CN106388824B - Respiration rate extraction method and device - Google Patents
Respiration rate extraction method and device Download PDFInfo
- Publication number
- CN106388824B CN106388824B CN201610899933.8A CN201610899933A CN106388824B CN 106388824 B CN106388824 B CN 106388824B CN 201610899933 A CN201610899933 A CN 201610899933A CN 106388824 B CN106388824 B CN 106388824B
- Authority
- CN
- China
- Prior art keywords
- respiratory rate
- weight factor
- breath signal
- obtains
- characteristic root
- 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.)
- Active
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/08—Measuring devices for evaluating the respiratory organs
- A61B5/0816—Measuring devices for examining respiratory frequency
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7221—Determining signal validity, reliability or quality
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Surgery (AREA)
- Public Health (AREA)
- Pathology (AREA)
- Veterinary Medicine (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Physiology (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Pulmonology (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Cardiology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a respiratory rate extraction method, which comprises the following steps: extracting the received electrocardiosignals through a pre-trained neural network model to obtain first respiratory signals, and calculating to obtain a first respiratory rate at the current moment according to the first respiratory signals; extracting the electrocardiosignals through a constructed autoregressive model to obtain second respiration signals, and calculating according to the second respiration signals to obtain a second respiration rate of the current moment; analyzing the first respiratory signal and the second respiratory signal based on a signal quality index to obtain a first weight factor corresponding to the first respiratory signal and a second weight factor corresponding to the second respiratory signal; and calculating to obtain the breathing rate of the current moment according to the first breathing rate, the first weight factor, the second breathing rate and the second weight factor. The invention also discloses a respiration rate extraction device which can conveniently and effectively extract respiration signals and calculate to obtain accurate and stable respiration rates.
Description
Technical field
The present invention relates to breathing detection field more particularly to a kind of respiratory rate extracting method and devices.
Background technique
Breathing is the important physiology course of human body, and the monitoring detection to human body respiration is also the one of modern medicine monitoring technology
A important component.Patient whether the lesion of respiratory system itself or the pathological development of other important organs to certain journey
Degree can all influence respiratory center.The failure of respiratory function is often involved in the failure of multi viscera system function, and the failure of respiratory function is again
Lead to the failure of other organs function, reciprocal causation.
The prior art mainly detects respiratory movement using following method: impedance volumetric method: measuring chest with high-frequency constant current source
Respiration information is extracted in the variation of portion's impedance；Sensor method: temperature, pressure, humidity and gas flow transducer is used to pass as nostril
Sensor；Capacitance method: capacitance is caused to generate corresponding variation when breathing；Breath sound method: by picking up breath sound identification of breathing；
Ultrasonic method: Doppler phenomenon is generated using ultrasonic wave, detects respiratory rate.It not only needs to increase signal using these methods and adopt
Collect component, and the shadow noon by movement and environment, is not suitable for daily monitoring.
A large amount of clinical datas show that respiratory movement can cause the variation of electrocardiogram.By electrocardiogram, we can observe that
By the change of the caused ecg wave form peak-to-peak value of chest exercise and cardiac position variation within the respiratory cycle.This is because breathing
In period, the heart electric axis rotation in description heart electric wave main propagation direction causes QRS complex form to be changed.QRS wave
Refer to the maximum wave group of amplitude in normal ECG, reflects the overall process of sequences of ventricular depolarization.Normal ventricle depolarization starts from interventricular septum
Portion, direction depolarization from left to right, therefore one small downward q wave is first presented in QRS complex.Normal chest leads QRS complex form is more permanent
It is fixed.It is a kind of breath signal detection skill that breath signal (ECG-DerivedRespiration, EDR) is extracted from electrocardiosignal
Art, this technology do not need sensor special and hardware module detection breath signal, it is only necessary to obtain electrocardio with ECG monitor
Signal avoids constraint of the above two detection method to human body, makes it possible dynamic breathing detection.
But the existing technology that breath signal is extracted from electrocardiosignal, Waveform Method is mainly used when calculating, and this method is logical
The average value (i.e. baseline value) of interior waveform after a period of time rises or falls trend to determine that current respiratory wave is in, uses extreme value
Method acquire the wave crest of waveform, trough.Effective wave crest or trough are determined according to certain threshold condition, further according to effective
The period of wave crest or trough calculates wave period, to obtain respiratory rate.Although this algorithm has, relatively more intuitive, operand is small
The advantages of, but the respiratory waveform obtained in the actual process more or less will receive the influence of electrocardio-activity, when base occurs in waveform
Line drift about when, the baseline value of calculating can not update quickly, and will lead to waveform missing inspection causes respiratory rate value relatively low, result have compared with
Large deviation.
Summary of the invention
In view of the above-mentioned problems, realization accurately may be used the purpose of the present invention is to provide a kind of respiratory rate extracting method and device
The measurement of the respiratory rate leaned on, and can reduce measurement fluctuation or error due to caused by extraneous or environment interference.
The present invention provides a kind of respiratory rate extracting methods, comprising:
The electrocardiosignal received is extracted by the preparatory trained neural network model about breath signal,
The first breath signal is obtained, and first respiratory rate at current time is calculated according to first breath signal；
The electrocardiosignal is extracted by the autoregression model about breath signal built, second is obtained and exhales
Signal is inhaled, and second respiratory rate at current time is calculated according to second breath signal；
Based on signal quality index, first breath signal and second breath signal are analyzed, obtain with
Corresponding first weight factor of first breath signal and the second weight factor corresponding with second breath signal；
According to first respiratory rate, the first weight factor, the second respiratory rate and the second weight factor, it is calculated current
The respiratory rate at moment.
Preferably, the electrocardiosignal received is extracted by preparatory trained neural network model described,
The first breath signal is obtained, and before first respiratory rate at current time is calculated according to first breath signal, is also wrapped
It includes:
Multi-lead electrocardiosignal is received, the RR interphase and R peak amplitude of each lead electrocardiosignal is calculated separately, is inputted
Sample space, wherein the dimension in the input sample space is p, and p/2 is the lead number of multi-lead electrocardiosignal；
The covariance matrix formed according to the input sample space is handled based on Principal Component Analysis, is led
Component score matrix；
Using the principal component scores matrix and the target breath signal obtained by impedance method synchronous acquisition as training sample
To being trained, neural network model is obtained.
Preferably, described that the covariance matrix formed according to the input sample space is carried out based on Principal Component Analysis
Processing, obtains principal component scores matrix, specifically includes:
Data normalization processing is carried out to the input sample space；
According to data normalization, treated that the input sample space obtains covariance matrix；
Calculate the covariance matrix characteristic root and feature vector corresponding with each characteristic root；Wherein, the feature
The quantity of root is p, and the p characteristic root is in magnitude order；
It obtains in the p characteristic root, the sum of contribution rate is greater than the preceding m characteristic root of predetermined threshold；Wherein, Mei Gete
The contribution rate for levying root is equal to the value of the characteristic root divided by the sum of the value of p whole characteristic roots；
According to feature vector corresponding with the preceding m characteristic root and the input sample space, obtains principal component and obtain
Sub-matrix.
Preferably, the autoregression model is the autoregression model after moving average method optimizes.
It is preferably, described according to first respiratory rate, the first weight factor, the second respiratory rate and the second weight factor,
The respiratory rate at current time is calculated specifically:
When judging that first weight factor is greater than preset a reference value and second weight factor and is less than the benchmark
When value, it sets first respiratory rate to the respiratory rate at current time；
When judging that first weight factor is less than preset a reference value and second weight factor and is greater than the benchmark
When value, it sets second respiratory rate to the respiratory rate at current time；
When judging that first weight factor and second weight factor are all larger than preset a reference value, according to described
First weight factor and second weight factor are weighted summation to first respiratory rate and the second respiratory rate, calculate
To the respiratory rate at current time.
The present invention also provides a kind of respiratory rate extraction elements, comprising:
Neural network extraction unit, for passing through the preparatory trained neural network model about breath signal to reception
To electrocardiosignal extract, obtain the first breath signal, and current time is calculated according to first breath signal
The first respiratory rate；
Autoregression extraction unit, for the autoregression model about breath signal by building to the electrocardiosignal
It extracts, obtains the second breath signal, and second respiratory rate at current time is calculated according to second breath signal；
Signal quality analytical unit exhales first breath signal and described second for being based on signal quality index
Inhale signal analyzed, obtain the first weight factor corresponding with first breath signal and with second breath signal pair
The second weight factor answered；
Respiratory rate computing unit, for according to first respiratory rate, the first weight factor, the second respiratory rate and the second power
The respiratory rate at current time is calculated in repeated factor.
Preferably, further includes:
Input sample space acquiring unit calculates separately each lead electrocardiosignal for receiving multi-lead electrocardiosignal
RR interphase and R peak amplitude, obtain input sample space, wherein the dimension in the input sample space be p, p/2 is multi-lead
The lead number of electrocardiosignal；
Principal component analysis unit, for based on Principal Component Analysis to the covariance formed according to the input sample space
Matrix is handled, and principal component scores matrix is obtained；
Neural metwork training unit, for the principal component scores matrix and the mesh obtained by impedance method synchronous acquisition
Marking breath signal is training sample to neural metwork training is carried out, and obtains neural network model.
Preferably, the principal component analysis unit specifically includes:
Standardization module, for input sample Standardization of Spatial Data processing；
Covariance matrix computing module, for treated that the input sample space obtains association side according to data normalization
Poor matrix；
Feature calculation module, for calculate the covariance matrix characteristic root and feature corresponding with each characteristic root to
Amount；Wherein, the quantity of the characteristic root is p, and the p characteristic root is in magnitude order；
Screening module, for obtaining in the p characteristic root, the sum of contribution rate is greater than the preceding m feature of predetermined threshold
Root；Wherein, the contribution rate of each characteristic root is equal to the value of the characteristic root divided by the sum of the value of p whole characteristic roots；
Score matrix obtains module, for basis and the corresponding feature vector of preceding m characteristic root and the input
Sample space obtains principal component scores matrix.
Preferably, the autoregression model is the autoregression model after moving average method optimizes.
Preferably, the respiratory rate computing unit specifically includes:
First judgment module judges that first weight factor is greater than preset a reference value and second weight for working as
When the factor is less than a reference value, it sets first respiratory rate to the respiratory rate at current time；
Second judgment module judges that first weight factor is less than preset a reference value and second weight for working as
When the factor is greater than a reference value, it sets second respiratory rate to the respiratory rate at current time；
Third judgment module, for preset when judging that first weight factor and second weight factor are all larger than
When a reference value, according to first weight factor and second weight factor to first respiratory rate and the second respiratory rate into
The respiratory rate at current time is calculated in row weighted sum.
Respiratory rate extracting method provided by the invention and device, by utilizing neural network model and auto-regressive time series
The mode processing cardioelectric signals that technology combines obtain the first respiratory rate and the second respiratory rate, and according to first respiratory rate
Corresponding first weight factor and the second weight factor corresponding with second respiratory rate obtain the respiratory rate at current time, phase
Than obtaining the scheme of breath signal from electrocardiosignal by monotechnics in existing, calculated result is more acurrate reliable, and can reduce by
Measurement fluctuation or error caused by extraneous or environment interference, so as to obtain more accurate stable measurement result.
Detailed description of the invention
Fig. 1 is a kind of flow chart of respiratory rate extracting method provided in an embodiment of the present invention；
Fig. 2 is the waveform diagram of original electro-cardiologic signals provided in an embodiment of the present invention；
Fig. 3 is the waveform diagram of the electrocardiosignal to be processed after notch filter provided in an embodiment of the present invention；
Fig. 4 is the waveform of the first breath signal provided in an embodiment of the present invention extracted by neural network model
Figure；
Fig. 5 is the waveform diagram of the second breath signal provided in an embodiment of the present invention extracted by autoregression model.
Fig. 6 is the structural schematic diagram of respiratory rate extraction element provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The present invention provides a kind of respiratory rate extracting methods, for extracting respiration information from electrocardiosignal, due to exhaling
Baseline drift in electrocardiogram caused by suction acts on, regards respiration information as the low-frequency component of electrocardiosignal, is exhaled by removal
The signal other than frequency is inhaled, thus the respiration information extracted needed for obtaining.
Referring to Fig. 1, it is a kind of respiratory rate extracting method provided in an embodiment of the present invention, includes the following steps:
S1: the electrocardiosignal received is mentioned by the preparatory trained neural network model about breath signal
It takes, obtains the first breath signal, and first respiratory rate at current time is calculated according to first breath signal.
It should be noted that as shown in Figure 2 since original electrocardiosignal usually contains a large amount of Hz noise, need into
Row 50Hz notch filter, to filter out Hz noise, referring to Fig. 3, to be believed according to the electrocardio after the notch filter of the embodiment of the present invention
Number waveform diagram.
In embodiments of the present invention, when extracting using neural network model to the electrocardiosignal received, elder generation is needed
The neural network model that can be used for extracting breath signal is obtained by training.
Specifically, in embodiments of the present invention, the calculating of neural network model can be carried out by the following method:
S01 receives multi-lead electrocardiosignal, calculates separately the RR interphase and R peak amplitude of each lead electrocardiosignal, obtain
Input sample space, wherein the dimension in the input sample space is p, and p/2 is the lead number of multi-lead electrocardiosignal.
In embodiments of the present invention, input sample space X=[x1, x2 ..., xn] indicates m*n dimensional vector, wherein x1 table
Show that length is the column vector of m, by calculating the RR interphase of 1 lead, obtains x1, calculate the R peak amplitude of 1 lead, obtain x2, calculate
The RR interphase of 2 leads, obtains x3；The R peak amplitude for calculating 2 leads, obtains x4；... and so on obtain.
S02 handles the covariance matrix formed according to the input sample space based on Principal Component Analysis, obtains
To principal component scores matrix.
In embodiments of the present invention, it is contemplated that standard multi-lead electrocardiosignal is 12 leads, calculates separately each lead
After RR interphase and R peak amplitude, the characteristic value for needing to be input to neural network reaches 24, and interrelated between each lead, leads
It causes the dimension of input sample larger and input item containing linear correlation, is unfavorable for calculating analysis, need to utilize principal component thus
Analytic approach carries out dimensionality reduction to it.
Specifically, step S12 can include:
S021 carries out data normalization processing to the input sample space.
Specifically,
Wherein:
Wherein, X 'ijIt is the new data after standardization；Mj、SjRespectively indicate a certain column of initial data arithmetic mean of instantaneous value and
Standard (inclined) is poor.
S022, according to data normalization, treated that the input sample space obtains covariance matrix.
Wherein, covariance matrix D=XTX, it may be assumed that
Wherein:
S023, calculate the covariance matrix characteristic root and feature vector corresponding with each characteristic root；Wherein, described
The quantity of characteristic root is p, and the p characteristic root is in magnitude order.
Wherein, (6) DP=P λ
When only considering j-th of characteristic value, there is DPj=Pjλj, that is, solve | D- λjI |=0.Each λ is successively solved, and is made
Sequence arranges by size for it, i.e. λ1≥λ2≥…,≥λp≥0；Then each characteristic value corresponding feature vector P, Jin Erte can be found out
Equation solution is levied to complete.
S024 is obtained in the p characteristic root, and the sum of contribution rate is greater than the preceding m characteristic root of predetermined threshold；Wherein,
The contribution rate of each characteristic root is equal to the value of the characteristic root divided by the sum of the value of p whole characteristic roots.
Firstly, calculating the contribution rate of single principal component and being added up, the number of principal component is determined according to contribution rate of accumulative total
M, so that it is determined that the principal component of required selection.The calculation formula of contribution rate such as formula (7) is described.Contribution rate of accumulative total is m i.e. preceding
The accumulation of contribution rate and, as shown in formula (8).The threshold value Dmax generally takes between 85%~95%.According to previous step
In characteristic root sort it is found that λ1≥λ2≥…,≥λp>=0, successively characteristic root is carried out from front to back (and from big to small) tired
Add, works as contribution rate of accumulative totalWhen greater than Dmax, stopping calculating, the number of the characteristic root λ of cumulative calculation is m at this time,
M principal component before then only needing to choose.
S025, according to feature vector corresponding with the preceding m characteristic root and the input sample space, obtain it is main at
Get sub-matrix.
Wherein, the principal component scores matrix
It should be noted that in embodiments of the present invention, can also calculate the load of principal component, wherein the principal component carries
Lotus mainly reflects principal component scores and former variable xjCorrelation degree, calculation formula are as follows:After obtaining the load of each principal component, so that it may know each of selection
The corresponding primitive character of a principal component go back if it is desired, can be converted according to the dimension of primitive character.
S13 is training with the principal component scores matrix and by the target breath signal that impedance method synchronous acquisition obtains
Sample obtains neural network model to neural metwork training is carried out.
In embodiments of the present invention, in the target for obtaining principal component scores matrix T and being obtained by impedance method synchronous acquisition
Breath signal Y, so that it may the training of neural network is carried out, specifically:
Neural network model is made of input layer (feature), hidden layer, output layer.Wherein, one is all corresponded to accordingly for every layer
Function.In training, the principal component eigenmatrix T is input to input layer, the mesh that will be obtained by impedance method synchronous acquisition
Mark breath signal is input to output layer (such as output layer uses softmax function), i.e., the available parameter for obtaining hidden layer is (such as
The parameter of sigmod function).Wherein, the parameter of neural network model, including weight, biasing etc..
In embodiments of the present invention, the hidden layer of neural network model can for 1 (desirable other values, under present case, 1
Can satisfy requirement), the number of hidden nodes is K, and learning rate is usually selected between 0.02~0.2 by learning rate μ.
In order to obtain optimal network training effect, trial and error procedure can be used and determine e-learning rate.
In embodiments of the present invention, the learning algorithm of neural network can choose Levenberg-Marquart algorithm, can also
Using other learning algorithms, the present invention is not specifically limited.
In embodiments of the present invention, after obtaining trained neural network model, the electrocardio to be processed need to only be believed
Number as the input layer for being input to the neural network model, it can obtain mentioning from the electrocardiosignal in output layer
The first breath signal as shown in Figure 4 is obtained, hereafter, the first breathing can be calculated according to first breath signal
Rate.
It specifically, can be by asking extremum method to find the wave of the first breath signal in the waveform diagram of first breath signal
Peak (or trough), referring to fig. 4 in point label；
By extracting the time interval between two wave crests being newly generated, to obtain the cycle T at current time.
Carrying out sampling rate conversion to the period can be obtained the first respiratory rate R1 at current time.
For example, R1=60/T1.
S2: extracting the electrocardiosignal by the autoregression model about breath signal built, obtains
Two breath signals, and second respiratory rate at current time is calculated according to second breath signal.
In embodiments of the present invention, when extracting using the autoregression model to the electrocardiosignal received, elder generation is needed
Autoregression model is constructed, building process is as follows:
For autoregression model AR (p), may be expressed as:
φ(B)yt=at (10)
Wherein, B is delay operator, Byt=yt-1；P is the order of model, indicates autoregression item number, ytFor time series
Current value；atFor random disturbances.Meet stationarity condition.In AR model, current time
Observe ytIt is indicated by the observation and the random disturbances at a current time of p historical juncture in past.
In embodiments of the present invention, for noise reduction, especially white noise, also autoregression can be optimized using moving average method
Model, it is assumed that the order of moving average method is q, then θ (B)=1- θ1-...-θqBq, moving average model MA (q) such as 11 institute of formula
Show, the observation y at current timetIt is indicated by the observation and the random disturbances at a current time of q historical juncture in past, yt
For the current value of time series；atFor random disturbances.Autoregression model is optimized using the model, then can be obtained and return certainly
Return-moving average model ARMA (p, q), wherein p, q are model order (p is autoregression item number, and q is sliding average item number), such as
Shown in formula 12.
yt=θ (B) at (11)
φ(B)yt=θ (B) at (12)
In embodiments of the present invention, after obtaining the auto-regressive moving-average model, mentioning for breath signal can be carried out
It takes.Specifically, auto-regressive moving-average model is a kind of method for extracting signal of blind source separating.Firstly, by estimation model
Weighting parameters, calculate ECG mixed signal (and described electrocardiosignal to be processed, which includes breath signals) ARMA (p,
Q) coefficient matrix of model, the feature as breath signal；Secondly, the feature of the breath signal obtained in conjunction with estimation, using certainly
Related separation algorithm extracts the electrocardiosignal to be processed, achievees the purpose that clean ECG signal and breath signal separation
Extraction obtains breath signal.
Referring to the waveform diagram of Fig. 5, the autoregression model to be implemented according to the present invention the second breath signal extracted.
In embodiments of the present invention, after obtaining second breath signal, the second respiratory rate R2 can be calculated, specifically
Are as follows:
By asking extremum method to find wave crest (or the wave of the second breath signal in the waveform diagram of second breath signal
Paddy), it is marked referring to the point in Fig. 5；
By extracting the time interval between two wave crests being newly generated, to obtain cycle T 2.
Real-time second respiratory rate R2 can be obtained according to sampling rate conversion.
S3: it is based on signal quality index, first breath signal and second breath signal are analyzed, obtained
And corresponding first weight factor of the first breath signal and the second weight factor corresponding with second breath signal.Tool
Body are as follows:
Power spectrumanalysis (or peak value spectrum analysis) is carried out to first breath signal and second breath signal, analysis
The Spectral structure of first breath signal and second breath signal obtains the first power corresponding with first breath signal
Repeated factor and the second weight factor corresponding with second breath signal.
S4: it according to first respiratory rate, the first weight factor, the second respiratory rate and the second weight factor, is calculated
The respiratory rate at current time.
It in a preferred embodiment, can be by first respiratory rate, the first weight factor, the second respiratory rate and
Two weight factors are weighted and averaged to obtain the respiratory rate at current time, it may be assumed that
R=μ 1*R1+ μ 2*R2 (13)
It should be noted that before being weighted and averaged need that first μ 1 and μ 2 is normalized, it is specifically, false
If μ 1+ μ 2=a, then need that μ 1 and μ 2 are normalized multiplied by normalization coefficient 1/a respectively, the μ 1+ μ 2 after guaranteeing normalization
=1.
In another embodiment, the respiratory rate that current time is calculated specifically includes:
When judging that first weight factor is greater than preset a reference value and second weight factor and is less than the benchmark
When value, it sets first respiratory rate to the respiratory rate at current time.
When second weight factor be less than a reference value when, it is believed that the signal quality of the second breath signal compared with
Difference, at this point, directly setting the first respiratory rate R1 to the respiratory rate R at current time.
When judging that first weight factor is less than preset a reference value and second weight factor and is greater than the benchmark
When value, it sets second respiratory rate to the respiratory rate at current time.
When first weight factor be less than a reference value when, it is believed that the signal quality of the first breath signal compared with
Difference, at this point, directly setting the first respiratory rate R1 to the respiratory rate R at current time.
When judging that first weight factor and second weight factor are all larger than preset a reference value, according to described
First weight factor and second weight factor are weighted summation to first respiratory rate and the second respiratory rate, calculate
To the respiratory rate at current time.
That is: R=μ 1*R1+ μ 2*R2.
If weight factor is smaller, illustrates that corresponding breath signal is second-rate, then directly remove and exhaled with second-rate
The corresponding respiratory rate of signal is inhaled, guarantee the accurate of calculated result and is stablized.
In the embodiment of the present invention, the heart is handled in such a way that neural network model is combined with auto-regressive time series technology
Electric signal obtains the first respiratory rate and the second respiratory rate, and according to the first weight factor corresponding with first respiratory rate and with
Corresponding second weight factor of second respiratory rate obtains the respiratory rate at current time, and calculated result is more acurrate reliable, and can
Mitigate measurement fluctuation or error due to caused by extraneous or environment interference, so as to obtain more accurate stable measurement
As a result.
Refering to Fig. 6, the embodiment of the present invention also provides a kind of respiratory rate extraction element 100, comprising:
Neural network extraction unit 10, for being docked by the neural network model about breath signal trained in advance
The electrocardiosignal received extracts, when obtaining the first breath signal, and being calculated current according to first breath signal
The first respiratory rate carved.
Autoregression extraction unit 20 believes the electrocardio for the autoregression model about breath signal by building
It number extracts, obtains the second breath signal, and second breathing at current time is calculated according to second breath signal
Rate.
Signal quality analytical unit 30, for being based on signal quality index, to first breath signal and described second
Breath signal is analyzed, obtain the first weight factor corresponding with first breath signal and with second breath signal
Corresponding second weight factor.
Respiratory rate computing unit 40, for according to first respiratory rate, the first weight factor, the second respiratory rate and second
The respiratory rate at current time is calculated in weight factor.
In the embodiment of the present invention, by neural network model in such a way that auto-regressive time series technology combines
Reason electrocardiosignal obtains the first respiratory rate and the second respiratory rate, and according to the first weight factor corresponding with first respiratory rate
The respiratory rate at current time is obtained with the second weight factor corresponding with second respiratory rate, calculated result is more acurrate reliable,
And can reduce measurement fluctuation or error due to caused by extraneous or environment interference, it is more accurate stable so as to obtain
Measurement result.
Preferably, the respiratory rate extraction element 100 further include:
Input sample space acquiring unit 50 calculates separately each lead electrocardio letter for receiving multi-lead electrocardiosignal
Number RR interphase and R peak amplitude, obtain input sample space, wherein the dimension in the input sample space be p, p/2 be lead more
Join the lead number of electrocardiosignal；
Principal component analysis unit 60, for based on Principal Component Analysis to the association side formed according to the input sample space
Poor matrix is handled, and principal component scores matrix is obtained；
Neural metwork training unit 70, for what is obtained with the principal component scores matrix and by impedance method synchronous acquisition
Target breath signal is training sample to neural metwork training is carried out, and obtains neural network model.
Preferably, the principal component analysis unit 60 specifically includes:
Standardization module 61, for input sample Standardization of Spatial Data processing；
Covariance matrix computing module 62, for treated that the input sample space is assisted according to data normalization
Variance matrix；
Feature calculation module 63, for calculate the covariance matrix characteristic root and feature corresponding with each characteristic root
Vector；Wherein, the quantity of the characteristic root is p, and the p characteristic root is in magnitude order；
Screening module 64, for obtaining in the p characteristic root, the sum of contribution rate is greater than the preceding m spy of predetermined threshold
Levy root；Wherein, the contribution rate of each characteristic root is equal to the value of the characteristic root divided by the sum of the value of p whole characteristic roots；
Score matrix obtains module 65, for according to feature vector corresponding with the preceding m characteristic root and described defeated
Enter sample space, obtains principal component scores matrix.
Preferably, the autoregression model is the autoregression model after moving average method optimizes.
Preferably, the respiratory rate computing unit 40 specifically includes:
First judgment module 41 judges that first weight factor is greater than preset a reference value and second power for working as
When repeated factor is less than a reference value, it sets first respiratory rate to the respiratory rate at current time；
Second judgment module 42 judges that first weight factor is less than preset a reference value and second power for working as
When repeated factor is greater than a reference value, it sets second respiratory rate to the respiratory rate at current time；
Third judgment module 43, for default when judging that first weight factor and second weight factor are all larger than
A reference value when, according to first weight factor and second weight factor to first respiratory rate and the second respiratory rate
It is weighted summation, the respiratory rate at current time is calculated.
Above disclosed is only two kinds of preferred embodiments of the invention, cannot limit the power of the present invention with this certainly
Sharp range, those skilled in the art can understand all or part of the processes for realizing the above embodiment, and weighs according to the present invention
Benefit requires made equivalent variations, still belongs to the scope covered by the invention.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the program can be stored in a computer-readable storage medium
In, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, the storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access
Memory, RAM) etc..
Claims (10)
1. a kind of respiratory rate extracting method characterized by comprising
The electrocardiosignal received is extracted by the preparatory trained neural network model about breath signal, is obtained
First breath signal, and first respiratory rate at current time is calculated according to first breath signal；
The electrocardiosignal is extracted by the autoregression model about breath signal built, obtains the second breathing letter
Number, and second respiratory rate at current time is calculated according to second breath signal；
Based on signal quality index, first breath signal and second breath signal are analyzed, obtain with it is described
Corresponding first weight factor of first breath signal and the second weight factor corresponding with second breath signal；
According to first respiratory rate, the first weight factor, the second respiratory rate and the second weight factor, current time is calculated
Respiratory rate.
2. respiratory rate extracting method according to claim 1, which is characterized in that it is described by advance it is trained about
The neural network model of breath signal extracts the electrocardiosignal received, obtains the first breath signal, and according to described
First breath signal is calculated before first respiratory rate at current time, further includes:
Multi-lead electrocardiosignal is received, the RR interphase and R peak amplitude of each lead electrocardiosignal is calculated separately, obtains input sample
Space, wherein the dimension in the input sample space is p, and p/2 is the lead number of multi-lead electrocardiosignal；
The covariance matrix formed according to the input sample space is handled based on Principal Component Analysis, obtains principal component
Score matrix；
The target breath signal obtained using the principal component scores matrix and by impedance method synchronous acquisition as training sample into
Row training, obtains neural network model.
3. respiratory rate extracting method according to claim 2, which is characterized in that
It is described that the covariance matrix formed according to the input sample space is handled based on Principal Component Analysis, it is led
Component score matrix, specifically includes:
Data normalization processing is carried out to the input sample space；
According to data normalization, treated that the input sample space obtains covariance matrix；
Calculate the covariance matrix characteristic root and feature vector corresponding with each characteristic root；Wherein, the characteristic root
Quantity is p, and the p characteristic root is in magnitude order；
It obtains in the p characteristic root, the sum of contribution rate is greater than the preceding m characteristic root of predetermined threshold；Wherein, each characteristic root
Contribution rate be equal to the value of the characteristic root divided by the sum of the value of p whole characteristic roots；
According to feature vector corresponding with the preceding m characteristic root and the input sample space, principal component scores square is obtained
Battle array.
4. respiratory rate extracting method according to claim 1, which is characterized in that the autoregression model is flat by sliding
Autoregression model after equal method optimization.
5. respiratory rate extracting method according to any one of claims 1 to 4, which is characterized in that described according to described
One respiratory rate, the first weight factor, the second respiratory rate and the second weight factor, the respiratory rate that current time is calculated are specific
Are as follows:
When judging that first weight factor is greater than preset a reference value and second weight factor is less than a reference value,
Set first respiratory rate to the respiratory rate at current time；
When judging that first weight factor is less than preset a reference value and second weight factor is greater than a reference value,
Set second respiratory rate to the respiratory rate at current time；
When judging that first weight factor and second weight factor are all larger than preset a reference value, according to described first
Weight factor and second weight factor are weighted summation to first respiratory rate and the second respiratory rate, are calculated and work as
The respiratory rate at preceding moment.
6. a kind of respiratory rate extraction element characterized by comprising
Neural network extraction unit, for by the preparatory trained neural network model about breath signal to receiving
Electrocardiosignal extracts, and obtains the first breath signal, and be calculated the of current time according to first breath signal
One respiratory rate；
Autoregression extraction unit carries out the electrocardiosignal for the autoregression model about breath signal by building
It extracts, obtains the second breath signal, and second respiratory rate at current time is calculated according to second breath signal；
Signal quality analytical unit believes first breath signal and second breathing for being based on signal quality index
It number is analyzed, is obtained and corresponding first weight factor of first breath signal and corresponding with second breath signal
Second weight factor；
Respiratory rate computing unit, for according to first respiratory rate, the first weight factor, the second respiratory rate and the second weight because
The respiratory rate at current time is calculated in son.
7. respiratory rate extraction element according to claim 6, which is characterized in that further include:
Input sample space acquiring unit calculates separately the RR of each lead electrocardiosignal for receiving multi-lead electrocardiosignal
Interphase and R peak amplitude, obtain input sample space, wherein the dimension in the input sample space is p, and p/2 is multi-lead electrocardio
The lead number of signal；
Principal component analysis unit, for based on Principal Component Analysis to the covariance matrix formed according to the input sample space
It is handled, obtains principal component scores matrix；
Neural metwork training unit, for being exhaled with the principal component scores matrix and by the target that impedance method synchronous acquisition obtains
Inhaling signal is training sample to neural metwork training is carried out, and obtains neural network model.
8. respiratory rate extraction element according to claim 7, which is characterized in that the principal component analysis unit specifically wraps
It includes:
Standardization module, for input sample Standardization of Spatial Data processing；
Covariance matrix computing module, for treated that the input sample space obtains covariance square according to data normalization
Battle array；
Feature calculation module, for calculate the covariance matrix characteristic root and feature vector corresponding with each characteristic root；
Wherein, the quantity of the characteristic root is p, and the p characteristic root is in magnitude order；
Screening module, for obtaining in the p characteristic root, the sum of contribution rate is greater than the preceding m characteristic root of predetermined threshold；Its
In, the contribution rate of each characteristic root is equal to the value of the characteristic root divided by the sum of the value of p whole characteristic roots；
Score matrix obtains module, for basis and the corresponding feature vector of preceding m characteristic root and the input sample
Space obtains principal component scores matrix.
9. respiratory rate extraction element according to claim 6, which is characterized in that the autoregression model is flat by sliding
Autoregression model after equal method optimization.
10. according to respiratory rate extraction element described in claim 6 to 9 any one, which is characterized in that the respiratory rate calculates
Unit specifically includes:
First judgment module judges that first weight factor is greater than preset a reference value and second weight factor for working as
When less than a reference value, it sets first respiratory rate to the respiratory rate at current time；
Second judgment module judges that first weight factor is less than preset a reference value and second weight factor for working as
When greater than a reference value, it sets second respiratory rate to the respiratory rate at current time；
Third judgment module judges that first weight factor and second weight factor are all larger than preset benchmark for working as
When value, first respiratory rate and the second respiratory rate are added according to first weight factor and second weight factor
Power summation, is calculated the respiratory rate at current time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610899933.8A CN106388824B (en) | 2016-10-14 | 2016-10-14 | Respiration rate extraction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610899933.8A CN106388824B (en) | 2016-10-14 | 2016-10-14 | Respiration rate extraction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106388824A CN106388824A (en) | 2017-02-15 |
| CN106388824B true CN106388824B (en) | 2019-04-16 |
Family
ID=58011630
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610899933.8A Active CN106388824B (en) | 2016-10-14 | 2016-10-14 | Respiration rate extraction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106388824B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106901694A (en) * | 2017-02-20 | 2017-06-30 | 广州视源电子科技股份有限公司 | Respiration rate extraction method and device |
| CN109009125B (en) * | 2018-06-07 | 2019-05-28 | 上海交通大学 | Driver's fine granularity monitoring of respiration method and system based on audio frequency of mobile terminal |
| CN114190951B (en) * | 2021-12-31 | 2023-11-03 | 上海联影医疗科技股份有限公司 | ECG signal processing method, system, device and medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101991418A (en) * | 2009-08-14 | 2011-03-30 | 深圳市理邦精密仪器股份有限公司 | Method for improving respiratory rate detection accuracy |
| CN105769122A (en) * | 2016-03-16 | 2016-07-20 | 中国科学院电子学研究所 | Method for measuring sleep apnea hyponea index |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060058590A1 (en) * | 2004-08-24 | 2006-03-16 | Shaw Geoffrey M | Method and system for assaying agitation |
-
2016
- 2016-10-14 CN CN201610899933.8A patent/CN106388824B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101991418A (en) * | 2009-08-14 | 2011-03-30 | 深圳市理邦精密仪器股份有限公司 | Method for improving respiratory rate detection accuracy |
| CN105769122A (en) * | 2016-03-16 | 2016-07-20 | 中国科学院电子学研究所 | Method for measuring sleep apnea hyponea index |
Non-Patent Citations (3)
| Title |
|---|
| 《基于ARM的4个生理参数的快速测量系统》;郎晓华;《山东大学 硕士论文》;20110520;全文 |
| 《基于光电容积脉搏波的呼吸频率监测》;王跃俊;《北京理工大学 硕士论文》;20150101;全文 |
| 《基于心电和脉搏波数据融合的呼吸率估计》;邓宝芸;《山东大学 硕士论文》;20120418;全文 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106388824A (en) | 2017-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106388825B (en) | Respiration rate extraction method and device | |
| CN106108889B (en) | Electrocardiogram classification method based on deep learning algorithm | |
| US9042973B2 (en) | Apparatus and method for measuring physiological signal quality | |
| CN101801263B (en) | Monitoring physiological condition and detecting abnormalities | |
| US7225013B2 (en) | Adaptive prediction of changes of physiological/pathological states using processing of biomedical signals | |
| CN106073755A (en) | The implementation method that in a kind of miniature holter devices, atrial fibrillation identifies automatically | |
| CN106073784B (en) | Respiration rate extraction method and device | |
| CN107137071A (en) | It is a kind of to analyze the method that heart impact signal is used for calculating short-term heart beat value | |
| CN109152529A (en) | More Disease management | |
| CN106066933A (en) | The method obtaining atrial fibrillation identification artificial neural network weighted value matrix | |
| WO2009100133A1 (en) | System, method and computer program product for detection of changes in health status and risk of imminent illness | |
| Risk et al. | Beat detection and classification of ECG using self organizing maps | |
| CN103961089A (en) | Sinus heart rate turbulence tendency detecting method based on segmented straight line fitting | |
| CN106388824B (en) | Respiration rate extraction method and device | |
| CN111000551A (en) | Heart disease risk diagnosis method based on deep convolutional neural network model | |
| CN106901694A (en) | Respiration rate extraction method and device | |
| CN110840401A (en) | Heart rate value measuring method, device and equipment and readable storage medium | |
| WO2009127799A1 (en) | Method and apparatus for measuring breathing rate | |
| Paliakaitė et al. | Modeling of artifacts in the wrist photoplethysmogram: Application to the detection of life-threatening arrhythmias | |
| CN106580324B (en) | Respiratory signal extraction method and device | |
| CN106955101A (en) | The method and device of breath signal is extracted from electrocardiosignal | |
| CN106725487A (en) | Respiration rate extraction method and device | |
| CN118747325A (en) | A method for evaluating the quality of electrocardiogram lead data | |
| Akhbari et al. | Twave alternans detection in ecg using Extended Kalman Filter and dualrate EKF | |
| Sanamdikar et al. | Extraction of different features of ECG signal for detection of cardiac arrhythmias by using wavelet transformation Db 6 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |