CN114912478A - Interference detection method for pulse wave signals - Google Patents
Interference detection method for pulse wave signals Download PDFInfo
- Publication number
- CN114912478A CN114912478A CN202210340616.8A CN202210340616A CN114912478A CN 114912478 A CN114912478 A CN 114912478A CN 202210340616 A CN202210340616 A CN 202210340616A CN 114912478 A CN114912478 A CN 114912478A
- Authority
- CN
- China
- Prior art keywords
- interference
- signal
- score
- point
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Biophysics (AREA)
- Heart & Thoracic Surgery (AREA)
- Psychiatry (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Physiology (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
Abstract
A method for detecting disturbance of a pulse wave signal, the method comprising the steps of: 1) data acquisition and preprocessing: (1) slicing data: (2) baseline wander removal: removing the baseline drift of each slice data by adopting a mean value baseline drift removing method, (3) smoothing: the PPG signal is often affected by high frequency interference, random noise or low perfusion, and the baseline shift algorithm cannot suppress these interferences, so it is important to smooth each signal point. (4) Normalization: in order to reduce the influence of PPG signals of different acquisition devices and different dimensions on the algorithm, maximum and minimum normalization processing needs to be carried out on each signal after baseline drift is removed, so that the original form of data can be reserved, and the influence of the dimensions can be eliminated. The invention can accurately determine whether the positioning signal of each signal point has interference or not, the signal interference level and realize the splicing of effective signals.
Description
Technical Field
The invention relates to an interference detection method of pulse wave signals, and belongs to the technical field of medical signal processing.
Background
Photoplethysmography (PPG) is one of important physiological signals of a human body, contains rich microcirculation physiological and pathological information, intuitively reflects the state of pulsatile transmission of blood in a blood vessel of the human body caused by each heart beat, and is an important information source for researching the circulatory system of the human body. The application field of the non-invasive detection probe is developed from a human body circulatory system to a respiratory system, and the non-invasive detection probe comprises various non-invasive detection applications such as blood oxygen, cerebral oxygen, muscle oxygen, blood pressure, blood flow, blood sugar, pulse rate, respiratory volume and the like. At present, the PPG signal acquisition technology is relatively mature, and the PPG signal acquisition technology has the advantages of simplicity in operation, low cost, safety, reliability and the like, can realize continuous PPG signal monitoring, and has better measurement experience.
However, the utilization of PPG signals by various existing medical instrument devices and smart wearable devices includes only a few parameters such as blood oxygen, pulse rate, respiration rate, etc. Most of the reasons are that the PPG signal has a lot of noises, including the interference of various factors such as ambient light, motion artifact, myoelectricity interference, power frequency interference, baseline drift, high frequency interference, random noise, and low perfusion. Therefore, it is necessary to develop an effective method for detecting interference of the photoplethysmography signals to improve the application capability of the PPG signals. The patent develops a set of full-chain interference detection method for signal splicing from the beginning of an original PPG signal to the end of interference detection, and can reduce the influence of interference signals on physiological parameter detection to a great extent.
Disclosure of Invention
The invention aims to provide an interference detection method for a photoplethysmography signal, which can be used for accurately detecting whether interference exists in a positioning signal of each signal point or not, and realizing signal interference level and splicing effective signals.
The invention is realized by the following technical scheme: a method for detecting interference of pulse wave signals, which is accurate to the existence of interference in positioning signals of each signal point, the signal interference level and the splicing of effective signals, comprises the following steps:
1) data acquisition and preprocessing:
(1) slicing data:
at the beginning of processing the PPG signal, slicing the PPG data according to the length of alpha second, and increasing the slice to the length of beta second along with the time lapse, wherein specifically, when the length of the PPG signal to be subjected to interference detection is less than alpha second, the data is not subjected to interference detection; over time, the data length of each slice is then α seconds, α +1 seconds, … …, β seconds, and finally the slice is sliced according to the β second length; finally, each slice data retains the latest beta second length data for iterative update, and interference detection of each second is completed;
(2) baseline wander removal:
removing the baseline drift of each slice data by adopting a mean value baseline drift removing method;
(3) smoothing:
the PPG signal is often affected by high frequency interference, random noise or low perfusion, and the baseline shift algorithm cannot suppress these interferences, so it is important to smooth each signal point;
(4) normalization:
in order to reduce the influence of PPG signals of different acquisition equipment and different dimensions on the algorithm, maximum and minimum normalization processing needs to be carried out on each signal from which the baseline drift is removed, so that the original form of the data can be reserved, and the influence of the dimensions can be eliminated;
2) estimating the RR interval of the PPG signal:
in the method, an interference detection window changes in real time according to the pulse rate corresponding to a signal, an RR interval of a PPG signal needs to be estimated during primary interference detection, and an accurate RR interval is obtained by positioning an R wave position of the PPG signal per second in a later stage;
3) interference detection
In reality, PPG signals have a lot of noises, including the interference of various factors such as ambient light, motion artifact, myoelectricity interference, random noise and low perfusion, and the noises are divided according to the interference degree and can be divided into two types: one is large interference caused by long-term motion artifact, baseline drift, electromyographic interference, and the like; the other is small interference caused by high-frequency interference, random noise and the like, wherein primary interference detection is mainly used for identifying long-term interference or sudden interference, and secondary interference detection is mainly used for identifying interference which causes signal deformation and is difficult to use due to random noise, ambient light, low perfusion and the like;
4) signal synthesis
The common synthesis method after signal interference detection is direct and simple splicing, and the method has the advantages of simple algorithm, small calculated amount, and capability of keeping original signals as much as possible, but has very obvious defects, can cause a new round of artificial interference, and is not beneficial to subsequent signal application;
preferably, the method comprises the following steps: the specific formula of baseline wander removal in the step 1) is shown in formulas (1) to (3):
y 1 (i)＝y 0 (i)-m(i)，1≤i≤l (1)
wherein wnd 1 Is a baseline drift window, typically an odd number, y 0 Is the original signal, y 1 Is the signal after removal of the baseline drift, l is the signal length.
Preferably, the method comprises the following steps: the smoothing algorithm in the step 1) is shown in formulas (4) to (5):
wherein wnd 2 Is a smooth window, typically an odd number, y 1 Is to remove the signal after baseline drift, y 2 Is the smoothed signal and l is the signal length.
Preferably, the method comprises the following steps: the normalization algorithm in the step 1) is shown as a formula (6):
where N is the range of the normalized data, b is the normalized numberAccording to the baseline, y 2 Is the smoothed signal, y 3 Is the normalized signal and l is the signal length.
Preferably, the method comprises the following steps: the estimation of the RR interval of the PPG signal in step 1) is shown in equations (7) to (13):
z 0 (j)＝i，y 4 (i)＞t1 or y 4 (i)＜t2 (8)
z 1 (j)＝z 0 (i)-z 0 (i-1) (12)
z 2 (k)＝z 1 (j)，z1(j)＞α 3 (13)
wherein, y 3 Is the normalized signal, l is the signal length, α 1 Is the coefficient of the upper threshold line, requires α 1 ＞0，α 2 Is a coefficient of a lower threshold line, requires α 2 ＜0，α 3 Is a parameter for estimating RR interval, requires alpha 3 > 0, final pair z 2 (k) Calculating a median value, wherein the evaluated value is an estimated RR interval, and the numerical value is related to the sampling rate;
z is 2 (k) The median algorithm is as follows:
the first step is as follows: obtaining array XMinimum value as lower limit X lower Maximum value as upper limit X upper ；
The second step is that: traversing each point X in the array X i Judging whether the point is at [ X ] lower ，X upper ]Within;
1. if the point is not in the range, entering the second step to judge the next point X i+1 ；
2. Otherwise, entering the third step;
the third step: traverse each point X in array X j Judgment of X i And X j The relationship of (a);
1. initializing two counters, cou 1 ＝0；cou 2 ＝0；(cou 1 Statistic less than X i Number of (c), cou 2 Statistic equals X i Number of (2)
2. Judgment of X i And X j The relationship of (1);
if i is j (i.e. not compared with itself), go to the third step and judge the next point X j+1 ；
If X is j ＜X i Then counter cou 1 Self-increasing, entering the fourth step;
if X is j ＝X i Then counter cou 2 Self-increment, enter the third step, judge the next point X j+1 ；
The fourth step: judge cou 1 Relation to p, update X upper ；
1. If cou 1 ＞(p-1)，
If X is upper ＞X i Then update X upper From X i Assigning, entering the second step, and judging the next point X i+1 ；
Otherwise, entering the second step to judge the next point X i+1 ；
2. Otherwise, entering the third step and judging the next point X j+1 ；
The fifth step: judge cou 1 、cou 1 +cou 2 Relation to p, update X lower Returning a function result;
1. if cou 1 If (p-1), then the function result X is returned i ；
2. If cou 1 < (p-1), but ku 1 +cou 2 ≧ p-1, the function result X is returned i ；
3. If cou 1 +cou 2 ＜(p-1)，
If X is lower ＜X i Then update X lower From X i And (4) assigning values. Entering the second step, judging the next point X i+1 ；
Otherwise, entering the second step to judge the next point X i+1 。
Preferably, the method comprises the following steps: the interference detection in the step 3) mainly comprises the following steps:
1) primary interference detection
In order to detect long-term or abrupt interference as much as possible in the interference primary screening process, the primary interference detection part comprises three layers of interference detection algorithms and respectively identifies a long-term interference signal, an approximate horizontal interference signal and an abrupt slope interference signal;
2) composite interference score
The interference detection algorithm is divided into two stages, interference detection algorithms with different functions can be developed in each stage, and how interference scores after each interference detection algorithm are fused has important influence on the performance of the whole algorithm. Considering that different interference scores have different ranges of influence on signals, blind simple addition and subtraction fusion can influence the effect of the algorithm to a certain extent.
3) Secondary interference detection
After primary interference detection, large variable amplitude interference, horizontal interference and slope abrupt change interference are basically detected, and detected signals can be basically used for calculating average pulse rate, but the calculation of functions of blood oxygen saturation, blood pressure and the like is still difficult. For this reason, secondary interference detection has been developed, mainly for identifying small amplitude interference signals.
4) Signal synthesis
A common synthesis method after signal interference detection is direct and simple splicing, and the method has the advantages of simple algorithm, small calculation amount and capability of keeping original signals as much as possible. But the disadvantage is also very obvious, which can cause a new round of artificial interference and is not beneficial to the subsequent signal application.
Preferably, the method comprises the following steps: the step of respectively identifying the long-time interference signal, the approximate horizontal interference signal and the slope abrupt change interference signal is respectively calculated as follows:
identifying long-term amplitude interference signals
The specific calculation formula is shown in formulas (15) - (24).
z 2 (j)＝i，z 1 (i)＞t1 or z 1 (i)＜t2 (23)
Wherein, y 3 Is the normalized signal, l is the signal length, wnd s Is an interference detection window, M (x) is an optimized median algorithm, which is detailed in step 2) to predict the RR interval of the PPG signal, alpha 1 Is the coefficient of the upper threshold line, requires α 1 ＞0，α 2 Is the coefficient of the lower threshold line, requires α 2 ＞0。z 1 Is data converted by interference detection, z 2 Is a long-time amplitude interference signal detection point.
The equations (15) to (24) are generally divided into 5 steps:
step 1: equations (15) - (17) convert the original signal to increase the variation amplitude of the interference signal, so as to prepare for subsequent interference detection;
step 2: equations (18) - (19), calculating the mean value of the converted signal according to the window;
and 3, step 3: equations (20) - (21), calculating standard deviation for the converted signal according to the window;
and 4, step 4: equation (22), construct a new signal for interference detection;
and 5, step 5: and (5) in formulas (23) to (24), searching a threshold line and acquiring a long-time variable amplitude interference position. According to different alpha 1 And alpha 2 Constructing an interference Score array Score of the signal 1 ；
Identifying short-term horizontal interfering signals
The specific calculation formula is shown in formulas (25) to (28):
z 5 (j)＝i，z 4 (i)＜t3 (27)
t3＝α 3 *M(z 5 )，α 3 ＞0 (28)
wherein, y 3 Is the normalized signal, l is the signal length, wnd 4 Is an interference detection window, M (x) is an optimized median algorithm, which is detailed in step 2) to predict the RR interval of the PPG signal, alpha 3 Is the coefficient of the upper threshold line, requires α 3 ＞0。z 4 Is data converted by interference detection, z 5 Is a short-time horizontal interference signal detection point;
equations (25) - (28) are generally divided into 3 steps:
step 1: a formula (25) is used for constructing the square of the slope of the adjacent points, and is used for increasing the difference of interference signals and reducing the disturbance of normal signals;
step 2: equation (26), construct a new signal for horizontal interference detection;
and 3, step 3: and (5) formulas (27) - (28) for finding a threshold line and acquiring a horizontal interference position. According to different alpha 3 Constructing an interference Score array Score of the signal 2 ；
Identifying abrupt slope interference signals
The specific calculation formula is shown in formulas (29) to (36).
z 9 (j)＝i，z 6 (i)＞t4 or z 6 (i)＜t5 (31)
Wherein, y 3 Is the normalized signal, l is the signal length, α 4 Is the coefficient of the upper threshold line, requires α 4 ＞0，α 5 Is the coefficient of the lower threshold line, requires α 5 ＞0，z 6 Is data converted by interference detection, z 9 Is a slope abrupt change interference signal detection point;
equations (29) to (36) are generally divided into 2 steps:
step 1: formula (29), construct a new signal for slope mutation interference detection;
step 2: and (5) searching a threshold line and obtaining the slope mutation interference position according to the formulas (30) to (36). According to different alpha 4 And alpha 5 Constructing an interference Score array Score of the signal 3 ；
The three abnormal interference Score array Score 1 、Score 2 And Score 3 Calculating the final interference Score array Score of the first-level interference detection algorithm 4 And according to the normalized signal y before interference detection 3 Constructing a signal y after primary interference detection 6 For a secondary interference detection algorithm;
the specific construction method comprises the following steps: determining whether the window meets the requirement of interference score according to the detection granularity of the interference score array, and if so, retaining the section of signal; otherwise, setting the segment signal as
Wherein, N and b are parameters in the normalization process, and are detailed in formula (6).
Preferably, the method comprises the following steps: the fusion algorithm in the step 2) comprises the following steps:
the first step is as follows: traverse each point X in array X i Judging whether the point is an interference score cur _ score to be fused or not;
1. if the point X is i If the interference score cur _ score is not to be fused, the first step is entered to judge the next point X i+1 ；
2. Otherwise, entering the second step;
the second step is that: traverse X i In the interference radiation window [ X ] i-wnd ，X i+wnd ]Each point X in j Update Y j ；
1. If X is j ＜cur_score，
If Y is j If < cur _ score, Y is updated j (Y j ＝β 1 *cur_score+β 2 *Y j ) Entering the second step to judge the next point X j+1 ；
Otherwise, entering the second step to judge the next point X j+1 ；
2. If not, then the mobile terminal can be switched to the normal mode,
if Y is j ＜X j Then, Y is updated j From X j And (4) assigning values. Entering the second step, judging the next point X j+1 ；
Otherwise, entering the second step to judge the next point X j+1 ；
Wherein, beta 1 And beta 2 Is a weighting coefficient, satisfies beta 1 +β 2 ＝1，β 1 ＞0，β 2 And if the value is more than 0, when the interference score fusion function Q is called for the first time, the fused interference score array Y is given an initial value by the interference score array X to be fused.
Preferably, the method comprises the following steps: the secondary interference detection algorithm comprises the following steps:
the first step is as follows: traversing each point X in the array X i First evaluation X i Whether or not toIs a candidate peak point;
1. if X is i ≤X i-1 Or X i ＜X i+1 Then, go to the first step to judge the next point X i+1 ；
2. Otherwise, initializing a counter: 0 is used for counting less than or equal to X i The number of the second step is entered;
the second step is that: traverse X i Left and right side window [ X ] i-wnd1 ，X i+wnd2 ]Each point X in j Judgment of X i And X j Second evaluation of X i Whether it is a candidate peak point;
1. if Score j Not more than thre, and X j ≤X i If so, the counter cou is automatically increased, and the second step is entered to judge the next point X j+1 ；
2. Otherwise, entering the first step and judging the next point X i+1 ；
The third step: traverse X i Right side additional window [ X ] i+wnd2+1 ，X i+wnd2+1+wnd3 ]Each point X in k Judgment of X i And X k A third evaluation of X i Whether it is a candidate peak point;
1. if Score k Not more than thre, and X k ＜X i If yes, the position i is saved to the peak position array Y, and the first step is entered to judge the next point X i+1 ；
2. If not, then,
if X k ＞X i Then, go to the first step to judge the next point X i+1 ；
Otherwise, entering the third step and judging the next point X k+1 ；
Therein, the window wnd 1 、wnd 2 、wnd 3 Both are related to pulse rate, which is typically 1/4;
the fourth step: positioning valley points, wherein the valley positioning algorithm is based on peak algorithm positioning, and the position of the minimum value between peak points is the valley point; the fifth step: calculating the standard deviation of a curve formed by the peak point and the valley point;
and a sixth step: root of herbaceous plantCalculating a peak interference Score array Score according to the standard deviation and different threshold lines 5 And valley interference Score array Score 6 ；
The seventh step: construction of two-level interference algorithm comprehensive Score array 1-Score 7 Synthesizing the Score array Score according to a first-level interference algorithm 4 Score array of sum-peak interference scores 5 Constructing a secondary interference algorithm comprehensive Score array 1-Score by using an interference Score function Q 7 ；
Eighth step: 2-Score for constructing comprehensive Score array of secondary interference algorithm 8 And synthesizing the Score array 1-Score according to a secondary interference algorithm 7 And valley interference Score array Score 6 Constructing a secondary interference algorithm comprehensive Score array 2-Score by using an interference Score function Q 8 ；
The ninth step: construction of post-secondary interference detection signal y 8 And synthesizing the Score array 2-Score according to a secondary interference algorithm 8 And the smoothed signal y 2 Constructing a signal y after secondary interference detection by using a signal synthesis function 8 ；
Preferably, the method comprises the following steps: the signal synthesis method comprises the following steps:
the first step is as follows: traversing each Score in the Score array Score i To discriminate Score i Whether a valid signal score is satisfied;
1. if Score i If more than thre, the first step is entered to judge the next Score i+1 ；
2. Otherwise, entering the second step;
the second step: synthesizing signals;
1. if Score i-1 If more than thre, entering the third step;
2. if Score i+1 If yes, entering the fifth step;
3. otherwise, the segment signal [ X ] is retained (i-1)*wnd1+1 ：X i*wnd1 ]Into signal Y;
the third step: traverse the Score array [ Score ] i+1 ：Score i+wnd4 ]Each Score within Score j Judging Score j Whether or not to satisfy the requirements ofA valid signal score;
1. if Score j If more than thre, the first step is entered to judge the next Score i+1 ；
2. Otherwise, entering the third step and judging the next Score j+1 ；
The fourth step: synthesizing signal segments before an interference point;
1. saving signal segments [ X ] (i-1)wnd1-wnd2 ：X (i-1)*wnd1 ]The signal from the signal at the position of the maximum value to the signal section end is added into the signal Y;
2. hold the segment signal [ X (i-1)*wnd1+1 ：X i*wnd1 ]Into signal Y;
3. entering the first step, judging the next Score i+1 ；
The fifth step: traverse the Score array [ Score ] i-wnd4 ：Score i-1 ]Each Score within Score k Judging Score k Whether a valid signal score is satisfied;
1. if Score k If more than thre, the first step is entered to judge the next Score i+1 ；
2. Otherwise, the fifth step is carried out, and the next Score is judged k+1 ；
And a sixth step: signal segment synthesis after the interference point;
1. hold the segment signal [ X (i-1)*wnd1+1 ：X i*wnd1 ]Into signal Y;
2. saving signal segments [ X ] i*wnd1+1 ：X i*wnd1+wnd3 ]Starting from the signal section to the position where the maximum value is located, and sending the signal to a signal Y;
3. entering the first step, judging the next Score i+1 ；
Therein, the window wnd 1 Is the interference detection granularity, which is related to the sampling rate, and is generally 0.5s Fs or 1s Fs; window wnd 2 、wnd 3 The granularity of the interference signal is reserved, and is related to the sampling rate, and is generally 0.5s x Fs; window wnd 4 Related to the interference detection granularity, generally 5 granularity sizes are taken, and after signal synthesis, the signal is directly synthesizedThe signal may be used as a valid signal for the period of time.
The interference detection method of the photoplethysmography signals can accurately determine whether the positioning signals of each signal point have interference or not, the signal interference level and realize the splicing of effective signals.
Drawings
FIG. 1 is a flow chart of primary interference detection in the present invention;
FIG. 2 is a flow chart of secondary interference detection in the present invention;
FIG. 3 is the raw 3minPPG signal;
FIG. 4 shows the interference algorithm 30 sPG signal at the 1 st entry;
FIG. 5 is the 2 nd entry interference algorithm 31 sPGP signal;
FIG. 6 shows the 151 th entry into the interference algorithm 60 sPGS signal;
FIG. 7 illustrates the interference algorithm 30 sPGS signal removal baseline wander for the 1 st entry;
FIG. 8 is a graph of the 151 th entry into the interference algorithm 60 sPGS signal with baseline drift removed;
fig. 9 shows the interference algorithm 30sPPG signal smoothing process for the 1 st entry;
FIG. 10 shows the 151 th entry into the interference algorithm 60 sPGS signal smoothing process;
fig. 11 illustrates the normalization process of the interference algorithm 30sPPG signal at the 1 st entry;
FIG. 12 illustrates the 151 th entry interference algorithm 60 sPGS signal normalization process;
FIG. 13 shows the estimated RR interval-equation (8);
FIG. 14 shows the estimated RR interval-equation (12);
FIG. 15 shows the estimated RR interval-equation (13);
FIG. 16 is a segment of an interference detection signal;
FIG. 17 is a first order interference-long term amplitude interference detection;
fig. 18 is a primary interference-short time horizontal interference detection;
FIG. 19 is a first order interference-slope mutation interference detection;
fig. 20 is a primary interference detection-composite interference score;
FIG. 21 shows primary interference detection results;
FIG. 22 is a secondary interference detection-composite score of 1;
FIG. 23 is a secondary interference detection-composite score of 2;
FIG. 24 is a secondary interference detection-composite score;
FIG. 25 shows secondary interference detection results;
fig. 26 shows signal synthesis after interference detection.
Detailed Description
The invention will be described in detail below with reference to the following figures: as shown in fig. 1, a method for detecting interference of pulse wave signals, which is accurate to whether there is interference in a positioning signal of each signal point, a signal interference level, and realizes splicing of effective signals, comprises the following steps:
2) data acquisition and preprocessing:
(1) slicing data:
at the beginning of processing the PPG signal, slicing the PPG data according to the length of alpha second, and increasing the length of the PPG data to the length of beta second slice along with the time lapse, wherein specifically, when the length of the PPG signal to be subjected to interference detection is less than alpha second, the interference detection is not carried out on the data; over time, each slice data length is then α seconds, α +1 seconds, … …, β seconds, eventually slicing at β seconds length; finally, each slice data retains the latest beta second length data for iterative updating, and the interference detection of each second is completed;
(2) baseline wander removal:
removing the baseline drift of each slice data by adopting a mean value baseline drift removing method;
(3) smoothing:
the PPG signal is often affected by high frequency interference, random noise or low perfusion, and the baseline shift algorithm cannot suppress these interferences, so it is important to smooth each signal point;
(4) normalization:
in order to reduce the influence of PPG signals of different acquisition equipment and different dimensions on the algorithm, maximum and minimum normalization processing needs to be carried out on each signal after baseline drift is removed, so that the original form of data can be reserved, and the influence of the dimensions can be eliminated;
2) estimating the RR interval of the PPG signal:
in the method, an interference detection window changes in real time according to the pulse rate corresponding to a signal, an RR interval of a PPG signal needs to be estimated when interference detection is carried out for the first time, and an accurate RR interval is obtained through the R wave position for positioning the PPG signal per second in the later period;
3) interference detection
In reality, PPG signals have a lot of noises, including the interference of various factors such as ambient light, motion artifact, myoelectricity interference, random noise and low perfusion, and the noises are divided according to the interference degree and can be divided into two types: one is large interference caused by long-term motion artifact, baseline drift, electromyographic interference, and the like; the other is small interference caused by high-frequency interference, random noise and the like, wherein primary interference detection is mainly used for identifying long-term interference or sudden interference, and secondary interference detection is mainly used for identifying interference which causes signal deformation and is difficult to use due to random noise, ambient light, low perfusion and the like;
5) signal synthesis
The common synthesis method after signal interference detection is direct and simple splicing, and the method has the advantages of simple algorithm, small calculated amount, and capability of keeping original signals as much as possible, but has very obvious defects, can cause a new round of artificial interference, and is not beneficial to subsequent signal application;
preferably, the method comprises the following steps: the specific formula of baseline drift removal in the step 1) is shown in formulas (1) to (3):
y 1 (i)＝y 0 (i)-m(i)，1≤i≤l (1)
wherein wnd 1 Is a baseline drift window, typically an odd number, y 0 Is the original signal, y 1 Is the signal after removal of the baseline drift, l is the signal length.
The smoothing algorithm in the step 1) is shown in formulas (4) to (5):
wherein wnd 2 Is a smooth window, typically odd, y 1 Is to remove the signal after baseline drift, y 2 Is the smoothed signal and l is the signal length.
The normalization algorithm in the step 1) is shown as a formula (6):
where N is the range of the normalized data, b is the baseline of the normalized data, y 2 Is the smoothed signal, y 3 Is the normalized signal and l is the signal length.
The estimation of the RR interval of the PPG signal in step 1) is shown in equations (7) to (13):
z 0 (j)＝i，y 4 (i)＞t1 or y 4 (i)＜t2 (8)
z 1 (j)＝z 0 (i)-z 0 (i-1) (12)
z 2 (k)＝z 1 (j)，z 1 (j)＞α 3 (13)
wherein, y 3 Is the normalized signal, l is the signal length, α 1 Is the coefficient of the upper threshold line, requires α 1 ＞0，α 2 Is the coefficient of the lower threshold line, requires α 2 ＜0，α 3 Is a parameter for estimating RR interval, requires alpha 3 > 0, final pair z 2 (k) Calculating a median value, wherein the evaluated value is an estimated RR interval, and the numerical value is related to the sampling rate;
z is 2 (k) The median algorithm is as follows:
the first step is as follows: obtaining the minimum value of the array X as the lower limit X lower Maximum value as upper limit X upper ；
The second step: traverse each point X in array X i Judging whether the point is [ X ] or not lower ，X upper ]Within;
1. if the point is not within the range, go to the second step to determine the next point X i+1 ；
2. Otherwise, entering the third step;
the third step: traverse each point X in array X j Judgment of X i And X j The relationship of (1);
1. initializing two counters, cou 1 ＝0；cou 2 ＝0；(cou 1 Statistic is less than X i Number of (c), cou 2 Statistic equals X i Number of (2)
2. Judgment of X i And X j The relationship of (1);
if i is j (i.e. not compared with itself), go to the third step and judge the next point X j+1 ；
If X j ＜X i Then counter cou 1 Self-increasing, entering the fourth step;
if X is j ＝X i Then counter cou 2 Self-increment, enter the third step, judge the next point X j+1 ；
The fourth step: judge cou 1 Relation to p, update X upper ；
1. If cou 1 ＞(p-1)，
If X is upper ＞X i Then update X upper From X i Assigning, entering the second step, and judging the next point X i+1 ；
Otherwise, entering the second step to judge the next point X i+1 ；
2. Otherwise, entering the third step and judging the next point X j+1 ；
The fifth step: judge cou 1 、cou 1 +cou 2 Relation to p, update X lower Returning a function result;
1. if cou 1 If (p-1), then the result of the function X is returned i ；
2. If cou 1 < (p-1), but ku 1 +cou 2 ≧ p-1, the function result X is returned i ；
3. If cou 1 +cou 2 ＜(p-1)，
If X is lower ＜X i Then update X lower From X i And (4) assigning values. Entering the second step, judging the next point X i+1 ；
Otherwise, entering the second step to judge the next point X i+1 。
The interference detection in the step 3) mainly comprises the following steps:
1) primary interference detection
In order to detect long-term or abrupt interference as much as possible in the interference primary screening process, the primary interference detection part comprises three layers of interference detection algorithms and respectively identifies a long-term interference signal, an approximate horizontal interference signal and an abrupt slope interference signal;
2) composite interference score
The interference detection algorithm is divided into two stages, interference detection algorithms with different functions can be developed in each stage, and how interference scores after each interference detection algorithm are fused has important influence on the performance of the whole algorithm. Considering that different interference scores have different ranges of influence on signals, blind simple addition and subtraction fusion can influence the effect of the algorithm to a certain extent.
3) Secondary interference detection
After primary interference detection, large variable amplitude interference, horizontal interference and slope abrupt change interference are basically detected, and detected signals can be basically used for calculating average pulse rate, but the calculation of functions of blood oxygen saturation, blood pressure and the like is still difficult. For this reason, secondary interference detection has been developed, mainly for identifying small amplitude interference signals.
4) Signal synthesis
The common synthesis method after signal interference detection is direct and simple splicing, and the method has the advantages of simple algorithm, small calculation amount and capability of keeping the original signal as much as possible. But the disadvantage is also very obvious, which can cause a new round of artificial interference and is not beneficial to the subsequent signal application.
The step of respectively identifying the long-time interference signal, the approximate horizontal interference signal and the slope abrupt change interference signal is respectively calculated as follows: identifying long-term amplitude interference signals
The specific calculation formula is shown in formulas (15) - (24).
z 2 (j)＝i，z 1 (i)＞t1 or z 1 (i)＜t2 (23)
Wherein, y 3 Is the normalized signal, l is the signal length, wnd 3 Is an interference detection window, M (x) is an optimized median algorithm, which is detailed in step 2) to predict the RR interval of the PPG signal, alpha 1 Is the coefficient of the upper threshold line, requires α 1 ＞0，α 2 Is the coefficient of the lower threshold line, requires α 2 ＞0。z 1 Is data converted by interference detection, z 2 Is a long-time amplitude interference signal detection point.
The equations (15) to (24) are generally divided into 5 steps:
step 1: equations (15) - (17) convert the original signal to increase the variation amplitude of the interference signal, so as to prepare for subsequent interference detection;
step 2: equations (18) - (19), calculating the mean value of the converted signal according to the window;
and 3, step 3: equations (20) - (21), calculating standard deviation for the converted signal according to the window;
and 4, step 4: equation (22), construct a new signal for interference detection;
and 5, step 5: and (5) in formulas (23) to (24), searching a threshold line and acquiring a long-time variable amplitude interference position. According to different alpha 1 And alpha 2 Constructing an interference Score array Score of the signal 1 ；
Identifying short-term horizontal interfering signals
The specific calculation formula is shown in formulas (25) to (28):
z 5 (j)＝i，z 4 (i)＜t3 (27)
t3＝α 3 *M(z 5 )，α 3 ＞0 (28)
wherein, y 3 Is normalized signal, l is signal length, wnd 4 Is an interference detection window, M (x) is an optimized median algorithm, which is detailed in step 2) to predict the RR interval of the PPG signal, alpha 3 Is the coefficient of the upper threshold line, requires α 3 ＞0。z 4 Is data converted by interference detection, z 5 Is a short-time horizontal interference signal detection point;
the equations (25) to (28) are generally divided into 3 steps:
step 1: a formula (25) is used for constructing the square of the slope of the adjacent points, and is used for increasing the difference of interference signals and reducing the disturbance of normal signals;
step 2: equation (26), construct a new signal for horizontal interference detection;
and 3, step 3: and (5) formulas (27) - (28) for finding a threshold line and acquiring a horizontal interference position. According to different alpha 3 Constructing an interference Score array Score for the signal 2 ；
Identifying abrupt slope interference signals
The specific calculation formula is shown in formulas (29) to (36).
z 9 (j)＝i，z 6 (i)＞t4 or z 6 (i)＜t5 (31)
Wherein, y 3 Is the normalized signal, l is the signal length, α 4 Is the coefficient of the upper threshold line, requires α 4 ＞0，α 5 Is the coefficient of the lower threshold line, requires α 5 ＞0，z 6 Is data converted by interference detection, z 9 Is a slope abrupt change interference signal detection point;
equations (29) to (36) are generally divided into 2 steps:
step 1: formula (29), construct a new signal for slope mutation interference detection;
step 2: and (5) searching a threshold line and obtaining the slope mutation interference position according to the formulas (30) to (36). According to different alpha 4 And alpha 5 Constructing an interference Score array Score of the signal 3 ；
The three abnormal interference score arrays Scoe 1 、Score 2 And Score 3 Calculating the final interference Score array Score of the first-level interference detection algorithm 4 And according to the normalized signal y before interference detection 3 Constructing a signal y after primary interference detection 6 For a secondary interference detection algorithm;
the specific construction method comprises the following steps: determining whether the window meets the requirement of interference score according to the detection granularity of the interference score array, and if so, retaining the section of signal; otherwise, setting the segment signal as
Wherein, N and b are parameters in the normalization process, and are detailed in formula (6).
The fusion algorithm in the step 2) comprises the following steps:
the first step is as follows: traverse each point X in array X i Judging whether the point is an interference score cur _ score to be fused or not;
1. if the point X is i If the interference score cur _ score is not to be fused, the first step is entered to judge the next point X i+1 ；
2. Otherwise, entering the second step;
the second step is that: go through X i In the interference radiation window [ X ] i-wnd ，X i+wnd ]Each point X in j Update Y j ；
1. If X is j ＜cur_score，
If Y is j If < cur _ score, Y is updated j (Y j ＝β 1 *cur_score+β 2 *Y j ) Entering the second step to judge the next point X j+1 ；
Otherwise, entering the second step to judge the next point X j+1 ；
2. If not, then,
if Y is j ＜X j Then, Y is updated j From X j And (7) assigning values. Entering the second step, judging the next point X j+1 ；
Otherwise, entering the second step to judge the next point X j+1 ；
Wherein, beta 1 And beta 2 Is a weighting coefficient, satisfies beta 1 +β 2 ＝1，β 1 ＞0，β 2 And if the value is more than 0, when the interference score fusion function Q is called for the first time, the fused interference score array Y is given an initial value by the interference score array X to be fused.
The secondary interference detection algorithm comprises the following steps:
the first step is as follows: traverse each point X in array X i First evaluation X i Whether it is a candidate peak point;
1. if X is i ≤X i-1 Or X i ＜X i+1 Then, go to the first step to judge the next point X i+1 ；
2. Otherwise, initializing a counter: 0 is used for counting less than or equal to X i The number of the second step is entered;
the second step is that: go through X i Left and right side window [ X ] i-wnd1 ，X i+wnd2 ]Each point X in j Judgment of X i And X j Second evaluation of X i Whether it is a candidate peak point;
1. if Score j Not more than thre, and X j ≤X i If so, the counter cou is automatically increased, and the second step is entered to judge the next point X j+1 ；
2. Otherwise, entering the first step and judging the next point X i+1 ；
The third step: traverse X i Right side additional window [ X ] i+wnd2+1 ，X i+wnd2+1+wnd3 ]Each point X in k Judgment of X i And X k A third evaluation of X i Whether it is a candidate peak point;
1. if Score k Not more than thre and X k ＜X i If yes, the position i is saved to the peak position array Y, and the first step is entered to judge the next point X i+1 ；
2. If not, then,
if X is k ＞X i Then go to the first step to determine the next point X i+1 ；
Otherwise, entering the third step and judging the next point X k+1 ；
Therein, the window wnd 1 、wnd 2 、wnd 3 Both are related to pulse rate, which is typically 1/4;
the fourth step: positioning valley points, wherein the valley positioning algorithm is based on peak algorithm positioning, and the position of the minimum value between peak points is the valley point;
the fifth step: calculating the standard deviation of a curve formed by the peak point and the valley point;
and a sixth step: calculating a peak interference Score array Score according to the standard deviation and different threshold lines 5 And valley interference Score array Score 6 ；
The seventh step: construction of two-level interference algorithm comprehensive Score array 1-Score 7 And integrating the Score arrays according to a first-level interference algorithm 4 Score array of sum-peak interference scores 5 Constructing a secondary interference algorithm comprehensive Score array 1-Score by using an interference Score function Q 7 ；
The eighth step: 2-Score for constructing comprehensive Score array of secondary interference algorithm 8 And integrating the Score array 1-Score according to a secondary interference algorithm 7 And valley interference Score array Score 6 Constructing a secondary interference algorithm comprehensive Score array 2-Score by using an interference Score function Q 8 ；
The ninth step: construction of post-secondary interference detection signal y 8 And synthesizing the Score array 2-Score according to a secondary interference algorithm 8 And the smoothed signal y 2 Constructing a post-secondary interference detection signal using a signal synthesis function y8 ；
The signal synthesis method comprises the following steps:
the first step is as follows: traversing each Score in the Score array Score i To discriminate Score i Whether a valid signal score is satisfied;
1. if Score i If more than thre, then go to the first step to determine the next Score i+1 ；
2. Otherwise, entering the second step;
the second step is that: synthesizing signals;
1. if Score i-1 If more than thre, entering the third step;
2. if Score i+1 If yes, entering the fifth step;
3. otherwise, the segment signal [ X ] is retained (i-1)*wnd1+1 ：X i*wnd1 ]Into signal Y;
the third step: traverse the Score array [ Score ] i+1 ：Score i+wnd4 ]Each Score within Score j Judging Score j Whether a valid signal score is satisfied;
1. if Score j If more than thre, the first step is entered to judge the next Score i+1 ；
2. Otherwise, entering the third step and judging the next Score j+1 ；
The fourth step: synthesizing signal segments before an interference point;
1. saving signal segments [ X ] (i-1)*wnd1-wnd2 ：X (i-1)*wnd1 ]The signal from the signal at the position of the maximum value to the signal section end is added into the signal Y;
2. hold the segment signal [ X (i-1)*wnd1+1 ：X i*wnd1 ]Into signal Y;
3. entering the first step, judging the next Score i+1 ；
The fifth step: traverse the Score array [ Score i-wnd4 ：Score i-1 ]Each Score within Score k Judging Score k Whether a valid signal score is satisfied;
1. if Score k If the Score is more than the Score, the first step is carried out, and the next Score is judged i+1 ；
2. Otherwise, the fifth step is entered, and the next Score is judged k+1 ；
And a sixth step: signal segment synthesis after the interference point;
1. hold the segment signal [ X (i-1)*wnd1+1 ：X i*wnd1 ]Into signal Y;
2. saving signal segments [ X ] i*wnd1+1 ：X i*wnd1+wnd3 ]Starting from the signal section to the position where the maximum value is located, and sending the signal to a signal Y;
3. entering the first step, judging the next Score i+1 ；
Therein, the window wnd 1 Is the interference detection granularity, is related to the sampling rate, and is generally 0.5s x Fs or 1s x Fs; window wnd 2 、wnd 3 The granularity of the interference signal is reserved, and is related to the sampling rate, and is generally 0.5s x Fs; window wnd 4 The interference detection granularity is related to 5 granularity, and after signal synthesis, the signal is directly used as an effective signal of the period of time.
DETAILED DESCRIPTION OF EMBODIMENT (S) OF INVENTION
1. Data acquisition and preprocessing
(1) Data slicing
α is 30 and β is 60. Namely, the shortest signal processing length of the current interference algorithm is 30 seconds, the longest signal processing length is 60 seconds, and a data slicing mode meeting the requirements of the current interference algorithm can be designed according to requirements. Here, 3 minute PPG data acquired using a blood oxygen ring is shown, with a sampling rate (Fs) of 100 Hz. The first piece of data entered the disturb algorithm was 30 seconds in length, followed by 31 seconds, 32 seconds, … …, 60 seconds, and finally sliced at 60 seconds in length. Finally, each slice data retains the latest 60 seconds data for iterative updating until the interference detection of the PPG data with the length of 3 minutes is completed. The signal slicing results are shown in FIGS. 3 to 6.
(2) Baseline wander removal:
the mean value removal baseline drift calculation is fast and convenient to update, and the baseline drift window wnd in the algorithm 1 The signal was subjected to baseline wander removal processing according to equations (1) to (3) at 0.8 × Fs-1, and the signal after baseline wander removal is shown in fig. 7 to (8).
(3) Smoothing:
the smooth processing is carried out on the signal, so that the influence of the mutation point on the signal can be reduced, and more favorable information can be provided for the subsequent signal utilization. Here wnd 2 As shown in fig. 9 to 10, the signal is smoothed according to equations (4) to (5). (4) Normalization:
the difference of the PPG signals obtained by different acquisition devices, different light intensities and different environments is large or small and can not be estimated. In order to reduce the influence of the factors on the interference algorithm, before the interference detection is carried out, the signal is normalized, and the influence of the dimension on the algorithm is reduced. Where N is 500 and b is 1000, the signal is normalized according to equation (6), and the normalized signal is shown in fig. 11 to 12.
2. Estimating the RR interval of PPG signal
The patent relates to more window parameter selection, and real-time interference detection requirements of PPG signals make it necessary to adjust a window in real time according to the pulse rate. For this purpose, the RR interval of the PPG signal is estimated according to equations (7) - (13), taking the 30 sPGS signal after the 1 st entering interference algorithm normalization processing as an example, wherein alpha is 1 ＝1.5，α 2 ＝-1.5，α 3 The estimated process is shown in fig. 13 to 15, where Fs is 0.15. The red point value obtained from fig. 15 is the estimated RR interval, which is slightly smaller than the actual RR interval.
3. Interference detection
To better describe the effectiveness of the interference detection algorithm of this patent, a signal with a length of 60 seconds is selected for interference detection. The overall quality of the signal is good, but severe jitter interference and signal mutation exist, and respiratory interference and myoelectric interference exist to a certain degree. The signals are shown in figure (16);
(1) primary interference detection
In order to detect long-term or abrupt interference as much as possible in the interference primary screening process, the first-stage interference detection part comprises three layers of interference detection algorithms, and the three layers of interference detection algorithms respectively identify the long-term interference, the approximate horizontal interference and the slope abrupt interference. Here, the interference detection granularity is set to 0.5 seconds, i.e. each interference score in the array of interference scores describes a signal quality of 0.5 seconds.
Identifying long-time amplitude interference signals
Setting wnd 3 ＝2*Fs，α 1 、α 2 And Score 1 The corresponding relation is shown in table 1, and the long-time amplitude interference Score array Score is completed according to the formulas (15) to (24) 1 The calculation result of (2) is shown in fig. 17.
TABLE 1 alpha 1 、α 2 And Score 1 Corresponding relation table
| |
100 | 90 | 80 | 70 | 60 | 50 |
| |
8 | 6 | 4 | 3 | 2 | - |
| α 2 | - | - | - | - | 0.15 | 0.25 |
Identifying short-term horizontal interfering signals
Setting wnd 4 ＝0.2*Fs，α 3 Score of short-term horizontal interference Score array is performed according to equations (25) - (28) at 0.01 2 The calculation result of (2) is shown in fig. 18.
Identifying abrupt slope interference signals
Setting alpha 4 、α 5 And Score 3 The corresponding relationship is shown in Table 2, and the slope mutation interference Score array Score is completed according to the formulas (29) to (36) 3 The calculation result of (2) is shown in fig. 19.
TABLE 2. alpha 4 、α 5 And Score 1 Corresponding relation table
| |
100 | 90 | 80 | 70 | 60 | 50 |
| α 4 | - | - | 10 | - | 5 | - |
| α 5 | - | - | 10 | - | 6 | - |
(2) Composite interference score
The interference radiation window wnd is set to 1, that is, it is known from the interference detection granularity set in the "first-stage interference detection" to be 0.5 seconds, and the current interference point influence range is 0.5 seconds before and after the 0.5 second signal. According to the interference Score fusion function Q, completing the Score of the interference Score array Score 1 、Score 2 And Score 3 The fusion of (2) and the fusion result of the integrated interference scores are shown in FIG. 20.
Setting a first order interference detection threshold fractional line thre 1 50 when Score i If the signal is more than 50, the corresponding 0.5 second signal is an interference signal. The result after the first-order interference detection is shown in fig. 21.
(3) Secondary interference detection
Locating the peak point, calculating the interference score of the peak point
Set the left side window wnd 1 0.3 RR, right window wnd 2 0.3 RR, right sideAdditional window wnd 3 0.25 RR, where RR is "2 estimated RR intervals in the estimated PPG signal RR intervals" module. Alpha is alpha 6 、α 7 And Score 5 、Score 6 The correspondence of (a) is shown in Table 3. Score is calculated using the composite interference Score function Q 4 And Score 5 Fusion interference Score array Score of 7 The calculation results are shown in fig. 22.
TABLE 3. alpha 6 、α 7 And Score 5 、Score 6 Corresponding relation table
| Score 5 、 |
40 | 30 | 20 | 10 |
| α 6 、α 7 | 1.5 | 1 | 0.5 | - |
Locating valley point, calculating valley point interference score
Score is calculated using the composite interference Score function Q 7 And Score 6 Fusion interference Score array Score of 8 The calculation results are shown in fig. 23.
Secondary composite interference score
And (4) finishing the evaluation of the interference score of the whole signal according to the primary interference detection algorithm and the secondary interference detection algorithm, wherein the evaluation result is shown as a graph (24).
4. Signal synthesis
Set window wnd 1 0.5 Fs, window wnd2 0.5 Fs, window wnd 3 0.5 Fs, window wnd 4 5. Setting a first order interference detection threshold fractional line thre 2 50, two-level interference detection threshold fraction line thre 2 The ratio of (40,. After two-stage interference detection, the signal detection results are shown in fig. 25 to 26.
The core technical points of the invention are as follows:
1. data acquisition and preprocessing:
(1) slicing data:
at the beginning of PPG signal processing, PPG data is sliced at a-second length, incrementing to β -second length slices over time. Specifically, when the length of the PPG signal to be interference-detected is less than α seconds, no interference detection is performed on the data; over time, each slice data length is then α seconds, α +1 seconds, … …, β seconds, eventually slicing at β seconds length; and finally, each slice data retains the latest beta second length data for iterative updating, and the interference detection of each second is completed.
(2) Baseline wander removal:
and (3) removing the baseline drift for each slice of data by adopting a mean value baseline drift removing method, wherein the specific formula is shown in formulas (1) to (3).
(3) Smoothing:
the PPG signal is often affected by high frequency interference, random noise or low perfusion, and the baseline shift algorithm cannot suppress these interferences, so it is important to smooth each signal point. The smoothing algorithm is shown in equations (4) - (5).
(4) Normalization:
in order to reduce the influence of PPG signals of different acquisition devices and different dimensions on the algorithm, maximum and minimum normalization processing needs to be carried out on each signal after baseline drift is removed, so that the original form of data can be reserved, and the influence of the dimensions can be eliminated. The normalization algorithm is shown in equation (6).
2. Estimating RR interval of PPG signal
Interference detection window is fast and slow and real-time change according to the pulse rate that the signal corresponds in this patent, when carrying out interference detection for the first time, need predict the RR interphase of PPG signal, and the later stage obtains accurate RR interphase through the R ripples position to PPG signal location every second. The specific estimation formula is shown in equations (7) to (13).
The histogram median algorithm is a space-time conversion algorithm, and has a high operation performance speed on an integer array, but has no capability on a floating-point array. Therefore, the present invention provides a new median-taking algorithm, as shown in formula (14) and function P, which can be applied to unsigned, integer, and floating-point data, and can improve speed by hundreds of times compared with sorting median-taking, and hardly occupy memory.
3. Interference detection
PPG signal has many noises in reality, including the interference of various factors such as ambient light, motion artifact, myoelectricity interference, random noise, low perfusion, divides according to the interference degree, can divide into two kinds: one is large interference caused by long-term motion artifact, baseline drift, electromyographic interference, and the like; the other is small interference caused by high frequency interference, random noise, etc. In the patent, the primary interference detection is mainly used for identifying long-term interference or abrupt interference, and the secondary interference detection is mainly used for identifying interference which is difficult to use and causes signal deformation due to random noise, ambient light, low perfusion and the like.
(1) Primary interference detection
In order to detect long-term or abrupt interference as much as possible in the interference primary screening process, the first-stage interference detection part comprises three layers of interference detection algorithms, and the three layers of interference detection algorithms respectively identify the long-term interference, the approximate horizontal interference and the slope abrupt interference.
Identifying long-time amplitude interference signals
The specific calculation formula is shown in formulas (15) - (24).
The equations (15) to (24) are generally divided into 5 steps:
step 1: equations (15) - (17) convert the original signal to increase the variation amplitude of the interference signal, so as to prepare for subsequent interference detection;
step 2: equations (18) - (19), calculating the mean value of the converted signal according to the window;
and 3, step 3: equations (20) - (21), calculating standard deviation for the converted signal according to the window;
and 4, step 4: equation (22), construct a new signal for interference detection;
and 5, step 5: and (5) in formulas (23) to (24), searching a threshold line and acquiring a long-time variable amplitude interference position. According to different alpha 1 And alpha 2 Constructing an interference Score array Score for the signal 1 ；
Identifying short-term horizontal interfering signals
The specific calculation formula is shown in formulas (25) to (28).
Equations (25) - (28) are generally divided into 3 steps:
step 1: a formula (25) for constructing the square of the slope of the adjacent points, so as to increase the difference of interference signals and reduce the disturbance of normal signals;
step 2: equation (26), construct a new signal for horizontal interference detection;
and 3, step 3: and (5) formulas (27) - (28) for finding a threshold line and acquiring a horizontal interference position. According to different alpha 3 Constructing an interference Score array Score for the signal 2 ；
Identifying abrupt slope interference signals
The specific calculation formula is shown in formulas (29) to (36).
The equations (29) to (36) are generally divided into 2 steps:
step 1: formula (29), construct a new signal for slope mutation interference detection;
step 2: and (5) searching a threshold line and obtaining the slope mutation interference position according to the formulas (30) to (36). According to different alpha 4 And alpha 5 Constructing an interference Score array Score of the signal 3 ；
First-order interference detection algorithm detection result
Score array Score according to the three above abnormal interference scores 1 、Score 2 And Score 3 Calculating the final interference Score array Score of the first-level interference detection algorithm 4 . And according to the normalized signal y before interference detection 3 Constructing a signal y after primary interference detection 6 And the method is used for a secondary interference detection algorithm.
The specific construction method comprises the following steps: and determining whether the interference score requirement is met in the window according to the detection granularity of the interference score array. If yes, reserving the section of signal; otherwise, setting the segment signal as
Wherein, N and b are parameters in the normalization process, and are detailed in formula (6). A flow chart of the primary interference detection algorithm is shown in fig. 1.
(2) Composite interference score
The interference detection algorithm in the patent is divided into two stages, interference detection algorithms with different functions can be developed in each stage, and how interference scores after each interference detection algorithm are fused has important influence on the performance of the whole algorithm. Considering that different interference scores have different ranges of influence on signals, blind simple addition and subtraction fusion can influence the effect of the algorithm to a certain extent.
For this purpose, a new interference score fusion algorithm is proposed, which is specifically shown as "interference score fusion function Q" in "summary of the invention".
(3) Secondary interference detection
After primary interference detection, large variable amplitude interference, horizontal interference and slope abrupt change interference are basically detected, and detected signals can be basically used for calculating average pulse rate, but the calculation of functions of blood oxygen saturation, blood pressure and the like is still difficult. Therefore, the method develops secondary interference detection and is mainly used for identifying small amplitude interference signals. A flow chart for secondary interference detection is shown in fig. 2.
Specifically, the secondary interference detection algorithm is divided into 8 steps:
step 1: normalizing the post-primary interference detection signal. Normalization methods such as the normalization method in "data acquisition and preprocessing", see equation (6) for details;
step 2: and (5) positioning a peak point. The peak positioning algorithm in the patent adopts a pre-estimation window and a local maximum value method. Based on a conventional peak point positioning algorithm, the method is developed based on a primary interference detection algorithm and a primary interference detection comprehensive score array, and the specific algorithm is shown as a peak point positioning function in the invention content.
And 3, step 3: and positioning a valley point. The wave trough positioning algorithm is based on wave crest algorithm positioning, and the position of the minimum value between peak points is a trough point;
and 4, step 4: calculating the standard deviation of a curve formed by the peak point and the valley point;
and 5, step 5: calculating a peak interference Score array Score according to the standard deviation and different threshold lines 5 And valley interference Score array Score 6 ；
And 6, a step of: construction of two-level interference algorithm comprehensive Score array 1-Score 7 . Synthesizing Score array Score according to first-level interference algorithm 4 Score array of sum-peak interference scores 5 Constructing a secondary interference algorithm comprehensive Score array 1-Score by using an interference Score function Q 7 ；
And 7, step 7: 2-Score for constructing comprehensive Score array of secondary interference algorithm 8 . Comprehensive scoring array 1-Score according to secondary interference algorithm 7 And valley interference Score array Score 6 Constructing a secondary interference algorithm comprehensive Score array 2-Score by using an interference Score function Q 8 ；
And 8, step 8: construction of post-secondary interference detection signal y 8 . Comprehensive Score array 2-Score according to secondary interference algorithm 8 And the smoothed signal y 2 Constructing a signal y after secondary interference detection by using a signal synthesis function 8 ；
4. Signal synthesis
The common synthesis method after signal interference detection is direct and simple splicing, and the method has the advantages of simple algorithm, small calculation amount and capability of keeping the original signal as much as possible. But the disadvantage is also very obvious, which can cause a new round of artificial interference and is not beneficial to the subsequent signal application. The patent proposes a new signal synthesis method aiming at the defect, and the main expression is as shown in the signal synthesis function in the 'summary of the invention'.
Claims (10)
1. A method for detecting interference of pulse wave signals is accurate to the existence of interference of positioning signals of each signal point, the signal interference level and the splicing of effective signals, and is characterized in that: the method comprises the following steps:
1) data acquisition and preprocessing:
(1) slicing data:
at the beginning of processing the PPG signal, slicing the PPG data according to the length of alpha second, and increasing the length of the PPG data to the length of beta second slice along with the time lapse, wherein specifically, when the length of the PPG signal to be subjected to interference detection is less than alpha second, the interference detection is not carried out on the data; over time, each slice data length is then α seconds, α +1 seconds, … …, β seconds, eventually slicing at β seconds length; finally, each slice data retains the latest beta second length data for iterative updating, and the interference detection of each second is completed;
(2) baseline wander removal:
removing the baseline drift of each slice data by adopting a mean value baseline drift removing method;
(3) smoothing:
the PPG signal is often affected by high frequency interference, random noise or low perfusion, and the baseline shift algorithm cannot suppress these interferences, so it is important to smooth each signal point;
(4) normalization:
in order to reduce the influence of PPG signals of different acquisition equipment and different dimensions on the algorithm, maximum and minimum normalization processing needs to be carried out on each signal after baseline drift is removed, so that the original form of data can be reserved, and the influence of the dimensions can be eliminated;
2) estimating the RR interval of the PPG signal:
the interference detection window changes in real time according to the pulse rate corresponding to the signal, the RR interphase of the PPG signal needs to be estimated when the interference detection is carried out for the first time, and the accurate RR interphase is obtained by the R wave position for positioning the PPG signal per second in the later period;
3) interference detection
In reality, PPG signals have a lot of noises, including the interference of various factors such as ambient light, motion artifact, myoelectricity interference, random noise and low perfusion, and the noises are divided according to the interference degree and can be divided into two types: one is large interference caused by long-term motion artifact, baseline drift, electromyographic interference and the like; the other is small interference caused by high-frequency interference, random noise and the like, wherein primary interference detection is mainly used for identifying long-term interference or sudden interference, and secondary interference detection is mainly used for identifying interference which causes signal deformation and is difficult to use due to random noise, ambient light, low perfusion and the like;
4) signal synthesis
The common synthesis method after signal interference detection is direct and simple splicing, and the method has the advantages of simple algorithm, small calculation amount, and original signals reserved as much as possible, but has obvious defects, can cause a new round of artificial interference, and is not beneficial to subsequent signal application.
2. The method of detecting interference with a pulse wave signal according to claim 1, characterized in that: the specific formula of baseline drift removal in the step 1) is shown in formulas (1) to (3):
y 1 (i)＝y 0 (i)-m(i)，1≤i≤l (1)
wherein wnd 1 Is a baseline drift window, typically an odd number, y 0 Is the original signal, y 1 Is the signal after removal of the baseline drift, l is the signal length.
3. The method of detecting interference of a pulse wave signal according to claim 1, characterized in that: the smoothing algorithm in the step 1) is shown in formulas (4) to (5):
wherein wnd 2 Is a smooth window, typically an odd number, y 1 Is to remove the signal after baseline drift, y 2 Is the smoothed signal and l is the signal length.
4. The method of detecting interference of a pulse wave signal according to claim 1, characterized in that: the normalization algorithm in the step 1) is shown as a formula (6):
where N is the range of the normalized data, b is the baseline of the normalized data, y 2 Is the smoothed signal, y 3 Is the normalized signal and l is the signal length.
5. The method of detecting interference of a pulse wave signal according to claim 1, characterized in that: the estimation formula (7) - (13) for the RR interval of the PPG signal in the step 1) is shown as follows:
z 0 (j)＝i，y 4 (i)＞t1 or y 4 (i)＜t2 (8)
z 1 (j)＝z 0 (i)-z 0 (i-1) (12)
z 2 (k)＝z 1 (j)，z 1 (j)＞α 3 (13)
wherein, y 3 Is the normalized signal, l is the signal length, α 1 Is the coefficient of the upper threshold line, requires α 1 ＞0，α 2 Is the coefficient of the lower threshold line, requires α 2 ＜0，α 3 Is a parameter for estimating RR interval, requires alpha 3 > 0, final pair z 2 (k) Calculating a median value, wherein the evaluated value is an estimated RR interval, and the numerical value is related to a sampling rate;
z is 2 (k) The median algorithm is as follows:
the first step is as follows: obtaining the minimum value of the array X as the lower limit X lower Maximum value as the upper limit X upper ；
The second step: traverse each point X in array X i Judging whether the point is [ X ] or not lower ，X upper ]Within;
1. if the point is not in the range, entering the second step to judge the next point X i+1 ；
2. Otherwise, entering the third step;
the third step: traverse each point X in array X j Judgment of X i And X j In relation to (2)；
1. Initializing two counters, cou 1 ＝0；cou 2 ＝0；(cou 1 Statistic less than X i Number of (c), cou 2 Statistic equals X i Number of (2)
2. Judgment of X i And X j The relationship of (1);
if i is equal to j (i.e. not compared with itself), go to the third step to judge the next point X j+1 ；
If X is j ＜X i Then counter cou 1 Self-increasing, entering the fourth step;
if X is j ＝X i Then counter cou 2 Self-increasing, entering the third step, judging the next point X j+1 ；
The fourth step: judge cou 1 Relation to p, update X upper ；
1. If couu 1 ＞(p-1)，
If X is upper ＞X i Then update X upper From X i Assigning, entering the second step, and judging the next point X i+1 ；
Otherwise, entering the second step to judge the next point X i+1 ；
2. Otherwise, entering the third step and judging the next point X j+1 ；
The fifth step: judge cou 1 、cou 1 +cou 2 Relation to p, update X lower Returning a function result;
1. if cou 1 If (p-1), then the function result X is returned i ；
2. If cou 1 < (p-1), but ku 1 +cou 2 ≧ p-1, the function result X is returned i ；
3. If cou 1 +cou 2 ＜(p-1)，
If X is lower ＜X i Then update X lower From X i And (4) assigning values. Entering the second step, judging the next point X i+1 ；
Otherwise, entering the second step to judge the next point X i+1 。
6. The method of detecting interference of a pulse wave signal according to claim 1, characterized in that: the interference detection in the step 3) mainly comprises the following steps:
1) primary interference detection
In order to detect long-term or abrupt interference as much as possible in the interference primary screening process, the primary interference detection part comprises three layers of interference detection algorithms, and the three layers of interference detection algorithms respectively identify a long-term interference signal, an approximate horizontal interference signal and an abrupt slope interference signal;
2) composite interference score
The interference detection algorithm is divided into two stages, interference detection algorithms with different functions can be developed in each stage, and how interference scores after each interference detection algorithm are fused has important influence on the performance of the whole algorithm. Considering that different interference scores have different ranges of influence on signals, blind simple addition and subtraction fusion can influence the effect of the algorithm to a certain extent.
3) Secondary interference detection
After primary interference detection, large variable amplitude interference, horizontal interference and slope abrupt change interference are basically detected, and detected signals can be basically used for calculating average pulse rate, but the calculation of functions of blood oxygen saturation, blood pressure and the like is still difficult. For this reason, secondary interference detection has been developed, mainly for identifying small amplitude interference signals.
4) Signal synthesis
The common synthesis method after signal interference detection is direct and simple splicing, and the method has the advantages of simple algorithm, small calculation amount and capability of keeping the original signal as much as possible. But the disadvantage is also very obvious, which can cause a new round of artificial interference and is not beneficial to the subsequent signal application.
7. The method of detecting interference with a pulse wave signal according to claim 1, characterized in that: the step of respectively identifying the long-time interference signal, the approximate horizontal interference signal and the slope abrupt change interference signal is respectively calculated as follows:
identifying long-time amplitude interference signals
The specific calculation formula is shown in formulas (15) - (24).
z 2 (j)＝i，z 1 (i)＞t1 or z 1 (i)＜t2 (23)
Wherein, y 3 Is the normalized signal, l is the signal length, wnd 3 Is an interference detection window, M (x) is an optimized median algorithm, which is detailed in step 2) to predict the RR interval of the PPG signal, alpha 1 Is the coefficient of the upper threshold line, requires α 1 ＞0，α 2 Is the coefficient of the lower threshold line, requires α 2 ＞0。z 1 Is data converted by interference detection, z 2 Is a long-time amplitude interference signal detection point.
The equations (15) to (24) are generally divided into 5 steps:
step 1: equations (15) - (17) convert the original signal to increase the variation amplitude of the interference signal, so as to prepare for subsequent interference detection;
step 2: equations (18) - (19), calculating the mean value of the converted signal according to the window;
and 3, step 3: equations (20) - (21), calculating standard deviation for the converted signal according to the window;
and 4, step 4: equation (22), construct a new signal for interference detection;
and 5, step 5: and (5) according to formulas (23) to (24), searching a threshold line and acquiring a long-time variable amplitude interference position. According to different alpha 1 And alpha 2 Constructing an interference Score array Score of the signal 1 ；
Identifying short-term horizontal interfering signals
The specific calculation formula is shown in formulas (25) to (28):
z 5 (j)＝i，z 4 (i)＜t3 (27)
t3＝α 3 *M(z 5 )，α 3 ＞0 (28)
wherein, y 3 Is the normalized signal, l is the signal length, wnd 4 Is an interference detection window, M (x) is an optimized median algorithm, which is detailed in step 2) to predict the RR interval of the PPG signal, alpha 3 Is the coefficient of the upper threshold line, requires α 3 ＞0。z 4 Is data converted by interference detection, z 5 Is a short-time horizontal interference signal detection point;
the equations (25) to (28) are generally divided into 3 steps:
step 1: a formula (25) is used for constructing the square of the slope of the adjacent points, and is used for increasing the difference of interference signals and reducing the disturbance of normal signals;
step 2: equation (26), construct a new signal for horizontal interference detection;
and 3, step 3: and (5) formulas (27) - (28) for finding a threshold line and acquiring a horizontal interference position. According to different alpha 3 Constructing an interference Score array Score of the signal 2 ；
Identifying abrupt slope interference signals
The specific calculation formula is shown in formulas (29) to (36).
z 9 (j)＝i，z 6 (i)＞t4 or z 6 (i)＜t5 (31)
Wherein, y 3 Is the normalized signal, l is the signal length, α 4 Is the coefficient of the upper threshold line, requires α 4 ＞0，α 5 Is the coefficient of the lower threshold line, requires α 5 ＞0，z 6 Is data converted by interference detection, z 9 Is a slope abrupt change interference signal detection point;
equations (29) to (36) are generally divided into 2 steps:
step 1: formula (29), construct a new signal for slope mutation interference detection;
step 2: and (5) searching a threshold line and obtaining the slope mutation interference position according to the formulas (30) to (36). According to different alpha 4 And alpha 5 Constructing an interference Score array Score of the signal 3 ；
The three abnormal interference Score array Score 1 、Score 2 And Score 3 And calculating the final interference Score array Score of the first-level interference detection algorithm 4 And according to the normalized signal y before interference detection 3 Constructing a signal y after primary interference detection 6 For a secondary interference detection algorithm;
the specific construction method comprises the following steps: determining whether the window meets the requirement of interference score according to the detection granularity of the interference score array, and if so, retaining the section of signal; otherwise, setting the segment signal as
Wherein N and b are normalization processesThe parameters in (1) are shown in the formula (6).
8. The method of detecting interference with a pulse wave signal according to claim 7, wherein: the fusion algorithm in the step 2) comprises the following steps:
the first step is as follows: traversing each point X in the array X i Judging whether the point is an interference score cur _ score to be fused or not;
1. if the point X is i If the interference score is not cur _ score to be fused, the first step is entered to judge the next point X i+1 ；
2. Otherwise, entering the second step;
the second step is that: traverse X i In the interference radiation window [ X ] i-wnd ，X i+wnd ]Each point X in j Update Y j ；
1. If X j ＜cur_score，
If Y is j If < cur _ score, Y is updated j (Y j ＝β 1 *cur_score+β 2 *Y j ) Entering the second step to judge the next point X j+1 ；
Otherwise, entering the second step to judge the next point X j+1 ；
2. If not, then,
if Y is j ＜X j Then, Y is updated j From X j And (7) assigning values. Entering the second step, judging the next point X j+1 ；
Otherwise, entering the second step to judge the next point X j+1 ；
Wherein beta is 1 And beta 2 Is a weighting coefficient, satisfies beta 1 +β 2 ＝1，β 1 ＞0，β 2 And if the value is more than 0, when the interference score fusion function Q is called for the first time, the fused interference score array Y is given an initial value by the interference score array X to be fused.
9. The method of detecting interference with a pulse wave signal according to claim 7, wherein: the secondary interference detection algorithm comprises the following steps:
the first step is as follows: traverse each point X in array X i First evaluation X i Whether it is a candidate peak point;
1. if X is i ≤X i-1 Or X i ＜X i+1 Then, go to the first step to judge the next point X i+1 ；
2. Otherwise, initializing a counter: 0 is used for counting less than or equal to X i The number of the second step is entered;
the second step is that: go through X i Left and right side window [ X ] i-wnd1 ，X i+wnd2 ]Each point X in j Judgment of X i And X j Second evaluation of X i Whether it is a candidate peak point;
1. if Score j Not more than thre, and X j ≤X i If so, the counter cou is automatically increased, and the second step is entered to judge the next point X j+1 ；
2. Otherwise, entering the first step and judging the next point X i+1 ；
The third step: traverse X i Right side additional window [ X ] i+wnd2+1 ，X i+wnd2+1+wnd3 ]Each point X in k Judgment of X i And X k A third evaluation of X i Whether it is a candidate peak point;
1. if Score k Not more than thre, and X k ＜X i If yes, the position i is saved to the peak position array Y, and the first step is entered to judge the next point X i+1 ；
2. If not, then,
if X k ＞X i Then, go to the first step to judge the next point X i+1 ；
Otherwise, entering the third step and judging the next point X k+1 ；
Therein, the window wnd 1 、wnd 2 、wnd 3 Both are related to pulse rate, which is typically 1/4;
the fourth step: positioning valley points, wherein the valley positioning algorithm is based on peak algorithm positioning, and the position of the minimum value between peak points is the valley point;
the fifth step: calculating the standard deviation of a curve formed by the peak point and the valley point;
and a sixth step: calculating a peak interference Score array Score according to the standard deviation and different threshold lines 5 And valley interference Score array Score 6 ；
The seventh step: construction of two-level interference algorithm comprehensive Score array 1-Score 7 Synthesizing the Score array Score according to a first-level interference algorithm 4 Score array of sum-peak interference scores 5 And constructing a secondary interference algorithm comprehensive Score array 1-Score by using the interference Score function Q 7 ；
Eighth step: 2-Score for constructing comprehensive Score array of secondary interference algorithm 8 And synthesizing the Score array 1-Score according to a secondary interference algorithm 7 And valley interference Score array Score 6 And constructing a secondary interference algorithm comprehensive Score array 2-Score by using the interference Score function Q 8 ；
The ninth step: construction of post-secondary interference detection signal y 8 And synthesizing the Score array 2-Score according to a secondary interference algorithm 8 And the smoothed signal y 2 Constructing a signal y after secondary interference detection by using a signal synthesis function 8 。
10. The method of detecting interference with a pulse wave signal according to claim 7, wherein: the signal synthesis method comprises the following steps:
the first step is as follows: traversing each Score in the Score array Score i To discriminate Score i Whether a valid signal score is satisfied;
1. if Score i If more than thre, the first step is entered to judge the next Score i+1 ；
2. Otherwise, entering the second step;
the second step is that: synthesizing signals;
1. if Score i-1 If more than thre, entering the third step;
2. if Score i+1 If yes, entering the fifth step;
3. otherwise, the segment signal [ X ] is retained (i-1)*wnd1+1 ：X i*wnd1 ]Into signal Y;
the third step: traverse the Score array [ Score i+1 ：Score i+wnd4 ]Each Score within Score j Judging Score j Whether a valid signal score is satisfied;
1. if Score j If more than thre, the first step is entered to judge the next Score i+1 ；
2. Otherwise, entering the third step and judging the next Score j+1 ；
The fourth step: synthesizing signal segments before an interference point;
1. saving signal segments [ X ] (i-1)*wnd1-wnd2 ：X (i-1)*wnd1 ]The signal from the signal at the position of the maximum value to the signal section end is added into the signal Y;
2. hold the segment signal [ X (i-1)*wnd1+1 ：X i*wnd1 ]Into signal Y;
3. entering the first step, judging the next Score i+1 ；
The fifth step: traverse the Score array [ Score ] i-wnd4 ：Score i-1 ]Each Score within Score k Judging Score k Whether a valid signal score is satisfied;
1. if Score k If more than thre, the first step is entered to judge the next Score i+1 ；
2. Otherwise, the fifth step is entered, and the next Score is judged k+1 ；
And a sixth step: signal segment synthesis after the interference point;
1. hold the segment signal [ X (i-1)*wnd1+1 ：X i*wnd1 ]Into signal Y;
2. saving signal segments [ X ] i*wnd1+1 ：X i*wnd1+wnd3 ]Starting from the signal section to the position where the maximum value is located, and sending the signal to a signal Y;
3. entering the first step, judging the next Score i+1 ；
Therein, the window wnd 1 Is the interference detection granularity, is related to the sampling rate, and is generally 0.5s x Fs or 1s x Fs; window wnd 2 、wnd 3 The granularity of the interference signal is reserved, and is related to the sampling rate, and is generally 0.5s x Fs; window wnd 4 The interference detection granularity is related to 5 granularity, and after signal synthesis, the signal is directly used as an effective signal of the period of time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210340616.8A CN114912478A (en) | 2022-04-02 | 2022-04-02 | Interference detection method for pulse wave signals |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210340616.8A CN114912478A (en) | 2022-04-02 | 2022-04-02 | Interference detection method for pulse wave signals |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114912478A true CN114912478A (en) | 2022-08-16 |
Family
ID=82762487
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210340616.8A Pending CN114912478A (en) | 2022-04-02 | 2022-04-02 | Interference detection method for pulse wave signals |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114912478A (en) |
-
2022
- 2022-04-02 CN CN202210340616.8A patent/CN114912478A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210259613A1 (en) | Apparatuses for tracking physiological parameters of mother and fetus during pregnancy | |
| CN108814590B (en) | Detection method of electrocardio QRS wave group and electrocardio analysis method thereof | |
| CN117033911B (en) | Step counting analysis method based on intelligent glasses data | |
| Gavrovska et al. | Classification of prolapsed mitral valve versus healthy heart from phonocardiograms by multifractal analysis | |
| WO2021184805A1 (en) | Blood pressure prediction method and device using multiple data sources | |
| AU2019101756A4 (en) | Method for analyzing heart rate variability signal based on extremum energy decomposition method | |
| CA2357059A1 (en) | Signal processing apparatus | |
| CN108814591B (en) | Method for detecting width of electrocardio QRS wave group and electrocardio analysis method thereof | |
| CN114469124B (en) | Method for identifying abnormal electrocardiosignals in movement process | |
| CN112545472B (en) | PPG signal quality evaluation method, device, equipment and storage medium | |
| CN106798547A (en) | A kind of computational methods of the arm ankle pulse wave time delay based on statistics | |
| CN111419279A (en) | Self-adaptive noise elimination algorithm for ultrasonic Doppler fetal heart signals | |
| CN108992054A (en) | A kind of pulse signal peak point detection method and device | |
| US20200187804A1 (en) | Fetal heart rate extraction within a processor constrained environment | |
| CN111839494A (en) | Heart rate monitoring method and system | |
| US20230309899A1 (en) | Determination method and determination apparatus for beginning of t-wave, storage medium and computer program product | |
| CN114912478A (en) | Interference detection method for pulse wave signals | |
| CN111493821B (en) | PPG signal real-time denoising method based on MODWT and median filtering | |
| CN110226925B (en) | Blood pressure detection device based on pulse wave | |
| CN110151156B (en) | Fetal heart rate extraction method and system based on in-window automatic weighted average algorithm | |
| Vadrevu et al. | Effective systolic peak detection algorithm using variational mode decomposition and center of gravity | |
| CN111528900A (en) | Heart sound segmentation method and device based on Butterworth filter and Shannon entropy method | |
| CN110432885A (en) | A kind of photoplethysmographic noise remove method | |
| Jang et al. | A simple and robust method for determining the quality of cardiovascular signals using the signal similarity | |
| TWM530126U (en) | Apparatus for examining and processing heartbeat signal based on fitting curve |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |