CN117224092B - Photoelectric volume pulse wave interference band real-time detection method and system based on decision tree - Google Patents
Photoelectric volume pulse wave interference band real-time detection method and system based on decision tree Download PDFInfo
- Publication number
- CN117224092B CN117224092B CN202311526698.6A CN202311526698A CN117224092B CN 117224092 B CN117224092 B CN 117224092B CN 202311526698 A CN202311526698 A CN 202311526698A CN 117224092 B CN117224092 B CN 117224092B
- Authority
- CN
- China
- Prior art keywords
- peak
- data
- signal
- pulse
- real
- 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
- 238000000034 method Methods 0.000 title claims abstract description 171
- 238000011897 real-time detection Methods 0.000 title claims abstract description 56
- 238000003066 decision tree Methods 0.000 title claims abstract description 42
- 238000013145 classification model Methods 0.000 claims abstract description 13
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 230000008569 process Effects 0.000 claims description 69
- 238000004364 calculation method Methods 0.000 claims description 40
- 238000005070 sampling Methods 0.000 claims description 30
- 238000001514 detection method Methods 0.000 claims description 29
- 238000012549 training Methods 0.000 claims description 23
- 238000010606 normalization Methods 0.000 claims description 20
- 238000001914 filtration Methods 0.000 claims description 17
- 239000011159 matrix material Substances 0.000 claims description 17
- 230000005057 finger movement Effects 0.000 claims description 15
- 238000013186 photoplethysmography Methods 0.000 claims description 15
- 230000035945 sensitivity Effects 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 9
- 230000003139 buffering effect Effects 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 7
- 238000001276 Kolmogorov–Smirnov test Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 description 5
- 230000033001 locomotion Effects 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000002790 cross-validation Methods 0.000 description 3
- 206010003119 arrhythmia Diseases 0.000 description 2
- 230000006793 arrhythmia Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000003183 myoelectrical effect Effects 0.000 description 2
- 238000002203 pretreatment Methods 0.000 description 2
- 206010003210 Arteriosclerosis Diseases 0.000 description 1
- 208000024172 Cardiovascular disease Diseases 0.000 description 1
- 208000011775 arteriosclerosis disease Diseases 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 230000036772 blood pressure Effects 0.000 description 1
- 230000036996 cardiovascular health Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000003750 conditioning effect Effects 0.000 description 1
- 208000029078 coronary artery disease Diseases 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 210000000624 ear auricle Anatomy 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 210000003811 finger Anatomy 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003211 malignant effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000029058 respiratory gaseous exchange Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002792 vascular Effects 0.000 description 1
- 210000000707 wrist Anatomy 0.000 description 1
Abstract
The invention discloses a real-time detection method and a system for photoelectric volume pulse wave interference bands based on a decision tree, wherein the method comprises the following steps: acquiring pulse signals and preprocessing; extracting pulse signal peak values to obtain pulse main wave interval data; extracting specific variation parameters of pulse main wave interval data to form a feature vector set; based on the decision tree classification model, an interference section real-time detection model is constructed and trained, and the trained interference section real-time detection model is used for carrying out interference section real-time detection. The interference section of the photoelectric volume pulse wave can be stably, rapidly and accurately detected in real time.
Description
Technical Field
The invention belongs to the technical field of PPG signal detection, and relates to a method and a system for detecting a photoelectric volume pulse wave interference section in real time based on a decision tree.
Background
Photoplethysmography (Photoplethysmography) technology is a common method of measuring physiological parameters such as pulse, heart rate, etc. By placing the sensor on the skin surface of a body part such as a finger, earlobe, wrist, temple, etc., the varying vascular pressure is converted into an electrical signal, called PPG signal, i.e. photoplethysmography wave. Parameters such as heart rate, blood oxygen saturation, blood pressure and the like can be extracted from the PPG signal and used for human health state assessment, and the PPG signal can also be used for diagnosing cardiovascular diseases such as arrhythmia, arteriosclerosis, coronary heart disease and the like. Therefore, the PPG technology is widely applied to wearable devices such as intelligent watches, intelligent bracelets and the like, and monitors the heart rate, activity level and cardiovascular health condition of a human body in real time.
In the acquisition process, the PPG signal is easily influenced by factors such as human body movement, body surface static electricity, respiration and the like, so that the acquired signal contains various noises such as motion artifacts, baseline drift, power frequency interference, myoelectric interference and the like, and the quality of the signal is influenced. The baseline drift, the power frequency interference and the myoelectricity interference can be effectively filtered through a designed filter, and the frequency band of the motion artifact overlaps with the frequency band of the PPG signal, so that the motion artifact is difficult to inhibit through a filtering technology, and the waveform characteristics of the normal PPG signal are changed, which is called an interference section. Research shows that the false alarm rate of the clinical physiological signal monitor and the wearable equipment is high due to the interference section, so that the accurate detection of the interference section is a problem to be solved urgently.
PPG signal interference section detection is an effective means for improving PPG signal quality, and the existing PPG signal interference section detection technology has the defects of poor instantaneity, complex implementation algorithm and the like, requires a large amount of computing resources and time, and is difficult to apply to wearable equipment with limited signal processing capability.
The patent of application number 2022105241738 discloses a malignant arrhythmia recognition and prediction system based on a pulse main wave interval, wherein the detection of an interference section is disclosed to remove the influence of motion artifact and signal section loss caused by sliding or falling of a sensor, and a detection method of the interference section is specifically proposed for different types of interference sections according to the difference of signal characteristics of the interference section and a normal section. However, the interference section detection technology of the method has poor real-time performance and complex implementation algorithm.
Disclosure of Invention
The invention aims to provide a real-time detection method and a real-time detection system for a photoelectric volume pulse wave interference section based on a decision tree, which can stably, rapidly and accurately detect the interference section of the photoelectric volume pulse wave in real time.
The technical solution for realizing the purpose of the invention is as follows:
a real-time detection method of photoelectric volume pulse wave interference section based on decision tree comprises the following steps:
s01: acquiring pulse signals and preprocessing;
s02: extracting pulse signal peak values to obtain pulse main wave interval data;
s03: extracting specific variation parameters of pulse main wave interval data to form a feature vector set;
s04: based on the decision tree classification model, an interference section real-time detection model is constructed and trained, and the trained interference section real-time detection model is used for carrying out interference section real-time detection.
In a preferred embodiment, the pretreatment method in step S01 includes:
filtering the pulse signal by 0Hz and 50Hz integer coefficient notch filters and a low pass filter with a cut-off frequency of 62.5 Hz;
the filter calculation formula of the low-pass filter with the cutoff frequency of 62.5 and Hz comprises the following steps:
in the method, in the process of the invention,is the signal after being filtered by the low-pass filter, < >>For the filter order +.>Is the original pulse signal +>Sampling points->Is a 19 th order filter coefficient obtained by the fir1 function in Matlab, < >>Is the original pulse signal;
in the method, in the process of the invention,for the signal array after low-pass filter, the size of array buffer is 321, +.>The low-pass filtering result of the earliest sampling point is put in the highest position of the buffer zone in the window and marked as +.>Every time data is updated, the low-pass filtering result is shifted by one bit from high to low, and the earliest result is +.>Is covered by high order, new data is put in +.>；
The 0Hz and 50Hz integer coefficient notch filter calculation formula:
in the method, in the process of the invention,is the filtered signal of the integer notch filter, < >>Is the signal filtered by the 1 st order integer notch filter, +>Is the signal filtered by the 6 th order integer notch filter, +>Is the signal filtered by the 11 th order integer notch filter, +>Is the signal filtered by the 1 st low-pass filter,>is the signal filtered by the 156 th low-pass filter,>is the signal filtered by the 161 th low-pass filter,>is the 321 th low pass filter filtered signal.
In a preferred embodiment, the pretreatment method in step S01 further includes: the sliding standard normalization method for the pulse signals comprises the following steps:
calculating the mean value and variance of the data in the sliding window, then carrying out standard normalization on the data, and finally repeating the standard normalization on the data in the sliding window after shifting;
the pulse signal sliding window formula is:
in the method, in the process of the invention,for in-window caching data +.>For the signal buffer size,/>Is the%>Sampling points->For the earliest sampling point, the sampling point which is acquired latest is placed at the highest position of a buffer zone in a window and is marked as +.>Every time data is updated, the sampling point moves one bit from high to low, and the earliest acquired +.>Covered by high order bits, new data is placed in；
Pulse signal sliding standard normalization formula:
in the method, in the process of the invention,normalized results for data, ++>For buffering data mean value in the filter back window, +.>For filtering the variance of the buffered data in the back window, < >>For buffering the data sum in the filter back window, the normalized result is stored in an arrayIn (I)>For the 1 st normalization result, analogize, < + >>And normalizing the result for the last time when the signal is stopped.
In a preferred embodiment, the method for extracting the peak value of the pulse signal in the step S02 includes:
build size ofConstructing a matrix comprising only elements 0 and 1 +.>：
In the method, in the process of the invention,for sliding normalized data, +.>For 40 sampling points +.>, ，/>；
Reconstruction matrixCalculate->Sum of each row:
in the method, in the process of the invention,is provided with->One-dimensional matrix of 0 elements->Representing matrix->Is>The number of rows of the device is,；
peak detection:
in the method, in the process of the invention,is the%>And peak point index values.
In a preferred embodiment, the extracting the specific variation parameter of the pulse main wave interval data in the step S03 includes:
threshold crossing count, slope of adjacent peak points, ratio of pulse period, skewness coefficient of pulse signal, slope of data between adjacent peaks of pulse signal;
threshold crossing count calculation formula:
in the method, in the process of the invention,representation->Pass threshold->Is>Indicate->Peak and->Data set between peaks +.>，/>Index value of peak value, < >>Is a function->Input parameters of->，/>For stopping the total number of peak points extracted at the time of acquisition, < > for>Is->Index increment of (a);
slope calculation formula of adjacent peak points:
in the method, in the process of the invention,for peak amplitude +.>Is->Peak point and->The slope of the point of the peak,；
the ratio calculation formula of pulse period:
in the method, in the process of the invention,is->Peak point position and->Ratio of index values of the peak points;
the calculation formula of the deviation coefficient of the pulse signal comprises the following steps:
in the method, in the process of the invention,is->Peak and->Deviation coefficient of data between peak values +.>Is->Peak and->Standard deviation of data between peaks +.>Is->Peak and->Mean value of data between peaks +.>Is->Index increment initial value, ++>Is->Peak and->Data length of data between peaks.
The slope calculation formula of pulse wave between adjacent wave peaks:
in the method, in the process of the invention,is->Peak and->Slope of data between peaks, +.>Is->Peak and->Maximum value of data between peaks, +.>Is->Peak and->The minimum value of the data between the individual peaks,/>is->Peak and->Minimum position of data between peaks.
In a preferred embodiment, the step S03 further includes:
classifying the acquired pulse signals according to the interference section generation type, including: a normal segment signal, a sensor placement signal, a finger movement signal;
the normal segment signal, the sensor placement signal, and the finger movement signal are grouped in pairs, and feature selection is performed using a double sample KS test, excluding features that do not have significant differences.
In a preferred embodiment, in the step S04, a kappa coefficient is used in the training interference section real-time detection model to evaluate the classification accuracy, and a kappa coefficient calculation formula is as follows:
in the method, in the process of the invention,is kappa coefficient,/->、/>And->Respectively the column matrix +.>Line element and->Column element and element sum located on diagonal, +.>Is the number of features, +.>Representing a classification category;
calculating the average accuracy of each groupThe method comprises the following steps:
wherein,、/>、/>。
in the preferred technical scheme, in the step S04, the accuracy, the specificity and the sensitivity are adopted to evaluate the omission factor, the false detection rate and the accuracy of the model, and the calculation formula is as follows:
in the method, in the process of the invention,representing accuracy, ->Express specificity,/->Indicating sensitivity, & lt>Indicating that the positive class samples are predicted the correct amount, +.>Indicating that the negative class samples are predicted the correct amount, +.>Representing the number of prediction errors for the negative class sample; />Indicating the number of prediction errors to be made for the positive class samples.
The invention also discloses a real-time detection system of the photoelectric volume pulse wave interference section based on the decision tree, which comprises the following steps:
the pulse signal acquisition processing module acquires pulse signals and performs preprocessing;
the pulse signal peak value extraction module is used for extracting a pulse signal peak value to obtain pulse main wave interval data;
the specific variation parameter extraction module is used for extracting specific variation parameters of the pulse main wave interval data to form a feature vector set;
the interference section real-time detection model training detection module is used for constructing an interference section real-time detection model based on the decision tree classification model, training, and carrying out interference section real-time detection by using the trained interference section real-time detection model.
The invention also discloses a computer storage medium, on which a computer program is stored, which is executed to realize the method for detecting the photoelectric volume pulse wave interference section based on the decision tree in real time.
Compared with the prior art, the invention has the remarkable advantages that:
(1) Compared with the existing interference section detection method, the method selects the decision tree model and trains the model in a cross-validation mode. The classification model is trained for many times according to the pulse signals of the normal section and the pulse signals of different types of interference sections, so that the model with strong processing performance, high accuracy and low complexity is obtained, and the interference sections can be rapidly and accurately identified at the mobile terminal with limited computing resources.
(2) The specific variation parameters extracted by the method do not need to be calculated in a large amount, the calculation time is short, the algorithm real-time performance is high, the complexity is low, the interference sections generated under different conditions can be accurately identified, and the method can be embedded into different mobile terminals to realize the real-time detection of pulse signal interference sections.
Drawings
FIG. 1 is a flowchart of a real-time detection method of the photo-volume pulse wave interference band based on a decision tree in the embodiment;
FIG. 2 is a schematic block diagram of a real-time detection system for the photoplethysmography interference band based on a decision tree in the present embodiment;
FIG. 3 is a flow chart of the online detection of the photoplethysmography interference band;
fig. 4a-4d are waveform diagrams of interference segments in the acquired pulse wave signals.
Fig. 5 is a schematic structural diagram of a real-time detection system for the photoplethysmography interference section according to this embodiment.
Detailed Description
The principle of the invention is as follows: firstly, preprocessing signals, filtering and normalizing the signals, inhibiting baseline drift, power frequency interference and myoelectric interference, and eliminating amplitude difference of the signals. And then, carrying out real-time extraction on the peak value of the signal, extracting characteristic parameters of pulse waves between adjacent wave peaks, forming a characteristic vector set, and carrying out characteristic selection. And finally, on-line identification and removal of the interference section based on the trained decision tree, and storage of a clean pulse signal.
Example 1:
as shown in fig. 1, a method for detecting the interference section of the photoplethysmogram based on a decision tree in real time includes the following steps:
s01: acquiring pulse signals and preprocessing;
s02: extracting pulse signal peak values to obtain pulse main wave interval data;
s03: extracting specific variation parameters of pulse main wave interval data to form a feature vector set;
s04: based on the decision tree classification model, an interference section real-time detection model is constructed and trained, and the trained interference section real-time detection model is used for carrying out interference section real-time detection.
In a preferred embodiment, the preprocessing method in step S01 includes:
filtering the pulse signal by 0Hz and 50Hz integer coefficient notch filters and a low pass filter with a cut-off frequency of 62.5 Hz;
the low-pass filter with cut-off frequency 62.5 Hz filters the formula:
in the method, in the process of the invention,is the signal after being filtered by the low-pass filter, < >>For the filter order +.>Is the original pulse signal +>Sampling points->Is a 19 th order filter coefficient obtained by the fir1 function in Matlab, < >>Is the original pulse signal;
in the method, in the process of the invention,for the signal array after low-pass filter, the size of array buffer is 321, +.>The low-pass filtering result of the earliest sampling point is put in the highest position of the buffer zone in the window and marked as +.>Every time data is updated, the low-pass filtering result is shifted by one bit from high to low, and the earliest result is +.>Is covered by high order, new data is put in +.>；
0Hz and 50Hz integer coefficient notch filter calculation formula:
in the method, in the process of the invention,is the filtered signal of the integer notch filter, < >>Is the signal filtered by the 1 st order integer notch filter, +>Is the signal filtered by the 6 th order integer notch filter, +>Is the signal filtered by the 11 th order integer notch filter, +>Is the signal filtered by the 1 st low-pass filter,>is the signal filtered by the 156 th low-pass filter,>is the signal filtered by the 161 th low-pass filter,>is the 321 th low pass filter filtered signal.
In a preferred embodiment, the preprocessing method in step S01 further includes: the sliding standard normalization method for the pulse signals comprises the following steps:
calculating the mean value and variance of the data in the sliding window, then carrying out standard normalization on the data, and finally repeating the standard normalization on the data in the sliding window after shifting;
the pulse signal sliding window formula is:
in the method, in the process of the invention,for in-window caching data +.>For the signal buffer size,/>Is the%>Sampling points->For the earliest sampling point, the latest sampling point is placed in the buffer area in the windowThe most significant bit, denoted->Every time data is updated, the sampling point moves one bit from high to low, and the earliest acquired +.>Covered by high order bits, new data is placed in；
In the method, in the process of the invention,normalized results for data, ++>For buffering data mean value in the filter back window, +.>For filtering the variance of the buffered data in the back window, < >>For buffering the data sum in the filter back window, the normalized result is stored in an arrayIn (I)>For the 1 st normalization result, analogize, < + >>And normalizing the result for the last time when the signal is stopped.
In a preferred embodiment, the method for extracting the pulse signal peak in step S02 includes:
build size ofConstructing a matrix comprising only elements 0 and 1 +.>：
In the method, in the process of the invention,for sliding normalized data, +.>For 40 sampling points +.>, ，/>；
Reconstruction matrixCalculate->Sum of each row:
in the method, in the process of the invention,is provided with->One-dimensional matrix of 0 elements->Representing matrix->Is>The number of rows of the device is,；
peak detection:
in the method, in the process of the invention,is the%>And peak point index values.
In a preferred embodiment, the extracting the specific variation parameters of the pulse main interval data in step S03 includes:
threshold crossing count, slope of adjacent peak points, ratio of pulse period, skewness coefficient of pulse signal, slope of data between adjacent peaks of pulse signal;
threshold crossing count calculation formula:
in the method, in the process of the invention,representation->Pass threshold->Is>Indicate->Peak and->Data set between peaks +.>，/>Index value of peak value, < >>Is a function->Input parameters of->，/>Is->Index increment of (a);
slope calculation formula of adjacent peak points:
in the method, in the process of the invention,for peak amplitude +.>Is->Peak point and->The slope of the point of the peak,，/>the total number of peak points extracted when the acquisition is stopped;
the ratio calculation formula of pulse period:
in the method, in the process of the invention,is->Peak point position and->Ratio of index values of the peak points;
the calculation formula of the deviation coefficient of the pulse signal comprises the following steps:
in the method, in the process of the invention,is->Peak and->Deviation coefficient of data between peak values +.>Is->Peak and->Standard deviation of data between peaks +.>Is->Peak and->Mean value of data between peaks +.>Is->Index increment initial value, ++>Is->Peak and->Data length of data between peaks.
The slope calculation formula of pulse wave between adjacent wave peaks:
in the method, in the process of the invention,is->Peak and->Slope of data between peaks, +.>Is->Peak and->Maximum value of data between peaks, +.>Is->Peak and->The minimum value of the data between the individual peaks,/>is->Peak and->Minimum position of data between peaks.
In a preferred embodiment, step S03 further includes:
classifying the acquired pulse signals according to the interference section generation type, including: a normal segment signal, a sensor placement signal, a finger movement signal;
the normal segment signal, the sensor placement signal, and the finger movement signal are grouped in pairs, and feature selection is performed using a double sample KS test, excluding features that do not have significant differences.
In a preferred embodiment, in step S04, the classification accuracy is evaluated by using kappa coefficients in the training interference section real-time detection model, and the kappa coefficient calculation formula is as follows:
/>
in the method, in the process of the invention,is kappa coefficient,/->、/>And->Respectively the column matrix +.>Line element and->Column element and element sum located on diagonal, +.>Is the number of features, +.>Representing a classification category;
calculating the average accuracy of each groupThe method comprises the following steps:
wherein,、/>、/>。
in a preferred embodiment, in step S04, the accuracy, specificity and sensitivity are used to evaluate the omission factor, the false detection rate and the accuracy of the model, and the calculation formula is as follows:
in the method, in the process of the invention,representing accuracy, ->Express specificity,/->Indicating sensitivity, & lt>Indicating that the positive class samples are predicted the correct amount, +.>Indicating that the negative class samples are predicted the correct amount, +.>Representing the number of prediction errors for the negative class sample; />Indicating the number of prediction errors to be made for the positive class samples.
In another embodiment, a computer storage medium has a computer program stored thereon, and the computer program when executed implements the method for detecting the photoplethysmography interference section based on the decision tree.
In yet another embodiment, as shown in fig. 2, a real-time detection system for a volume pulse wave interference band based on a decision tree includes:
the pulse signal acquisition and processing module 10 acquires pulse signals and performs preprocessing;
the pulse signal peak value extraction module 20 extracts the pulse signal peak value to obtain the main pulse interval data;
the specific variation parameter extraction module 30 extracts specific variation parameters of the pulse main wave interval data to form a feature vector set;
the interference section real-time detection model training detection module 40 constructs and trains an interference section real-time detection model based on the decision tree classification model, and performs interference section real-time detection by using the trained interference section real-time detection model.
Specifically, the following description will be given by taking a preferred embodiment as an example, and the workflow of the real-time detection system for the photoplethysmography interference band based on the decision tree is as follows:
mainly comprises the following steps:
(1) Offline training of a real-time detection model of a photoelectric volume pulse wave interference section based on a decision tree;
(2) And (5) online detection of the photoelectric volume pulse wave interference section.
Wherein, (1) the offline training of the real-time detection model of the photoplethysmography interference band based on the decision tree, as shown in fig. 3, comprises:
step 1: collecting pulse signals of normal sections of different subjects in a sitting state and pulse signals subjected to interference in a finger movement state of the subjects, preprocessing the signals, reducing the influence of various noise signals such as power frequency interference, baseline deviation, myoelectricity interference and the like on the signal accuracy, and improving the detection accuracy of the interference sections;
the acquired signal was filtered using 0Hz and 50Hz integer coefficient notch filters and a low pass filter with a cut-off frequency of 62.5 Hz and the signal was normalized for sliding criteria.
The low-pass filter with cut-off frequency 62.5 Hz filters the formula:
in the method, in the process of the invention,is the signal after being filtered by the low-pass filter, < >>For the filter order +.>Is the original pulse signal +>Sampling points->Is a 19 th order filter coefficient obtained by the fir1 function in Matlab, < >>Is the original pulse signal;
in the method, in the process of the invention,for the signal array after low-pass filter, the size of array buffer is 321, +.>The low-pass filtering result of the earliest sampling point is put in the highest position of the buffer zone in the window and marked as +.>Every time data is updated, the low-pass filtering result is shifted by one bit from high to low, and the earliest result is +.>Is covered by high order, new data is put in +.>；
0Hz and 50Hz integer coefficient notch filter calculation formula:
in the method, in the process of the invention,is the filtered signal of the integer notch filter.
The sliding standard normalization process of the pulse signals is as follows: setting the size of the signal buffer area asAnd calculating the mean value and variance of the data in the window, carrying out standard normalization on the data, and repeating standard normalization after the data in the window is shifted.
Pulse signal sliding window formula:
in the method, in the process of the invention,data is cached for the window. />Is the%>Sampling points->For the earliest sampling point, the sampling point which is acquired latest is placed at the highest position of a buffer zone in a window and is marked as +.>. Every time data is updated, the sampling point is shifted by one bit from high to low, and the earliest sampling is carried out>Is covered by high order, new data is put in +.>。
Pulse signal sliding standard normalization formula:
/>
in the method, in the process of the invention,normalized results for data, ++>For buffering data mean value in the filter back window, +.>For filtering the variance of the buffered data in the back window, < >>For buffering the data sum in the filter back window, the normalized result is stored in an arrayIn (I)>For normalization result 1,/>And normalizing the result for the last time when the signal is stopped.
Step 2: extracting a peak value of a pulse signal, and acquiring data of a main pulse interval;
the pulse signal peak value extraction process comprises the following steps: build size ofConstructing a matrix comprising only elements 0 and 1 +.>Reconstruction matrix->Calculate->Peak detection.
Constructing a matrix containing only elements 0 and 1The calculation formula is as follows:
in the method, in the process of the invention,for sliding normalized data, +.>For a number of sampling points of 40,, />，/>。
reconstruction matrixCalculate->The sum formula of each row:
in the method, in the process of the invention,is provided with->One-dimensional matrix of 0 elements->Representing matrix->Is>The number of rows of the device is,。
peak detection calculation formula:
in the method, in the process of the invention,is the%>Index value of peak point->，/>，/>The number of peak points extracted in the window.
Step 3: extracting specific parameters of pulse main wave interval data as characteristics of interference section detection to form a characteristic vector set;
the specific variation parameters of the pulse main wave interval data comprise: threshold crossing count, slope of adjacent peak points, ratio of pulse periods, skewness coefficient of pulse signals, slope of data between adjacent peaks of pulse signals.
Threshold crossing count calculation formula:
in the method, in the process of the invention,representation->Pass threshold->Is>Index value of peak value, < >>Indicate->Peak and->Data set between peaks, +.>For stopping the total number of peak points extracted at the time of acquisition, < > for>Is->Index increment, function->The following are provided:
in the method, in the process of the invention,is a function->Is provided.
Slope calculation formula of adjacent peak points:
in the method, in the process of the invention,for peak amplitude +.>Is->Peak point and->Slope of the peak points.
The ratio calculation formula of pulse period:
in the method, in the process of the invention,is->Peak point position and->Ratio of peak index values.
The calculation formula of the deviation coefficient of the pulse signal comprises the following steps:
in the method, in the process of the invention,is->Peak and->Deviation coefficient of data between peak values +.>Is->Peak and->Standard deviation of data between peaks +.>Indicate->Peak and->Data set between peaks, +.>Is the firstPeak and->Mean value of data between peaks +.>Is->Index increment initial value, ++>Is->Peak and->Data length of data between peaks.
The slope calculation formula of pulse wave between adjacent wave peaks:
in the method, in the process of the invention,is->Peak and->Slope of data between peaks, +.>Is->Peak and->Maximum value of data between peaks, +.>Is->Peak and->The minimum value of the data between the individual peaks,/>is->Peak and->Inter-peak dataIs the minimum position of (c).
Step 3.1: classifying the acquired pulse signals according to the interference section generation type, including: normal segment signal, sensor placement signal, finger movement signal.
Step 3.2: extracting specific variation parameters from pulse signals of a normal section, sensor placement and finger movement as input of a designed intelligent detection model of an interference section, wherein the extracted specific variation parameters comprise: threshold crossing count, slope of adjacent peak points, ratio of pulse periods, skewness coefficient of pulse signals, slope of data between adjacent peaks of pulse signals.
Step 4: selecting the extracted features, and eliminating the features without obvious difference;
some of the features were not normally distributed and feature selection was performed using the Kolmogorov-Smirno (KS) test.
Step 5: constructing a real-time detection model of a photoelectric volume pulse wave interference section based on a decision tree, and training the model;
step 5.1: training a photoelectric volume pulse wave interference band detection model based on a decision tree; and taking characteristic parameters of the normal section signals, the sensor placement signals and the finger movement signals as input training decision tree classification models. The model is trained by using a cross-validation mode, and the classification model is trained for multiple times aiming at the pulse signals of the normal section and the pulse signal characteristics of different types of interference sections.
Step 6: the measured data is used to evaluate the average performance of the decision tree model.
Step 6.1: the classification accuracy was evaluated using kappa coefficients, which were calculated using the formula:
in the method, in the process of the invention,is kappa coefficient,/->、/>And->Respectively the column matrix +.>Line element and->Column elements and elements located on the diagonal. Parameter->Is the number of features. />Representing a classification category. Results of kappa coefficientThe closer the value is to 1, the better the result.
Defined as the average accuracy of each group, the formula is as follows: />
Step 6.2: the accuracy, specificity and sensitivity are adopted to evaluate the omission factor, false detection rate and accuracy of the model, and the calculation formula is as follows:
in the method, in the process of the invention,representing accuracy, ->Express specificity,/->Indicating sensitivity. />Indicating that the positive class samples are predicted the correct amount, +.>Indicating that the negative class samples are predicted the correct amount, +.>Representing the number of prediction errors for the negative class sample; />Representing misprediction of positive class samplesNumber of errors.
Specifically, in step 1, the experimental data set includes an offline signal of a normal pulse signal section and an interference section, which are collected and stored by the built pulse signal collection system. The pulse signal acquisition system includes: ESP32 singlechip, HKG-07B infrared pulse sensor, system integration base, signal conditioning circuit and bluetooth transmission module.
Specifically, in step 2, after the pulse signal peak value is extracted, the pulse signal is subjected to cycle segmentation, and data between each two adjacent peaks is analyzed.
Specifically, in step 4, the normal segment signal, the sensor placement signal, and the finger movement signal are grouped in pairs, and then the feature of significant change is selected by using the double-sample KS test.
Specifically, in step 5.1, 80% of the data is used as a training set training model, and the remaining 20% of the data is used as a test set to evaluate model performance.
Specifically, in step 6, accuracy, sensitivity and specificity are adopted to evaluate the average performance, the real-time detection result of the pulse wave interference section is evaluated, and the average value +/-standard deviation of each parameter is used as the evaluation result.
Specifically, in step 6.2, the normal section signal, the sensor placement signal and the finger movement signal are grouped in pairs, and then the accuracy, the specificity and the sensitivity of classification are calculated.
(2) And (5) online detection of the photoelectric volume pulse wave interference section.
And using the trained model for real-time detection of the photoelectric volume pulse wave interference section based on the decision tree for online detection of the photoelectric volume pulse wave interference section.
Specifically, fig. 5 is a schematic structural diagram of a real-time detection system for the photoplethysmogram interference band based on a decision tree. The system consists of a signal acquisition lower computer, a wireless communication module and a signal processing upper computer.
Furthermore, the signal acquisition lower computer consists of a pulse sensor and a microprocessor. The pulse sensor is used for acquiring pulse signals of a preset time length at the fingertip part of the human body. The microprocessor filters and normalizes the signals, and the microprocessor can be a single chip microcomputer, a DSP, and the like, but is not limited to the processors.
Still further, the communication manner of the wireless communication module may be bluetooth, a local area network or Wifi, but is not limited to these communication manners.
Still further, the signal processing host computer may be a microprocessor, a wearable device, a PC, a smart phone. The photoelectric volume pulse wave interference section real-time detection model based on the decision tree after training can be embedded into a microprocessor, wearable equipment, PC and smart phone to realize detection of the photoelectric volume pulse wave interference section. The microprocessor may be a single-chip microcomputer, a DSP, or the like, but is not limited to these processors. Meanwhile, the operating systems of the PC and the smart phone may be Windows, android or iOS.
The following is a description of specific examples:
in this embodiment, the PPG signals of 20 students in the sitting state of the university and healthy university are collected, the sampling frequency of the signals is 250Hz, the collection time is 90 seconds, 30 minutes of data are collected, and the data set has 4880 groups of pulse interval data.
The experimental equipment configuration includes: the AMD Ryzen 7 5800H processor, with a clock frequency set at 3.2GHz, was equipped with 16GB RAM and was loaded with a 64-bit Windows-11 operating system. Simulation software: matlab 2021b.
In this embodiment, fig. 4a-4d are pulse signal waveforms including interference sections, the photoelectric pulse sensor is clamped at the finger tip, and the finger movement causes the sensor to slide or even slide on the finger tip. In the experiment, when the sensor is worn by the subject, the interference section shown in fig. 4a is generated, and the sensor slides down due to the finger force and then is replaced, so that the interference section shown in fig. 4b is generated. The finger is moved "side-to-side", "up-down", "rotating" causing the sensor to slide over the fingertip creating the interference zone shown in figures 4c and 4 d.
In this embodiment, further, for the feature threshold crossing count, a threshold crossing method is selected, 3 different thresholds are selected according to the characteristics of the normalized data, respectively 1, -0.8 and-0.4, and the number of times of crossing between the signal and the different thresholds is calculated respectively. Table 1 shows the results of feature selection.
TABLE 1 feature extraction results (mean.+ -. Standard deviation)
In the present embodiment of the present invention,threshold crossing count representing threshold 1, +.>Threshold crossing count representing a threshold of-0.4, ">Threshold crossing count representing a threshold of-0.8, ">Represents the slope of adjacent peak points, +.>Representing the ratio of pulse periods +.>Deviation coefficient representing pulse signal, +.>Representing the slope of the data between adjacent peaks of the pulse.
In this embodiment, further, fig. 4a and 4b are both interference sections formed by sensor placement, so they are classified as sensor placement type interference sections, and are given to the tag 2; both fig. 4c and fig. 4d are interference segments due to finger movement, thus classifying them as finger movement type interference segments, giving the tag 3; the normal segment signal is given to tag 1.
In this embodiment, three types of data are grouped three times, the data of the tags 1 and 2 are the first group, the data of the tags 1 and 3 are the second group, the data of the tags 2 and 3 are the third group, and the double sample KS test is used to select the significant changeFeatures. Table 2 shows the results of feature selection. The results show that the method has the advantages of,without significant difference, the feature should be removed, and the other features all have significant difference +.>. Thus in this study the division characteristic +.>All other features are used to detect the interference section.
Table 2 feature selection results
In this embodiment, 80% of the data is used as a training set, the remaining 20% of the data is used as a test set, the model is trained by cross validation, at least 2 samples are arranged on the leaf nodes in the process of constructing the decision tree, and the characteristics of the training data set are randomly changed for 200 times, and the classification model is obtained by training for 200 times.
In this embodiment, the signal is subjected to three-class discrimination, and the kappa coefficient is used to evaluate the classification accuracy. Table 3 shows kappa coefficients of the classification model.
Table 3 decision Tree results of three classifications of tag 1, 2, 3 data
In this embodiment, in order to calculate the accuracy, specificity and sensitivity of the proposed interference section detection algorithm, the data of the tag 1 in the first group represents a positive class sample, and the data of the tag 2 represents a negative class sample; in the second group, the data of the tag 1 is a positive type sample, and the data of the tag 3 is a negative type sample; in the third group, the data of the label 2 is a positive class sample, the data of the label 3 is a negative class sample, and a classification decision tree model is trained to perform two classifications. 80% of the data is used as a training set training model, and the remaining 20% of the data is used as a test set to evaluate model performance. At least 2 samples are arranged on the leaf nodes, the characteristics of the training data set are randomly changed for 200 times, and the training is carried out for 200 times, so that the average result of the accuracy, the specificity and the sensitivity of three groups is obtained. Table 4 shows the average results of accuracy, specificity and sensitivity of the classification model.
TABLE 4 accuracy, specificity, sensitivity results (mean.+ -. Standard deviation) for three sets of data
The foregoing examples are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the foregoing examples, and any other changes, modifications, substitutions, combinations, and simplifications that do not depart from the spirit and principles of the present invention should be made therein and are intended to be equivalent substitutes within the scope of the present invention.
Claims (9)
1. A real-time detection method of photoelectric volume pulse wave interference section based on decision tree is characterized by comprising the following steps:
s01: acquiring pulse signals and preprocessing;
s02: extracting pulse signal peak values to obtain pulse main wave interval data;
s03: extracting specific variation parameters of pulse main wave interval data to form a feature vector set; the specific variation parameters include:
threshold crossing count, slope of adjacent peak points, ratio of pulse period, skewness coefficient of pulse signal, slope of data between adjacent peaks of pulse signal;
threshold crossing count calculation formula:
，
，
in the method, in the process of the invention,representation->Pass threshold->Is>Indicate->Peak and->Data set between peaks +.>，/>Index value of peak value, < >>Is a function->Is used for the input parameters of the (a),，/>for stopping the total number of peak points extracted at the time of acquisition, < > for>Is->Index increment of (a);
slope calculation formula of adjacent peak points:
，
in the method, in the process of the invention,for peak amplitude +.>Is->Peak point and->Slope of each peak point;
the ratio calculation formula of pulse period:
，
in the method, in the process of the invention,is->Peak point position and->Ratio of index values of the peak points;
the calculation formula of the deviation coefficient of the pulse signal comprises the following steps:
，
in the method, in the process of the invention,is->Peak and->Deviation coefficient of data between peak values +.>Is->Peak and->Standard deviation of data between peaks +.>Is->Peak and->Mean value of data between peaks +.>Is->Index increment initial value, ++>Is->Peak and->The data length of the data between the peak values;
the slope calculation formula of pulse wave between adjacent wave peaks:
，
in the method, in the process of the invention,is->Peak and->Slope of data between peaks, +.>Is->Peak and thMaximum value of data between peaks, +.>Is->Peak and->Minimum of data between peaks, +.>Is->Peak and->Minimum position of data between peaks;
s04: based on the decision tree classification model, an interference section real-time detection model is constructed and trained, and the trained interference section real-time detection model is used for carrying out interference section real-time detection.
2. The method for real-time detection of photoplethysmography interference according to claim 1, wherein the preprocessing method in step S01 includes:
filtering the pulse signal by 0Hz and 50Hz integer coefficient notch filters and a low pass filter with a cut-off frequency of 62.5 Hz;
the filter calculation formula of the low-pass filter with the cutoff frequency of 62.5 and Hz comprises the following steps:
，
in the method, in the process of the invention,is the signal after being filtered by the low-pass filter, < >>For the filter order +.>Is the original pulse signal +>Sampling points->Is a 19 th order filter coefficient obtained by the fir1 function in Matlab, < >>Is the original pulse signal;
the low-pass filter sliding window formula is:
，
in the method, in the process of the invention,for the signal array after low-pass filter, the size of array buffer is 321, +.>The low-pass filtering result of the earliest sampling point is put in the highest position of the buffer zone in the window and marked as +.>Every time data is updated, the low-pass filtering result is shifted by one bit from high to low, and the earliest result is +.>Covered by high order bits, new data is placed in；
The 0Hz and 50Hz integer coefficient notch filter calculation formula:
，
in the method, in the process of the invention,is the filtered signal of the integer notch filter, < >>Is the signal filtered by the 1 st order integer notch filter, +>Is the signal filtered by the 6 th order integer notch filter, +>Is the signal filtered by the 11 th order integer notch filter, +>Is the signal filtered by the 1 st low-pass filter,>is the 156 th low pass filter filtered signal,is the signal filtered by the 161 th low-pass filter,>is the 321 th low pass filter filtered signal.
3. The method for real-time detection of photoplethysmography interference according to claim 2, wherein the preprocessing method in step S01 further comprises: performing sliding standard normalization on the pulse signals, wherein the sliding standard normalization method comprises the following steps:
calculating the mean value and variance of the data in the sliding window, then carrying out standard normalization on the data, and finally repeating the standard normalization on the data in the sliding window after shifting;
the pulse signal sliding window formula is:
，
in the method, in the process of the invention,for in-window caching data +.>For the signal buffer size,/>Is the%>A number of sampling points are used to sample the sample,for the earliest sampling point, the sampling point which is acquired latest is placed at the highest position of a buffer zone in a window and is marked as +.>Every time data is updated, the sampling point moves one bit from high to low, and the earliest acquired +.>Covered by high order bits, new data is placed in；
Pulse signal sliding standard normalization formula:
，
，
，
，
in the method, in the process of the invention,normalized results for data, ++>For buffering data mean value in the filter back window, +.>For filtering the variance of the buffered data in the back window, < >>For buffering the data sum in the filter back window, the normalized result is stored in an arrayIn (I)>For the 1 st normalization result, analogize, < + >>And normalizing the result for the last time when the signal is stopped.
4. The method for real-time detection of photoplethysmography interference band based on decision tree according to claim 3, wherein the method for extracting pulse signal peak value in step S02 includes:
build size ofConstructing a matrix comprising only elements 0 and 1 +.>：
，
In the method, in the process of the invention,for sliding normalized data, +.>For 40 sampling points +.>, />，/>，/>Representing matrix->Is>A row;
reconstruction matrixCalculate->Sum of each row:
，
in the method, in the process of the invention,is provided with->One-dimensional matrix of 0 elements->；
Peak detection:
，
in the method, in the process of the invention,is the%>And peak point index values.
5. The method for real-time detection of photoplethysmography interference band based on decision tree according to claim 1, wherein the step S03 further comprises:
classifying the acquired pulse signals according to the interference section generation type, including: a normal segment signal, a sensor placement signal, a finger movement signal;
the normal segment signal, the sensor placement signal, and the finger movement signal are grouped in pairs, and feature selection is performed using a double sample KS test, excluding features that do not have significant differences.
6. The method for real-time detection of photoplethysmography interference band based on decision tree according to claim 1, wherein the training interference band real-time detection model in step S04 uses kappa coefficients to evaluate classification accuracy, and the kappa coefficients calculate the formula:
，
，
，
in the method, in the process of the invention,is kappa coefficient,/->、/>And->Respectively the column matrix +.>Line element and->Column element and element sum located on diagonal, +.>Is the number of features, +.>Representing a classification category;
calculating the average accuracy of each groupThe method comprises the following steps:
，
wherein,、/>、/>。
7. the method for real-time detection of photoplethysmogram interference band based on decision tree according to claim 1, wherein in step S04, the accuracy, the specificity and the sensitivity are adopted to evaluate the model miss rate, the false rate and the accuracy, and the calculation formula is as follows:，
，
，
in the method, in the process of the invention,representing accuracy, ->Express specificity,/->Indicating sensitivity, & lt>Indicating that the positive class samples are predicted the correct amount, +.>Indicating that the negative class samples are predicted the correct amount, +.>Representing the number of prediction errors for the negative class sample;indicating the number of prediction errors to be made for the positive class samples.
8. A real-time detection system of photoelectric volume pulse wave interference section based on decision tree is characterized by comprising:
the pulse signal acquisition processing module acquires pulse signals and performs preprocessing;
the pulse signal peak value extraction module is used for extracting a pulse signal peak value to obtain pulse main wave interval data;
the specific variation parameter extraction module is used for extracting specific variation parameters of the pulse main wave interval data to form a feature vector set; the specific variation parameters include:
threshold crossing count, slope of adjacent peak points, ratio of pulse period, skewness coefficient of pulse signal, slope of data between adjacent peaks of pulse signal;
threshold crossing count calculation formula:
，
，
in the method, in the process of the invention,representation->Pass threshold->Is>Indicate->Peak and->Data set between peaks +.>，/>Index value of peak value, < >>Is a function->Is used for the input parameters of the (a),，/>for stopping the total number of peak points extracted at the time of acquisition, < > for>Is->Index increment of (a);
slope calculation formula of adjacent peak points:
，
in the method, in the process of the invention,for peak amplitude +.>Is->Peak point and->Slope of each peak point;
the ratio calculation formula of pulse period:
，
in the method, in the process of the invention,is->Peak point position and->Ratio of index values of the peak points;
the calculation formula of the deviation coefficient of the pulse signal comprises the following steps:
，
in the method, in the process of the invention,is->Peak and->Deviation coefficient of data between peak values +.>Is->Peak and->Standard deviation of data between peaks +.>Is->Peak and->Mean value of data between peaks +.>Is->Index increment initial value, ++>Is->Peak and->The data length of the data between the peak values;
the slope calculation formula of pulse wave between adjacent wave peaks:
，
in the method, in the process of the invention,is->Peak and->Slope of data between peaks, +.>Is->Peak and thMaximum value of data between peaks, +.>Is->Peak and->Minimum of data between peaks, +.>Is->Peak and->Minimum position of data between peaks;
the interference section real-time detection model training detection module is used for constructing an interference section real-time detection model based on the decision tree classification model, training, and carrying out interference section real-time detection by using the trained interference section real-time detection model.
9. A computer storage medium having stored thereon a computer program, which when executed implements the decision tree-based method for real-time detection of photoplethysmography wave interference bands according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311526698.6A CN117224092B (en) | 2023-11-16 | 2023-11-16 | Photoelectric volume pulse wave interference band real-time detection method and system based on decision tree |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311526698.6A CN117224092B (en) | 2023-11-16 | 2023-11-16 | Photoelectric volume pulse wave interference band real-time detection method and system based on decision tree |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117224092A CN117224092A (en) | 2023-12-15 |
| CN117224092B true CN117224092B (en) | 2024-02-09 |
Family
ID=89098880
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311526698.6A Active CN117224092B (en) | 2023-11-16 | 2023-11-16 | Photoelectric volume pulse wave interference band real-time detection method and system based on decision tree |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117224092B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102440768A (en) * | 2010-10-13 | 2012-05-09 | 兰州理工大学 | Method for extracting pulse waveform feature points |
| CN106659900A (en) * | 2014-08-18 | 2017-05-10 | 卡梅伦保健公司 | Implantable medical device with self-correlation means and with analysis means for estimating cardiac rate |
| CN107157492A (en) * | 2017-05-19 | 2017-09-15 | 国家电网公司 | A kind of embedded human physiologic information non-invasive detection system and data processing method |
| CN109157202A (en) * | 2018-09-18 | 2019-01-08 | 东北大学 | A kind of cardiovascular disease early warning system based on more physiological signal depth integrations |
| KR20190100888A (en) * | 2018-02-21 | 2019-08-29 | 연세대학교 원주산학협력단 | Sleep stage classification device and method using pulse oximeter |
| CN110956197A (en) * | 2019-10-28 | 2020-04-03 | 新绎健康科技有限公司 | Method and system for establishing pulse wave noise signal identification model based on convolutional neural network |
| CN114886403A (en) * | 2022-05-13 | 2022-08-12 | 常熟理工学院 | Malignant arrhythmia identification and prediction system based on pulse main wave interval |
| CN116616762A (en) * | 2023-05-25 | 2023-08-22 | 武汉中旗生物医疗电子有限公司 | Pulse wave blood oxygen saturation detection method and blood oxygen saturation detection device |
| CN116649924A (en) * | 2023-06-06 | 2023-08-29 | 湖南敬凯投资管理有限公司 | Pulse analysis method and device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3918980A1 (en) * | 2020-06-02 | 2021-12-08 | CSEM Centre Suisse d'Electronique et de Microtechnique SA - Recherche et Développement | Method for classifying photoplethysmography pulses and monitoring of cardiac arrhythmias |
-
2023
- 2023-11-16 CN CN202311526698.6A patent/CN117224092B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102440768A (en) * | 2010-10-13 | 2012-05-09 | 兰州理工大学 | Method for extracting pulse waveform feature points |
| CN106659900A (en) * | 2014-08-18 | 2017-05-10 | 卡梅伦保健公司 | Implantable medical device with self-correlation means and with analysis means for estimating cardiac rate |
| CN107157492A (en) * | 2017-05-19 | 2017-09-15 | 国家电网公司 | A kind of embedded human physiologic information non-invasive detection system and data processing method |
| KR20190100888A (en) * | 2018-02-21 | 2019-08-29 | 연세대학교 원주산학협력단 | Sleep stage classification device and method using pulse oximeter |
| CN109157202A (en) * | 2018-09-18 | 2019-01-08 | 东北大学 | A kind of cardiovascular disease early warning system based on more physiological signal depth integrations |
| CN110956197A (en) * | 2019-10-28 | 2020-04-03 | 新绎健康科技有限公司 | Method and system for establishing pulse wave noise signal identification model based on convolutional neural network |
| CN114886403A (en) * | 2022-05-13 | 2022-08-12 | 常熟理工学院 | Malignant arrhythmia identification and prediction system based on pulse main wave interval |
| CN116616762A (en) * | 2023-05-25 | 2023-08-22 | 武汉中旗生物医疗电子有限公司 | Pulse wave blood oxygen saturation detection method and blood oxygen saturation detection device |
| CN116649924A (en) * | 2023-06-06 | 2023-08-29 | 湖南敬凯投资管理有限公司 | Pulse analysis method and device |
Non-Patent Citations (1)
| Title |
|---|
| 脉搏信号的干扰段检测与质量评估;张爱华 等;中国医疗器械杂志;第235-239页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117224092A (en) | 2023-12-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11529103B2 (en) | Artificial intelligence self-learning-based automatic electrocardiography analysis method and apparatus | |
| CN107495962B (en) | Sleep automatic staging method for single-lead electroencephalogram | |
| Smital et al. | Real-time quality assessment of long-term ECG signals recorded by wearables in free-living conditions | |
| WO2021135672A1 (en) | Signal detection method and system for assessing sleep apnea | |
| WO2020119245A1 (en) | Wearable bracelet-based emotion recognition system and method | |
| WO2019100565A1 (en) | Method and device for self-learning dynamic electrocardiography analysis employing artificial intelligence | |
| CN109758145B (en) | Automatic sleep staging method based on electroencephalogram causal relationship | |
| CN113349752B (en) | Wearable device real-time heart rate monitoring method based on sensing fusion | |
| CN108392211A (en) | A kind of fatigue detection method based on Multi-information acquisition | |
| Banerjee et al. | PhotoECG: Photoplethysmographyto estimate ECG parameters | |
| CN112869716B (en) | Pulse feature identification system and method based on two-channel convolutional neural network | |
| CN112971795B (en) | Electrocardiosignal quality evaluation method | |
| CN111557658B (en) | PPG real-time heart rate signal quality evaluation method and device and storage medium | |
| KR20220150901A (en) | Fusion signal processing for detection of maternal uterine activity | |
| Gupta et al. | Higher order derivative-based integrated model for cuff-less blood pressure estimation and stratification using PPG signals | |
| CN112294264A (en) | Sleep staging method based on BCG and blood oxygen saturation rate | |
| CN107550484B (en) | Magnetocardiogram signal quality evaluation method and system | |
| CN117224092B (en) | Photoelectric volume pulse wave interference band real-time detection method and system based on decision tree | |
| CN108338777A (en) | A kind of pulse signal determination method and device | |
| CN113679369A (en) | Heart rate variability evaluation method, intelligent wearable device and storage medium | |
| Farhan et al. | Myocardial Ischemia Detection from Slope of ECG ST Segment | |
| Saadi et al. | Low-cost ecg monitoring system with classification using deep learning | |
| Talatov et al. | Methodology for processing and analysis of diagnostic indicators electrocardiogram based on Labview | |
| Jokić et al. | An efficient approach for heartbeat classification | |
| Paul et al. | Mental stress detection using multimodal characterization of PPG signal for personal healthcare applications |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |