamic spectrum optical length between current Different Individual, avoids the loss of effective information, described below:

A dynamic spectral data extracting method for tentering value, said method comprising the steps of:

(1) gather all band photoelectricity volume pulsation wave make logarithmic transformation, select the preset wavelength in all band, the amplitude p such as default ^λ;

(2) obtain rising edge and the trailing edge region in described all band logarithm pulse wave, with the amplitude p such as described ^λin each rising edge, obtain amplitude the matching with the sampling period for standard, the amplitude that all matchings obtain forms a tentering value Real-time and Dynamic spectrum;

(3) reject the tentering value Real-time and Dynamic spectrum that contains gross error, residue tentering value Real-time and Dynamic spectrum superposed average is obtained to final dynamic spectrum output.

The amplitude p such as described collection all band photoelectricity volume pulsation wave is also made logarithmic transformation, selects the preset wavelength in all band, default ^λstep be specially:

The described all band photoelectricity volume pulsation wave of at least one pulse wave cycle of synchronous acquisition detected part;

Described all band photoelectricity volume pulsation wave is made to logarithmic transformation, obtain described all band logarithm pulse wave;

Preset wavelength λ in selected all band, on preset wavelength λ, the amplitude such as default described is designated as p ^λ.

Described rising edge and the trailing edge region obtaining in described all band logarithm pulse wave, with the amplitude p such as described ^λin each rising edge, obtain amplitude the matching with the sampling period for standard, the step that the amplitude that all matchings obtain forms a tentering value Real-time and Dynamic spectrum is specially:

Obtain the logarithm pulse wave that is greater than average amplitude in described all band logarithm pulse wave, described logarithm pulse wave superposed average is obtained to pulse wave template;

Search all peak points and the valley point of described pulse wave template, by described peak point and described valley point, described all band logarithm pulse wave is divided into a series of rising edges and trailing edge region;

In the each rising edge of preset wavelength λ, obtain with first sampled point and differ the amplitude p such as described ^λsampled point;

The amplitude sampled point such as pass through and divide the rising edge that other logarithm pulse waves are corresponding;

In each rising edge, obtain valley point to etc. all band logarithm pulse wave data of amplitude sampled point region, ask for the match value of the amplitude sampled points such as each wavelength by least square fitting;

In each rising edge, full wave all match values form a tentering value Real-time and Dynamic spectrum.

The tentering value Real-time and Dynamic spectrum that described rejecting contains gross error, the step that residue tentering value Real-time and Dynamic spectrum superposed average is obtained to final dynamic spectrum output is specially:

To all tentering value Real-time and Dynamic spectrum superposed averages, obtain average dynamic spectrum; The similarity degree of all tentering value Real-time and Dynamic spectrum and described average dynamic spectrum is described by Euclidean distance;

Utilize 3 σ criterions according to described similarity degree, delete the tentering value Real-time and Dynamic spectrum that residual error is greater than 3 σ;

Residue tentering value Real-time and Dynamic spectrum superposed average is obtained to final tentering value dynamic spectrum output.

The beneficial effect of technical scheme provided by the invention is: this method is by being divided into pulse wave in rising edge and trailing edge region, ask for each rising edge etc. amplitude Real-time and Dynamic spectrum, by the impact of statistical average effect suppressing exception waveform and singular value, improve the signal to noise ratio of tentering value dynamic spectrum again; The more important thing is determining by amplitudes such as different detected part, in not changing dynamic spectrum effective information, contribute to reduce the impact of different measuring position optical length difference, and balance the signal to noise ratio of different detected part dynamic spectrums, the precision that detects calibration model for improving dynamic spectrum noinvasive blood constituent provides effective method.

Brief description of the drawings

Fig. 1 is the flow chart of the dynamic spectral data extracting method of a kind of tentering value provided by the invention;

Fig. 2 is the flow chart of the amplitude such as default provided by the invention;

Fig. 3 is the flow chart that obtains tentering value Real-time and Dynamic spectrum provided by the invention;

Fig. 4 is the flow chart of the final dynamic spectrum of output provided by the invention.

Detailed description of the invention

For making the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing, embodiment of the present invention is described further in detail.

In order to solve the problem of dynamic spectrum optical length between current Different Individual, avoid the loss of effective information, the embodiment of the present invention provides a kind of dynamic spectral data extracting method of tentering value, referring to Fig. 1, Fig. 2, Fig. 3 and Fig. 4, described below:

101: the amplitude p such as collection all band photoelectricity volume pulsation wave is also made logarithmic transformation, selects the preset wavelength in all band, default ^λ;

This step specifically comprises step 1011-1013, described below:

1011: all band photoelectricity volume pulsation wave of at least one pulse wave cycle of synchronous acquisition detected part;

Wherein, detected part can be the positions such as finger or ear-lobe, and when specific implementation, the embodiment of the present invention does not limit this.The all band photoelectricity volume pulsation wave gathering is expressed as formula (1), λ 1, and λ 2 ..., λ k is corresponding to k wavelength, and n represents the sampled point number of each wavelength pulse wave.

$\left(\begin{array}{cccc}{x}_1^{\mathrm{\λ}1}& x_2^{\mathrm{\λ}1}& \·\·\·& x_n^{\mathrm{\λ}1}\\ x_1^{\mathrm{\λ}2}& x_2^{\mathrm{\λ}2}& \·\·\·& x_n^{\mathrm{\λ}2}\\ \·\·\·& & & \\ x_1^{\mathrm{\λk}}& x_2^{\mathrm{\λk}}& \·\·\·& x_n^{\mathrm{\λk}}\end{array}\right)---\left(1\right)$

1012: all band photoelectricity volume pulsation wave is made to logarithmic transformation, obtain all band logarithm pulse wave and be expressed as formula (2);

$\left(\begin{array}{cccc}{y}_1^{\mathrm{\λ}1}& y_2^{\mathrm{\λ}1}& \·\·\·& y_n^{\mathrm{\λ}1}\\ y_1^{\mathrm{\λ}2}& y_2^{\mathrm{\λ}2}& \·\·\·& y_n^{\mathrm{\λ}2}\\ \·\·\·& & & \\ y_1^{\mathrm{\λk}}& y_2^{\mathrm{\λk}}& \·\·\·& y_n^{\mathrm{\λk}}\end{array}\right)---\left(2\right)$

1013: the preset wavelength λ in selected all band, on preset wavelength λ, the amplitude such as default is designated as p ^λ.

Wherein, the selection of wavelength X is set according to the needs in practical application, and the embodiment of the present invention does not limit this.

102: obtain rising edge and trailing edge region in all band logarithm pulse wave, taking etc. amplitude in each rising edge, obtain amplitude the matching with the sampling period as standard, amplitude that all matchings obtain forms a tentering value Real-time and Dynamic spectrum;

This step specifically comprises step 1021-1026, described below:

1021: obtain the logarithm pulse wave that is greater than average amplitude in all band logarithm pulse wave, logarithm pulse wave superposed average is obtained to pulse wave template;

If wavelength X a ..., λ b(1<a<b<k) logarithm pulse wave amplitude be greater than average amplitude, these logarithm pulse wave corresponding point superposed averages are obtained to pulse wave template, for example: add be expressed as formula (3),

(y ₁y ₂...y _n) (3)

1022: search all peak points and the valley point of pulse wave template, by peak point and valley point, all band logarithm pulse wave is divided into a series of rising edges and trailing edge region;

Wherein, peak point and the valley point method of searching pulse wave template are conventionally known to one of skill in the art, and the embodiment of the present invention does not repeat at this.For example: if find out 4 peak points and valley point is expressed as formula (4), subscript 1 is expressed as peak point, and subscript 0 is expressed as valley point.

$\left(\begin{array}{cccc}{n}_1^0& n_2^1& n_3^0& n_4^1\end{array}\right)---\left(4\right)$

By peak point and valley point, all band logarithm pulse wave is divided into a series of rising edges region and trailing edge region, be divided into three regions according to formula (4), be respectively rising edge region, trailing edge region and rising edge region, be expressed as formula (5)

$\left(\begin{array}{ccc}{y}_{n_1}^{\mathrm{\&lambda;}1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_2}^{\mathrm{\&lambda;}1}\\ y_{n_1}^{\mathrm{\&lambda;}2}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_2}^{\mathrm{\&lambda;}2}\\ & \&CenterDot;\&CenterDot;\&CenterDot;& \\ y_{n_1}^{\mathrm{\&lambda;k}}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_2}^{\mathrm{\&lambda;k}}\end{array}\right)\left(\begin{array}{ccc}{y}_{n_2}^{\mathrm{\&lambda;}1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_3}^{\mathrm{\&lambda;}1}\\ y_{n_2}^{\mathrm{\&lambda;}2}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_3}^{\mathrm{\&lambda;}2}\\ & \&CenterDot;\&CenterDot;\&CenterDot;& \\ y_{n_2}^{\mathrm{\&lambda;k}}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_3}^{\mathrm{\&lambda;k}}\end{array}\right)\left(\begin{array}{ccc}{y}_{n_3}^{\mathrm{\&lambda;}1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_4}^{\mathrm{\&lambda;}1}\\ y_{n_3}^{\mathrm{\&lambda;}2}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_4}^{\mathrm{\&lambda;}2}\\ & \&CenterDot;\&CenterDot;\&CenterDot;& \\ y_{n_3}^{\mathrm{\&lambda;k}}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_4}^{\mathrm{\&lambda;k}}\end{array}\right)---\left(5\right)$

1023: in the each rising edge of preset wavelength λ, obtain with first sampled point and the amplitude p such as differ ^λsampled point;

For example: in wavelength X, the rising edge of logarithm pulse wave is expressed as formula (6),

$\left(\begin{array}{ccc}{y}_{n_1}^{\mathrm{\&lambda;}}& ...& y_{n_2}^{\mathrm{\&lambda;}}\end{array}\right)$ With $\left(\begin{array}{ccc}{y}_{n_3}^{\mathrm{\&lambda;}}& ...& y_{n_4}^{\mathrm{\&lambda;}}\end{array}\right)---\left(6\right)$

The amplitude p such as suppose to differ ^λsampled point be respectively n _e1and n _e2point, n ₁<n _e1<n ₂, n ₃<n _e2<n ₄, sampled point meets formula (7),

$y_{n_{e1}}^{\mathrm{\&lambda;}}-y_{n_1}^{\mathrm{\&lambda;}}=p^{\mathrm{\&lambda;}},y_{n_{e2}}^{\mathrm{\&lambda;}}-y_{n_3}^{\mathrm{\&lambda;}}=p^{\mathrm{\&lambda;}}---\left(7\right)$

1024: the amplitude sampled point such as pass through and divide the rising edge that other logarithm pulse waves are corresponding;

With n _e1and n _e2the position at some place is divided rising edge corresponding to logarithm pulse wave beyond wavelength X.

1025: in each rising edge, obtain valley point to etc. all band logarithm pulse wave data of amplitude sampled point region, ask for the match value of the amplitude sampled points such as each wavelength by least square fitting;

The data of one of them rising edge can be expressed as formula (8),

$\left(\begin{array}{ccc}{y}_{n_1}^{\mathrm{\&lambda;}1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_{e1}}^{\mathrm{\&lambda;}1}\\ y_{n_1}^{\mathrm{\&lambda;}2}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_{e1}}^{\mathrm{\&lambda;}2}\\ & \&CenterDot;\&CenterDot;\&CenterDot;& \\ y_{n_1}^{\mathrm{\&lambda;k}}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{n_{e1}}^{\mathrm{\&lambda;k}}\end{array}\right)---\left(8\right)$

Data to each wavelength in each rising edge region are removed base value, for example: expression formula (8) medium wavelength λ 1 is removed to base value, and correspondence is expressed as formula (9),

Approximating method is least square fitting, obtains etc. the match value of amplitude sampled point by single order matching by that analogy other wavelength in this rising edge region are carried out to same processing, obtain the match value under each wavelength.

1026: in each rising edge, full wave all match values form a tentering value Real-time and Dynamic spectrum.

Be formula (10) for the tentering value Real-time and Dynamic spectral representation that in this rising edge of formula (8), full wave all match values form,

103: reject the tentering value Real-time and Dynamic spectrum that contains gross error, residue tentering value Real-time and Dynamic spectrum superposed average is obtained to final dynamic spectrum output.

This step specifically comprises 1031-1033, described below:

1031: to all tentering value Real-time and Dynamic spectrum superposed averages, obtain average dynamic spectrum; The similarity degree of all tentering value Real-time and Dynamic spectrum and average dynamic spectrum is described by Euclidean distance;

Wherein, describe the similarity degree of tentering value Real-time and Dynamic spectrum and average dynamic spectrum by Euclidean distance, Euclidean distance is less shows that the similarity degree of tentering value Real-time and Dynamic spectrum and average dynamic spectrum is higher.

1032: utilize 3 σ criterions according to similarity degree, delete the tentering value Real-time and Dynamic spectrum that residual error is greater than 3 σ;

Wherein, in measuring process the volume pulsation wave signal of certain time period have motion artifacts or noise larger; can affect the certainty of measurement that this section extracts tentering value Real-time and Dynamic spectrum, make tentering value Real-time and Dynamic spectrum and the average dynamic spectrum Euclidean distance of this segment signal larger.Calculate the meansigma methods of each tentering value Real-time and Dynamic spectrum and average dynamic spectrum Euclidean distance , residual error v _i, standard deviation sigma.If the residual error of tentering value Real-time and Dynamic spectrum is greater than 3 σ, think that this tentering value Real-time and Dynamic spectral error is large and reject, retain if be less than 3 σ.

1033: residue tentering value Real-time and Dynamic spectrum superposed average is obtained to final tentering value dynamic spectrum output.

The least-square fitting approach being applied in embodiment of the present invention method, Euclidean distance and 3 σ decision criterias are the known technology in data processing method, and the embodiment of the present invention does not repeat at this.

In sum, the embodiment of the present invention provides a kind of dynamic spectral data extracting method of tentering value, the method is by being divided into pulse wave in rising edge and trailing edge region, ask for each rising edge etc. amplitude Real-time and Dynamic spectrum, by the impact of statistical average effect suppressing exception waveform and singular value, improve the signal to noise ratio of tentering value dynamic spectrum again; The more important thing is determining by amplitudes such as different detected part, in not changing dynamic spectrum effective information, contribute to reduce the impact of different measuring position optical length difference, and balance the signal to noise ratio of different detected part dynamic spectrums, the precision that detects calibration model for improving dynamic spectrum noinvasive blood constituent provides effective method.

It will be appreciated by those skilled in the art that accompanying drawing is the schematic diagram of a preferred embodiment, the invention described above embodiment sequence number, just to describing, does not represent the quality of embodiment.

The foregoing is only preferred embodiment of the present invention, in order to limit the present invention, within the spirit and principles in the present invention not all, any amendment of doing, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.

Claims (4)

1. a dynamic spectral data extracting method for tentering value, is characterized in that, said method comprising the steps of:

(1) gather all band photoelectricity volume pulsation wave make logarithmic transformation, select the preset wavelength in all band, the amplitude p such as default ^λ;

(2) obtain rising edge and the trailing edge region in all band logarithm pulse wave, with the amplitude p such as described ^λin each rising edge, obtain amplitude the matching with the sampling period for standard, the amplitude that all matchings obtain forms a tentering value Real-time and Dynamic spectrum;

(3) reject the tentering value Real-time and Dynamic spectrum that contains gross error, residue tentering value Real-time and Dynamic spectrum superposed average is obtained to final dynamic spectrum output.

2. the amplitude p such as the dynamic spectral data extracting method of a kind of tentering value according to claim 1, is characterized in that, described collection all band photoelectricity volume pulsation wave is also made logarithmic transformation, selects the preset wavelength in all band, default ^λstep be specially:

The described all band photoelectricity volume pulsation wave of at least one pulse wave cycle of synchronous acquisition detected part;

Described all band photoelectricity volume pulsation wave is made to logarithmic transformation, obtain all band logarithm pulse wave;

Preset wavelength λ in selected all band, on preset wavelength λ, the amplitude such as default described is designated as p ^λ.

3. the dynamic spectral data extracting method of a kind of tentering value according to claim 1, is characterized in that, described in obtain rising edge and the trailing edge region in all band logarithm pulse wave, with the amplitude p such as described ^λin each rising edge, obtain amplitude the matching with the sampling period for standard, the step that the amplitude that all matchings obtain forms a tentering value Real-time and Dynamic spectrum is specially:

Obtain the logarithm pulse wave that is greater than average amplitude in described all band logarithm pulse wave, the logarithm pulse wave superposed average that is greater than average amplitude is obtained to pulse wave template;

Search all peak points and the valley point of described pulse wave template, by described peak point and described valley point, described all band logarithm pulse wave is divided into a series of rising edges and trailing edge region;

In the each rising edge of preset wavelength λ, obtain with first sampled point and differ the amplitude p such as described ^λsampled point;

The amplitude sampled point such as pass through and divide rising edge corresponding to logarithm pulse wave beyond preset wavelength λ;

In each rising edge, obtain valley point to etc. all band logarithm pulse wave data of amplitude sampled point region, ask for the match value of the amplitude sampled points such as each wavelength by least square fitting;

In each rising edge, full wave all match values form a tentering value Real-time and Dynamic spectrum.

4. the dynamic spectral data extracting method of a kind of tentering value according to claim 1, it is characterized in that, the tentering value Real-time and Dynamic spectrum that described rejecting contains gross error, the step that residue tentering value Real-time and Dynamic spectrum superposed average is obtained to final dynamic spectrum output is specially:

To all tentering value Real-time and Dynamic spectrum superposed averages, obtain average dynamic spectrum; The similarity degree of all tentering value Real-time and Dynamic spectrum and described average dynamic spectrum is described by Euclidean distance;

Utilize 3 σ criterions according to described similarity degree, delete the tentering value Real-time and Dynamic spectrum that residual error is greater than 3 σ;

Residue tentering value Real-time and Dynamic spectrum superposed average is obtained to final tentering value dynamic spectrum output.