CN106137167A - A kind of motion artifacts detection method based on photoplethysmographic signal - Google Patents

Abstract

The invention discloses a kind of motion artifacts detection method based on photoplethysmographic signal, the follow-up work for heart rate measurement lays the foundation.In the method, by reflective photoelectric sensor and motion sensor collection with the multiple photoplethysmographic signal in the time period and acceleration signal；Use PCA that acceleration signal is processed, produce the reference signal that motion artifacts is relevant, and combine minimum mean square self-adaption filter elimination componental movement noise；Then after processing, multiple photoplethysmographic signal and acceleration signal constitute spectral matrix, extract the sparse architectural feature of spectral matrix row and build sparse signal reconfiguring model；Sparse signal reconfiguring model is optimized, it is thus achieved that the spectrum peak position of motion artifacts in multiple photoplethysmographic signal frequency spectrums finally by regularization algorithm.The present invention can accurately detect out the motion artifacts in photoplethysmographic signal, it is achieved the high-acruracy survey of heart rate.

Description

A kind of motion artifacts detection method based on photoplethysmographic signal

Technical field

The present invention relates to physiological single processing field, particularly relate to a kind of motion based on photoplethysmographic signal and make an uproar Sound detection method.

Background technology

Heart rate measurement, as the objective evaluation index of human motion physiological stress, has been widely used in body building, competing The various aspects of skill athletic training.Though traditional heart rate measurement technology can reach higher certainty of measurement, but measuring condition limits Make more.In order to meet future electronic health monitoring requirement, method for measuring heart rate based on photoplethysmographic signal causes The highest attention of academia and industrial quarters.

But, photoplethysmographic is that a kind of signal intensity is weak, easily by the bio signal of noise jamming.At kinestate Under, due to tissue interference, venous blood volume and change in optical path length, easily produce the motion artifacts being sufficiently close to heart rate frequency, And then make heart rate measurement precise decreasing.

In recent years, research worker has carried out relevant grinding to the detection work of motion artifacts in photoplethysmographic signal Study carefully.Such as, patent of invention " a kind of motion artifacts detection method being applicable to heart rate signal " (application number: 2015108739783, Application publication number: CN105286846A) in propose joint sparse spectrum reconstruction model, whole spectral matrix is carried out by this model The restriction that row is sparse and the overall situation is sparse, and the optimal solution of this model is solved by inaccuracy augmented vector approach, enter And detect the position of motion artifacts.And the technology that the present invention uses quick Denoising Algorithm and signal reconstruction algorithm to combine is carried out The detection of the motion artifacts in photoplethysmographic signal.I.e. filter first with PCA and minimum mean square self-adaption The combination of ripple device eliminates the componental movement noise in photoplethysmographic；Secondly the architectural feature that spectral matrix row is sparse is extracted Build sparse signal reconfiguring model, and by regularization M-FOCUSS algorithm optimization reconstruction model.Present invention greatly improves The detection degree of accuracy of motion artifacts, reduces computation complexity.

Summary of the invention

The technical problem to be solved is the most effectively to detect strong movements in photoplethysmographic signal The spectrum peak position of noise, thus realize the high-acruracy survey of heart rate.

In order to solve above-mentioned technical problem, the invention provides a kind of motion artifacts based on photoplethysmographic signal Detection method, it is characterised in that:

Described acceleration signal is processed by described PCA, and combine described sef-adapting filter eliminate institute State the componental movement noise in multiple photoplethysmographic signal；Then, the multiple photoplethysmographic letter after processing Number and described acceleration signal constitute spectral matrix, according to above-mentioned spectral matrix architectural feature build described sparse signal reconfiguring Model, and optimize described sparse signal reconfiguring model；

The method comprises the steps:

Pretreated described acceleration signal is analyzed by described PCA, produces what motion artifacts was correlated with Reference signal, and combine minimum mean square self-adaption filter the componental movement noise in described photoplethysmographic signal is entered Row eliminates；

Meanwhile, the multiple photoplethysmographic signal and described acceleration signal of removing componental movement noise are constituted frequency Spectrum matrix, sets up described sparse signal reconfiguring model according to the architectural feature that above-mentioned spectral matrix row is sparse, and passes through regularization Reconstruction model described in M-FOCUSS algorithm optimization, it is thus achieved that the spectral peak position of motion artifacts in multiple photoplethysmographic signal frequency spectrums Put；

Preferably, described acceleration signal has strong correlation with the motion artifacts in described photoplethysmographic signal Property, the most described acceleration signal can describe " footprint " of motion artifacts on three direction of principal axis；Therefore described PCA pair Described acceleration signal processes, and selects and comprises the first principal component that " information " is most, and be correlated with as motion artifacts Reference signal；

Preferably, described minimum mean square self-adaption filter selects the reference signal that above-mentioned motion artifacts is relevant, according to all The criterion of side's error minimize, constantly updates filtering weighting, and the componental movement eliminated in described photoplethysmographic signal is made an uproar Sound；

Preferably, described spectral matrix is by multiple photoplethysmographic signal of above-mentioned elimination componental movement noise and institute State acceleration signal to constitute；The architectural feature sparse according to described spectral matrix row constructs described sparse signal reconfiguring model, its Object function is as follows:

$\underset{X}{\min }\frac{1}{2}\left|\right|Y-\mathrm{\Φ}X|{|}_F^2+\mathrm{\λ}|\left|X\right|{|}_{1.2}$

S.t:Y=Φ X+V

Wherein,

It is used for retraining spectral matrix row sparse, x_{I, j}It is that spectral matrix X the i-th row jth row are first Element, λ is used to weigh | | X | |_1,2The weights of importance；Y ∈ R in constraints^M×HIt is an observing matrix, X ∈ C^N×HIt it is phase The spectral matrix of induction signal, i.e. needs the sparse spectral matrix solvedIt it is a redundancy Discrete Fourier transform base, V is model error or measurement error matrix.

Compared with prior art, the technical scheme that the present invention provides uses quick Denoising Algorithm to tie mutually with signal reconstruction algorithm The technology closed accurately detects the motion artifacts in photoplethysmographic signal, greatly increases the detection essence of motion artifacts Degree, and reduce computation complexity.

Accompanying drawing explanation

Fig. 1 is the flow process signal of the motion artifacts detection method based on photoplethysmographic signal of the embodiment of the present invention Figure.

Detailed description of the invention

Describe embodiments of the present invention in detail below in conjunction with drawings and Examples, whereby how the present invention is applied skill Art means solve technical problem, and the process that realizes reaching relevant art effect can fully understand and implement according to this.

Technical scheme includes two parts, and first utilizes PCA to be analyzed acceleration signal, The reference signal that the motion artifacts that being produced from adaptive filtering needs is correlated with, and then utilize minimum mean square self-adaption filter to eliminate light Componental movement noise in Power Capacity pulse wave；The strong correlation of the second foundation acceleration signal and motion artifacts signal is at frequency domain Show as the spectrum peak position of motion artifacts in photoplethysmographic signal frequency spectrum and the spectrum peak position phase of acceleration signal frequency spectrum The feature construction sparse signal reconfiguring model of alignment.This technical scheme can accurately detect out in photoplethysmographic signal Motion artifacts, lays the foundation for the work of follow-up removal motion artifacts.

Embodiment one, motion artifacts detection method based on photoplethysmographic signal

Fig. 1 is the schematic flow sheet of the motion artifacts detection method based on photoplethysmographic signal of the present embodiment.

The present embodiment shown in Fig. 1, is the bulk flow of motion artifacts detection method based on photoplethysmographic signal Journey, mainly comprises the steps:

Step S210, utilizes two reflective photoelectric sensors being distributed in diverse location to gather the photocapacitance of two passages Long-pending pulse wave signal, the acceleration signal of three directions of motion that recycling motion sensor collection synchronizes；

Step S220, pretreatment include carrying out above-mentioned primary signal being down-sampled to operation that sample frequency is 25Hz and By the filtering operation of second order Butterworth filter that passband is 0.4Hz-4Hz；

Step S230, utilizes PCA to be analyzed acceleration signal after pretreatment, selects and comprise Reference signal X (n) that the first principal component that " information " is most, i.e. motion artifacts are relevant；

Step S240, the reference signal that the motion artifacts that minimum mean square self-adaption filter optional step S220 produces is relevant X (n), eliminates componental movement noise in photoplethysmographic signal by constantly updating filtering weighting W (n)；

In this step, typically, the mathematical expression that difference e (n) and filtering weighting W (n) update is as follows:

$\begin{array}{c}e\left(n\right)=S\left(n\right)+M\left(n\right)-M^{,}\left(n\right)\\ M^{,}\left(N\right)=W\left(n\right)*X\left(n\right)\\ W(n+1)=W\left(n\right)-\mathrm{\μ}*e\left(n\right)*X\left(n\right)\end{array}---\left(1\right)$

Wherein, S (n) represents clean photoplethysmographic signal, and M (n) represents motion artifacts signal, and M ' (n) represents Farthest approximate motion interference signal；

Step S250, utilizes two photoplethysmographic signal after step S240 processes and three acceleration signals Constitute spectral matrix；

Step S260, according to the architectural feature that above-mentioned spectral matrix row is sparse, constructs sparse signal reconfiguring model；

In this step, typically, formula (2) is the object function of sparse signal reconfiguring model:

$\begin{array}{c}\underset{X}{\min }\frac{1}{2}\left|\right|Y-\mathrm{\Φ}X|{|}_F^2+\mathrm{\λ}|\left|X\right|{|}_{1.2}\\ s.t:Y=\mathrm{\Φ}X+V\end{array}---\left(2\right)$

Wherein,It is used for retraining spectral matrix row sparse, x_{I, j}It it is spectral matrix X the i-th row jth Column element, λ is used to weigh | | X | |_1,2The weights of importance；Y ∈ R in constraints^M×HIt is an observing matrix, X ∈ C^N×H It is the spectral matrix of corresponding signal, i.e. needs the sparse spectral matrix solved,It it is one Redundant discrete Fourier transformation base, V is model error or measurement error matrix, H=5 in the present embodiment；

Step S270, can solve the object function of above-mentioned sparse signal reconfiguring model by regularization M-FOCUSS algorithm Optimal solution；

In this step, typically, the mathematic(al) representation of above-mentioned M-FOCUSS algorithm is as follows:

$\begin{array}{ccc}{W}_{k+1}=diag\left(c_k{\[i\]}^{1-\frac{p}{2}}\right),& where& c_k\[i\]={(\Σ{\left(x_k^{\left(l\right)}\[i\]\right)}^2)}^{\frac{1}{2}}\\ Q_{k+1}{A}_{k+1}^{+}B,& where& A_{k+1}={\mathrm{AW}}_{k+1}\\ X_{k+1}=W_{k+1}{Q}_{k+1}& & \end{array}---\left(3\right)$

Step S280, after above-mentioned steps processes, can obtain the spectrum of motion artifacts in photoplethysmographic signal frequency spectrum Peak position, thus complete to detect the work of motion artifacts.

In the present embodiment, it is to utilize two instead with two photoplethysmographic signal in the time period and acceleration signal Penetrate formula photoelectric sensor and motion sensor gathers at user's wrist；Utilize PCA to acceleration signal Process, it is thus achieved that the reference signal that motion artifacts is relevant, eliminate two photoelectricity volumes in conjunction with minimum mean square self-adaption filter Componental movement noise in pulse wave signal；Then utilize remove componental movement noise two photoplethysmographic signal and Acceleration signal constitutes spectral matrix, and the architectural feature extracting the row of above-mentioned spectral matrix sparse sets up sparse signal reconfiguring mould Type, by the sparse spectral matrix in the regularization above-mentioned reconstruction model of M-FOCUSS Algorithm for Solving；Finally obtain two photocapacitance The spectrum peak position of motion artifacts in long-pending pulse wave signal frequency spectrum.The method greatly increases the detection performance of motion artifacts, fall Low computation complexity, for realizing the high-acruracy survey based theoretical of heart rate.

Although the embodiment that disclosed herein is as above, but foregoing is only to facilitate understand the present invention and use Embodiment, be not limited to the present invention.On the premise of without departing from the spirit and scope that disclosed herein, can be in reality That executes and makees any modification and change in form in details, but the scope of patent protection of the present invention, still must be with appended right Claim is defined in the range of standard.

Claims (4)

1. a motion artifacts detection method based on photoplethysmographic signal, it is characterised in that:

Described acceleration signal is processed by described PCA, and it is described many to combine the elimination of described sef-adapting filter Componental movement noise in individual photoplethysmographic signal；Then, will process after multiple photoplethysmographic signal and Described acceleration signal constitutes spectral matrix, builds described sparse signal reconfiguring mould according to the architectural feature of above-mentioned spectral matrix Type, and optimize described sparse signal reconfiguring model；

The method comprises the steps:

Pretreated described acceleration signal is analyzed by described PCA, produces the reference that motion artifacts is relevant Signal, and combine minimum mean square self-adaption filter the componental movement noise in described photoplethysmographic signal is disappeared Remove；

Meanwhile, the multiple photoplethysmographic signal and described acceleration signal of removing componental movement noise are constituted frequency spectrum square Battle array, sets up described sparse signal reconfiguring model according to the architectural feature that above-mentioned spectral matrix row is sparse, and by regularization M- Reconstruction model described in FOCUSS algorithm optimization, it is thus achieved that the spectral peak position of motion artifacts in multiple photoplethysmographic signal frequency spectrums Put.

Motion artifacts detection method based on photoplethysmographic signal the most according to claim 1, it is characterised in that:

Described acceleration signal has strong correlation with the motion artifacts in described photoplethysmographic signal, the most described acceleration Degree signal can describe " footprint " of motion artifacts on three direction of principal axis；Therefore described PCA is to described acceleration signal Process, select and comprise the first principal component that " information " is most, and the reference signal being correlated with as motion artifacts.

Motion artifacts detection method based on photoplethysmographic signal the most according to claim 1, it is characterised in that:

Described minimum mean square self-adaption filter selects the reference signal that above-mentioned motion artifacts is relevant, minimizes according to mean square error Criterion, constantly update filtering weighting, eliminate the componental movement noise in described photoplethysmographic signal.

Motion artifacts detection method based on photoplethysmographic signal the most according to claim 1, it is characterised in that:

Described spectral matrix is believed by multiple photoplethysmographic signal and the described acceleration of above-mentioned elimination componental movement noise Number constitute；The architectural feature sparse according to described spectral matrix row constructs described sparse signal reconfiguring model, and its object function is such as Under:

S.t:Y=Φ X+V

Wherein,

It is used for retraining spectral matrix row sparse, x_{I, j}Being spectral matrix X the i-th row jth column element, λ is It is used for weighing | | X | |_1,2The weights of importance；Y ∈ R in constraints^M×HIt is an observing matrix, X ∈ C^N×HIt it is corresponding signal Spectral matrix, i.e. need the sparse spectral matrix solvedIt is redundant discrete Fu In leaf transformation base, V is model error or measurement error matrix.