CN113892939A - Method for monitoring respiratory frequency of human body in resting state based on multi-feature fusion - Google Patents
Method for monitoring respiratory frequency of human body in resting state based on multi-feature fusion Download PDFInfo
- Publication number
- CN113892939A CN113892939A CN202111130709.XA CN202111130709A CN113892939A CN 113892939 A CN113892939 A CN 113892939A CN 202111130709 A CN202111130709 A CN 202111130709A CN 113892939 A CN113892939 A CN 113892939A
- Authority
- CN
- China
- Prior art keywords
- respiratory frequency
- physiological
- data
- monitoring
- human body
- 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
- 230000036391 respiratory frequency Effects 0.000 title claims abstract description 83
- 238000012544 monitoring process Methods 0.000 title claims abstract description 57
- 238000000034 method Methods 0.000 title claims abstract description 27
- 230000000284 resting effect Effects 0.000 title claims abstract description 19
- 230000004927 fusion Effects 0.000 title claims abstract description 16
- 238000010801 machine learning Methods 0.000 claims abstract description 18
- 238000013186 photoplethysmography Methods 0.000 claims description 30
- 238000005070 sampling Methods 0.000 claims description 22
- 230000000241 respiratory effect Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 7
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 5
- 239000008280 blood Substances 0.000 claims description 5
- 210000004369 blood Anatomy 0.000 claims description 5
- 229910052760 oxygen Inorganic materials 0.000 claims description 5
- 239000001301 oxygen Substances 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 230000029058 respiratory gaseous exchange Effects 0.000 claims description 3
- 238000007500 overflow downdraw method Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 abstract description 13
- 238000001514 detection method Methods 0.000 abstract description 3
- 108010064719 Oxyhemoglobins Proteins 0.000 abstract 1
- 238000005259 measurement Methods 0.000 abstract 1
- 230000036387 respiratory rate Effects 0.000 description 9
- 238000000605 extraction Methods 0.000 description 7
- 210000000038 chest Anatomy 0.000 description 6
- 238000012545 processing Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011982 device technology Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000013138 pruning Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 210000000115 thoracic cavity Anatomy 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/08—Detecting, measuring or recording devices for evaluating the respiratory organs
- A61B5/0816—Measuring devices for examining respiratory frequency
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/024—Detecting, measuring or recording pulse rate or heart rate
- A61B5/02416—Detecting, measuring or recording pulse rate or heart rate using photoplethysmograph signals, e.g. generated by infrared radiation
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/145—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue
- A61B5/14542—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue for measuring blood gases
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7253—Details of waveform analysis characterised by using transforms
- A61B5/726—Details of waveform analysis characterised by using transforms using Wavelet transforms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Molecular Biology (AREA)
- Veterinary Medicine (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Heart & Thoracic Surgery (AREA)
- Public Health (AREA)
- Signal Processing (AREA)
- Physiology (AREA)
- Artificial Intelligence (AREA)
- Psychiatry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Cardiology (AREA)
- Pulmonology (AREA)
- Power Engineering (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Optics & Photonics (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a respiratory frequency monitoring method in a resting state of a human body based on multi-feature fusion, belongs to the field of medical special software, relates to the technical field of human body detection, and comprises the steps of integrating a filter frequency selection algorithm and a machine learning model, and fitting and outputting respiratory frequency based on multi-body-position electrocardiosignals, photoplethysmogram signals and statistical features thereof in combination with physiological parameters such as heart rate, pulse rate, oxyhemoglobin saturation and the like which are convenient to measure. The invention improves the accuracy and generalization performance of the respiratory frequency monitoring technology, reduces the cost of respiratory frequency monitoring, and is generally suitable for the respiratory frequency measurement in the monitoring scene.
Description
Technical Field
The invention relates to the technical field of human body detection, in particular to a respiratory frequency monitoring method based on multi-feature fusion in a human body resting state.
Background
Currently, respiratory rate monitoring technologies mainly include three monitoring technologies, namely monitoring based on respiratory motion, monitoring based on respiratory airflow, and monitoring based on other physiological signals. Observation and judgment of the three-concave sign in the respiratory motion-based monitoring technology are highly dependent on the professional knowledge of a monitor; the monitoring technology based on the respiratory airflow can not be separated from the complex equipment for measuring the air pressure, the temperature and the change of the thoracic impedance; compared with the two methods, the respiratory rate monitoring based on other physiological signals has the characteristics of low cost, high automation and high convenience.
The frequency selection of the filter and the extraction of machine learning characteristics are the main methods for processing other physiological signal data, and the automation degree of respiratory frequency monitoring is improved. Wavelet transforms and band-pass filtering are commonly used in other physiological signal processing based on filter frequency selection. Other physiological signal processing based on machine learning feature extraction mainly adopts electrocardiosignal ECG and photoplethysmography PPG signal, and the respiratory frequency is fitted through models such as logistic regression and support vector machine. Time domain, frequency domain and statistical features of PPG and ECG signals are also gradually introduced into respiratory rate detection techniques based on machine learning feature extraction. Meanwhile, the development of wearable equipment technology enables vital signs such as electrocardiosignals ECG, photoplethysmography PPG, heart rate HR, blood oxygen saturation SpO2 and the like to be monitored and recorded in real time through a wireless body area network or a body sensor network.
The existing respiratory frequency monitoring technology based on machine learning feature extraction has the following 3 main problems:
1. in the characteristic selection stage, electrocardiosignal ECG and photoplethysmography (PPG) signals are still taken as main components, and characteristics such as 104-dimensional P-peak amplitude, Q-peak amplitude, R-peak amplitude, S-peak amplitude and the like introduced by partial technologies are still calculated based on the electrocardiosignal ECG and the photoplethysmography (PPG) signals. However, advances in wearable device technology have enabled various physiological parameters such as heart rate HR, blood oxygen saturation SpO2, pulse rate pause, etc. to be monitored and recorded in real time.
2. Machine learning characteristic engineering such as principal component extraction, one-hot coding and the like cannot effectively process the pollution of noise to a data source. Meanwhile, machine learning models such as neural networks have high data processing capability, but have high requirements on computing resources and are not suitable for being embedded into respiratory frequency monitoring hardware equipment.
3. The imbalance between the physiological signal and the physiological parameter data causes the overfitting phenomenon of the respiratory rate monitoring model to be serious, and the diversified respiratory rate monitoring task in the actual monitoring scene cannot be objectively processed.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a method for monitoring the respiratory frequency of a human body in a resting state based on multi-feature fusion, so that the convenience and the automation degree of respiratory frequency monitoring are improved, the influence of noise on the fitting accuracy of respiratory signals is reduced, the accuracy and the generalization performance of a respiratory frequency monitoring technology are improved, and the cost of respiratory frequency monitoring is reduced.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows:
a respiratory frequency monitoring method under a resting state of a human body based on multi-feature fusion comprises the following steps:
step 1, collecting a plurality of physiological signals and physiological parameter data of a human body in a resting state to form a prediction sample library;
step 2, calculating physiological signals of a prediction sample library by using a wavelet denoising algorithm, and replacing original physiological signal data with denoised physiological signals as new physiological signal data to form a denoising prediction sample library;
step 3, extracting statistical characteristics of part of physiological signals of the noise reduction prediction sample library, and combining the statistical characteristics into the original noise reduction prediction sample library to form a preliminary respiratory frequency fitting library;
step 4, standardizing the data of each field in the preliminary respiratory frequency fitting library by using a data standardization algorithm, and replacing the original data with the standardized data to construct a respiratory frequency fitting library;
step 5, training a respiratory frequency fitting model by using a machine learning model based on a tree according to a respiratory frequency fitting library;
step 6, fusing the wavelet denoising algorithm in the step 2, the statistical characteristics of the partial physiological signals in the step 3 and the respiratory frequency fitting model in the step 5 to construct a respiratory frequency monitoring model;
and 7, inputting part of physiological signals of the person with the respiratory frequency to be monitored into the respiratory frequency monitoring model, and outputting the respiratory frequency of the person with the respiratory frequency to be monitored.
The technical scheme of the invention is further improved as follows: in step 1, the physiological signals comprise photoplethysmography PPG, chest lead electrocardiosignals VECG, standard lead electrocardiosignals IIECG, pressurized unipolar limb lead electrocardiosignals AVRECG and respiration signals RESP; the physiological parameters comprise heart rate HR, blood oxygen saturation SpO2 and pulse rate PLUSE; collecting age selection stages of a plurality of people including young, middle-aged and old, wherein the number of people of each tested person is not less than P, and simultaneously measuring physiological signals and physiological parameters within a period of continuous time A seconds, wherein the sampling frequency of the physiological signals is B Hz, and the sampling frequency of the physiological parameters is C Hz;
the step 1 comprises the following substeps:
1.1, according to the measured physiological signal data, using a physiological signal sampling TIME stamp TIME1 and an identifier ID of a person to be measured as a main key, and using the physiological signal in the step 1 as other fields to construct a physiological signal sample library;
1.2, according to the measured physiological parameter data, using a physiological data sampling TIME stamp TIME2 and an identifier ID of a person to be measured as a main key, and using the physiological parameter in the step 1 as other fields to construct a physiological parameter sample library;
1.3, performing left external connection on a measured physiological signal sample library and a physiological parameter sample library, and using a left external connection result as a prediction sample library; wherein the left external connection conditions are as follows: the testee identifiers ID are equal and the result of rounding down the physiological signal sampling timestamp TIME1 is equal to the result of rounding down the physiological parameter timestamp.
The technical scheme of the invention is further improved as follows: in step 2, the wavelet denoising algorithm is a wavelet threshold denoising algorithm.
The technical scheme of the invention is further improved as follows: in step 3, the partial physiological signals are photoplethysmography PPG, chest lead electrocardiosignals VECG, standard lead electrocardiosignals IIECG and pressurized unipolar limb lead electrocardiosignals AVRECG; the statistical characteristic types are Kurt and Skaew;
the method comprises the following substeps:
3.1, adding a kurtosis value field and a skewness value field of each partial physiological signal to a noise reduction prediction sample library;
3.2, respectively calculating the Kurt value and the Skaw value of each partial physiological signal as the values of the corresponding fields by using a Kurt value and skewness value Skaw calculation method according to the corresponding data of the partial physiological signal of each sample in the noise reduction prediction sample library;
3.3, deleting the ID field of the identifier of the person to be tested, the TIME1 field of the physiological signal sampling TIME stamp, the field of the physiological parameter sampling TIME stamp and the corresponding data thereof on the noise reduction prediction sample library to form a preliminary respiratory frequency fitting library.
The technical scheme of the invention is further improved as follows: in step 4, the data normalization algorithm is a standard deviation normalization algorithm.
The technical scheme of the invention is further improved as follows: in step 5, the tree-based machine learning model is an extreme gradient lifting tree model, the target output of the respiratory frequency fitting model is a respiratory signal RESP in a respiratory frequency fitting library, and the rest fields in the respiratory frequency fitting library are target inputs.
The technical scheme of the invention is further improved as follows: in step 6, the fusion method is as follows: according to the output data of the wavelet denoising algorithm, the statistical characteristics of partial physiological signals are calculated, and the output data of the wavelet denoising algorithm, the statistical characteristic data of the partial physiological signals and the physiological parameter data are horizontally spliced to be used as the input of the respiratory frequency fitting model.
The technical scheme of the invention is further improved as follows: in step 7, the partial physiological signal is defined the same as the partial physiological signal in step 3.1.
The technical scheme of the invention is further improved as follows: in step 1.3, the young subject is aged 18 to 45 years, the middle subject is aged 46 to 59 years, the elderly subject is aged 59 years or older, P is a positive integer of not less than 5, a is a positive integer of not less than 300, B is a positive integer of not less than 125, and C is a positive integer of not less than 1.
Due to the adoption of the technical scheme, the invention has the technical progress that:
the invention improves the convenience and automation degree of respiratory frequency monitoring, reduces the influence of noise on the fitting accuracy of respiratory signals, improves the accuracy and generalization performance of respiratory frequency monitoring technology, and reduces the cost of respiratory frequency monitoring. According to the invention, the physiological signals and physiological parameters which can be measured by wearable equipment in the sensor network are selected, the statistical characteristics of the physiological signals are combined, and the respiratory frequency is fitted based on multiple dimensions, so that the convenience of respiratory frequency monitoring is improved, and the cost of respiratory frequency monitoring is reduced.
The invention reduces the influence of noise pollution on physiological signal characteristic extraction by fusing a noise reduction algorithm and a machine learning model and using a wavelet noise reduction technology, reduces the operation cost in practical application by using a basic lightweight tree machine learning model, and improves the respiratory frequency monitoring speed.
The invention reduces the sampling error of the training data by collecting physiological signals and physiological parameter monitoring data of the tested persons in different ages, and improves the generalization performance of the respiratory frequency monitoring model by combining the special pruning process of the tree machine learning model.
The invention uses special software, particularly a respiratory frequency monitoring technology, adopts various physiological signals and physiological parameters which can be conveniently obtained, integrates a wavelet noise reduction method and a machine learning model, uses physiological data of detected persons in different age groups, ensures the accuracy and generalization performance of the respiratory frequency monitoring model, and effectively solves the problems of high cost and poor convenience of the respiratory frequency monitoring technology.
Drawings
FIG. 1 is a flow chart of a respiratory rate monitoring method of the present invention;
FIG. 2 is a sub-flow diagram of the construction of a prediction sample library in the present invention.
Detailed Description
The invention is described in further detail below with reference to the following figures and examples:
as shown in fig. 1, a method for monitoring respiratory rate of a human body in a resting state based on multi-feature fusion specifically includes the following steps:
step 1, collecting a plurality of physiological signals and physiological parameter data of a human body in a resting state to form a prediction sample library, as shown in fig. 2.
Defining physiological signals to be collected as photoplethysmography PPG, chest lead electrocardiosignals VECG, standard lead electrocardiosignals IIECG, pressurized unipolar limb lead electrocardiosignals AVRECG and respiration signals RESP;
defining the physiological parameters to be collected as heart rate HR, blood oxygen saturation SpO2 and pulse rate PLUSE;
the method also specifically comprises the following substeps:
1.1, selecting tested persons in the age stage of young, middle and old, wherein the number of people of each tested person is not less than P, and simultaneously measuring physiological signals and physiological parameters within a period of A seconds, wherein the sampling frequency of the physiological signals is B hertz, and the sampling frequency of the physiological parameters is C hertz;
wherein the age of young testees is 18-45 years old, the age of middle testees is 46-59 years old, the age of old testees is above 59 years old, P is a positive integer not less than 5, A is a positive integer not less than 300, B is a positive integer not less than 125, and C is a positive integer not less than 1.
In this example, P is 6, a is 480, B is 125, C is 1, and the number of subjects selected is 44, among which 9 subjects aged 18 to 45, 6 subjects aged 46 to 59, and 14 subjects aged 59 or older.
1.2, according to the measured physiological signal data, using the physiological signal sampling TIME stamp TIME1 and the identifier ID of the person to be measured as the main key, and using each physiological signal as the rest fields, a physiological signal sample library is constructed.
Two examples of samples of the physiological signal sample library in the present embodiment are: [ (ID:1), (TIME1:0), (PPG:0.68719), (VECG:0.51562), (IIECG:0.48047), (AVRECG:0.60588), (RESP:0.50635) ], [ (ID:1), (TIME1:0.008), (PPG:0.68328), (VECG:0.51562), (IIECG:0.48438), (AVRECG:0.59608), (RESP:0.5132) ].
And 1.3, according to the measured physiological parameter data, using a physiological data sampling TIME stamp TIME2 and an identifier ID of a person to be measured as a main key, and using each physiological parameter as the rest fields to construct a physiological parameter sample library.
An example of a sample of the physiological parameter sample library in this embodiment is: [ (ID:1), (TIME2:0), (HR:86), (SpO2:97), (PLUSE:88) ].
1.4, performing left external connection on the measured physiological signal sample library and the physiological parameter sample library, wherein the left external connection conditions are as follows: the testees' identifiers ID are equal and the result of rounding down the physiological signal sampling timestamp TIME1 is equal to the result of rounding down the physiological parameter timestamp, and the left outer connection result is used as the prediction sample library.
Two examples of samples of the prediction sample library in this embodiment are: [ (ID:1), (TIME1:0), (TIME2:0), (PPG:0.68719), (VECG:0.51562), (IIECG:0.48047), (AVRECG:0.60588), (HR:86), (SpO2:97), (PLUSE:88), (RESP:0.50635) ], [ (ID:1), (TIME1:0.008), (TIME2:0), (PPG:0.68328), (VECG:0.51562), (IIECG:0.48438), (AVRECG:0.59608), (HR:86), (SpO2:97), (PLUSE:88), (RESP:0.5132) ].
And 2, respectively calculating the physiological signals subjected to noise reduction on the prediction sample library by using a wavelet noise reduction algorithm to serve as new physiological signal data to replace the original physiological signal data to form a noise reduction prediction sample library.
The wavelet denoising algorithm used in the present embodiment is a wavelet threshold algorithm.
And 3, respectively extracting the statistical characteristics of partial physiological signals according to the noise reduction prediction sample library, and combining the statistical characteristics into the noise reduction prediction sample library to form a preliminary respiratory frequency fitting library.
Defining the partial physiological signals as photoplethysmography PPG, chest lead electrocardiosignals VECG, standard lead electrocardiosignals IIECG and pressurized unipolar limb lead electrocardiosignals AVRECG;
defining the statistical characteristics as Kurt and skewness Skaew, wherein the calculation formulas are respectively as follows:
wherein x isiIs the data of the i-th field,
is the average value of all sample data in the ith field, and n is the total number of samples.
The method specifically comprises the following substeps:
3.1, adding a kurtosis value field and a skewness value field of each partial physiological signal to a noise reduction prediction sample library;
the fields added in this embodiment are: the device comprises a photoplethysmography peak value PPG _ Kurt, a chest lead electrocardiosignal peak value VECG _ Kurt, a standard lead electrocardiosignal peak value IIECG _ Kurt, a pressurized unipolar limb lead electrocardiosignal peak value AVRECG _ Kurt, a photoplethysmography pulse skewness value PPG _ Skaew, a chest lead electrocardiosignal skewness value VECG _ Skaew, a standard lead electrocardiosignal skewness value IIECG _ Skaew and a pressurized unipolar limb lead electrocardiosignal skewness value AVRECG _ Skaew.
3.2, respectively calculating the Kurt value and the Skaw value of each partial physiological signal as the values of the corresponding fields by using a Kurt value and skewness value Skaw calculation method according to the corresponding data of the partial physiological signal of each sample in the noise reduction prediction sample library;
3.3, deleting the ID field of the identifier of the person to be tested, the TIME1 field of the physiological signal sampling timestamp, the TIME2 field of the physiological parameter sampling timestamp and corresponding data on the noise reduction prediction sample library to form a preliminary respiratory frequency fitting library.
One example of a sample of the respiratory rate fit library in this embodiment is: [ (PPG:0.68719), (VECG:0.51562), (IIECG:0.48047), (AVRECG:0.60588), (HR:86), (SpO2:97), (PLUSE:88), (RESP:0.50635), (PPG _ Kurt: -0.637488), (VECG _ Kurt:13.43306296), (IIECG _ Kurt:10.746857), (AVRECG _ Kurt:13.463033), (PPG _ Skaew: 0.084163), (VECG _ Skaew: -3.429179), (IIECG _ Skaew: 3.2293148), (AVRECG _ Skaew: -3.555338) ].
And 4, respectively calculating the normalized value of the data of each field on the preliminary respiratory frequency fitting library by using a data normalization algorithm, replacing the original data, and constructing the respiratory frequency fitting library.
The data normalization algorithm in this embodiment is a standard deviation normalization algorithm.
And 5, training a respiratory frequency fitting model by using a machine learning model based on a tree according to the respiratory frequency fitting library.
In this embodiment, the tree-based machine learning model is an extreme gradient lifting tree model, in which: the maximum depth of the tree is 9, the L1 regularization pre-term coefficient is 0.3, the L2 regularization pre-term coefficient is 0.1, and the learning rate is 0.1.
And 6, fusing the wavelet denoising algorithm in the step 2, the statistical characteristics of the partial physiological signals in the step 3 and the respiratory frequency fitting model in the step 5 to construct a respiratory frequency monitoring model.
The statistical characteristics of partial physiological signals are calculated according to the output data of the wavelet denoising algorithm, and the output data of the wavelet denoising algorithm, the statistical characteristic data of the partial physiological signals and the physiological parameter data are horizontally spliced to be used as the input of the respiratory frequency fitting model.
And 7, inputting part of physiological signals of the person with the respiratory frequency to be monitored into the respiratory frequency monitoring model, and outputting the respiratory frequency of the person with the respiratory frequency to be monitored.
The respiratory rate monitoring model in this embodiment is represented by the following test data: the average absolute error range is 0 to 0.05, the root mean square error range is 0 to 0.1, the square value range of the correlation coefficient is 0.9 to 0.98, the model error is low and the prediction effect is good.
In conclusion, the invention improves the convenience and automation degree of respiratory frequency monitoring and reduces the influence of model overfitting and noise on the respiratory signal monitoring accuracy rate by fusing the filter frequency selection and the machine learning feature extraction method and based on the diversified physiological data of the tested persons in different age groups.
Claims (9)
1. A respiratory frequency monitoring method under a resting state of a human body based on multi-feature fusion is characterized by comprising the following steps: the method comprises the following steps:
step 1, collecting a plurality of physiological signals and physiological parameter data of a human body in a resting state to form a prediction sample library;
step 2, calculating physiological signals of a prediction sample library by using a wavelet denoising algorithm, and replacing original physiological signal data with denoised physiological signals as new physiological signal data to form a denoising prediction sample library;
step 3, extracting statistical characteristics of part of physiological signals of the noise reduction prediction sample library, and combining the statistical characteristics into the original noise reduction prediction sample library to form a preliminary respiratory frequency fitting library;
step 4, standardizing the data of each field in the preliminary respiratory frequency fitting library by using a data standardization algorithm, and replacing the original data with the standardized data to construct a respiratory frequency fitting library;
step 5, training a respiratory frequency fitting model by using a machine learning model based on a tree according to a respiratory frequency fitting library;
step 6, fusing the wavelet denoising algorithm in the step 2, the statistical characteristics of the partial physiological signals in the step 3 and the respiratory frequency fitting model in the step 5 to construct a respiratory frequency monitoring model;
and 7, inputting part of physiological signals of the person with the respiratory frequency to be monitored into the respiratory frequency monitoring model, and outputting the respiratory frequency of the person with the respiratory frequency to be monitored.
2. The method for monitoring the respiratory frequency of the human body in the resting state based on the multi-feature fusion as claimed in claim 1, wherein: in step 1, the physiological signals comprise photoplethysmography PPG, chest lead electrocardiosignals VECG, standard lead electrocardiosignals IIECG, pressurized unipolar limb lead electrocardiosignals AVRECG and respiration signals RESP; the physiological parameters comprise heart rate HR, blood oxygen saturation SpO2 and pulse rate PLUSE; collecting age selection stages of a plurality of people including young, middle-aged and old, wherein the number of people of each tested person is not less than P, and simultaneously measuring physiological signals and physiological parameters within a period of continuous time A seconds, wherein the sampling frequency of the physiological signals is B Hz, and the sampling frequency of the physiological parameters is C Hz;
the step 1 comprises the following substeps:
1.1, according to the measured physiological signal data, using a physiological signal sampling TIME stamp TIME1 and an identifier ID of a person to be measured as a main key, and using the physiological signal in the step 1 as other fields to construct a physiological signal sample library;
1.2, according to the measured physiological parameter data, using a physiological data sampling TIME stamp TIME2 and an identifier ID of a person to be measured as a main key, and using the physiological parameter in the step 1 as other fields to construct a physiological parameter sample library;
1.3, performing left external connection on a measured physiological signal sample library and a physiological parameter sample library, and using a left external connection result as a prediction sample library; wherein the left external connection conditions are as follows: the testee identifiers ID are equal and the result of rounding down the physiological signal sampling timestamp TIME1 is equal to the result of rounding down the physiological parameter timestamp.
3. The method for monitoring the respiratory frequency of the human body in the resting state based on the multi-feature fusion as claimed in claim 1, wherein: in step 2, the wavelet denoising algorithm is a wavelet threshold denoising algorithm.
4. The method for monitoring the respiratory frequency of the human body in the resting state based on the multi-feature fusion as claimed in claim 1, wherein: in step 3, the partial physiological signals are photoplethysmography PPG, chest lead electrocardiosignals VECG, standard lead electrocardiosignals IIECG and pressurized unipolar limb lead electrocardiosignals AVRECG; the statistical characteristic types are Kurt and Skaew;
the method comprises the following substeps:
3.1, adding a kurtosis value field and a skewness value field of each partial physiological signal to a noise reduction prediction sample library;
3.2, respectively calculating the Kurt value and the Skaw value of each partial physiological signal as the values of the corresponding fields by using a Kurt value and skewness value Skaw calculation method according to the corresponding data of the partial physiological signal of each sample in the noise reduction prediction sample library;
3.3, deleting the ID field of the identifier of the person to be tested, the TIME1 field of the physiological signal sampling TIME stamp, the field of the physiological parameter sampling TIME stamp and the corresponding data thereof on the noise reduction prediction sample library to form a preliminary respiratory frequency fitting library.
5. The method for monitoring the respiratory frequency of the human body in the resting state based on the multi-feature fusion as claimed in claim 1, wherein: in step 4, the data normalization algorithm is a standard deviation normalization algorithm.
6. The method for monitoring the respiratory frequency of the human body in the resting state based on the multi-feature fusion as claimed in claim 1, wherein: in step 5, the tree-based machine learning model is an extreme gradient lifting tree model, the target output of the respiratory frequency fitting model is a respiratory signal RESP in a respiratory frequency fitting library, and the rest fields in the respiratory frequency fitting library are target inputs.
7. The method for monitoring the respiratory frequency of the human body in the resting state based on the multi-feature fusion as claimed in claim 1, wherein: in step 6, the fusion method is as follows: according to the output data of the wavelet denoising algorithm, the statistical characteristics of partial physiological signals are calculated, and the output data of the wavelet denoising algorithm, the statistical characteristic data of the partial physiological signals and the physiological parameter data are horizontally spliced to be used as the input of the respiratory frequency fitting model.
8. The method for monitoring the respiratory frequency of the human body in the resting state based on the multi-feature fusion as claimed in claim 1 or 4, wherein: in step 7, the partial physiological signal is defined the same as the partial physiological signal in step 3.1.
9. The method for monitoring the respiratory frequency of the human body in the resting state based on the multi-feature fusion as claimed in claim 2, wherein: in step 1.3, the young subject is aged 18 to 45 years, the middle subject is aged 46 to 59 years, the elderly subject is aged 59 years or older, P is a positive integer of not less than 5, a is a positive integer of not less than 300, B is a positive integer of not less than 125, and C is a positive integer of not less than 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111130709.XA CN113892939A (en) | 2021-09-26 | 2021-09-26 | Method for monitoring respiratory frequency of human body in resting state based on multi-feature fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111130709.XA CN113892939A (en) | 2021-09-26 | 2021-09-26 | Method for monitoring respiratory frequency of human body in resting state based on multi-feature fusion |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113892939A true CN113892939A (en) | 2022-01-07 |
Family
ID=79029660
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111130709.XA Pending CN113892939A (en) | 2021-09-26 | 2021-09-26 | Method for monitoring respiratory frequency of human body in resting state based on multi-feature fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113892939A (en) |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB0807039D0 (en) * | 2008-04-17 | 2008-05-21 | Oxford Biosignals Ltd | Method and apparatus for measuring breathing rate |
| US20150105632A1 (en) * | 2013-10-11 | 2015-04-16 | Xhale, Inc. | Fusion of data from multiple sources for non-invasive detection of respiratory parameters |
| CN106022477A (en) * | 2016-05-18 | 2016-10-12 | 国网信通亿力科技有限责任公司 | Intelligent analysis decision system and method |
| CN106897776A (en) * | 2017-01-17 | 2017-06-27 | 华南理工大学 | A kind of continuous type latent structure method based on nominal attribute |
| CN107122643A (en) * | 2017-04-07 | 2017-09-01 | 西安电子科技大学 | Personal identification method based on PPG signals and breath signal Fusion Features |
| CN107194138A (en) * | 2016-01-31 | 2017-09-22 | 青岛睿帮信息技术有限公司 | A kind of fasting blood-glucose Forecasting Methodology based on physical examination data modeling |
| CN108446676A (en) * | 2018-05-03 | 2018-08-24 | 南京信息工程大学 | Facial image age method of discrimination based on orderly coding and multilayer accidental projection |
| CN109394188A (en) * | 2018-11-27 | 2019-03-01 | 中山大学 | A kind of adnormal respiration detection method, device and equipment based on heart rate variability |
| EP3479758A1 (en) * | 2017-11-03 | 2019-05-08 | Tata Consultancy Services Limited | System and method for breathing pattern extraction from ppg signals |
| US20200093425A1 (en) * | 2018-09-21 | 2020-03-26 | Tata Consultancy Services Limited | System and method for non-apnea sleep arousal detection |
| CN110929644A (en) * | 2019-11-22 | 2020-03-27 | 南京甄视智能科技有限公司 | Heuristic algorithm-based multi-model fusion face recognition method and device, computer system and readable medium |
| CN111084618A (en) * | 2019-12-13 | 2020-05-01 | 安徽通灵仿生科技有限公司 | Wearable multifunctional respiration cycle detection system and method |
| CN111568400A (en) * | 2020-05-20 | 2020-08-25 | 山东大学 | Human body sign information monitoring method and system |
| CN111714105A (en) * | 2020-07-24 | 2020-09-29 | 长春理工大学 | Human vital sign perception system based on IPPG |
| CN112101326A (en) * | 2020-11-18 | 2020-12-18 | 北京健康有益科技有限公司 | Multi-person posture recognition method and device |
| US20210000356A1 (en) * | 2019-07-02 | 2021-01-07 | Tata Consultancy Services Limited | Method and system for screening and monitoring of cardiac diseases |
| CN112998690A (en) * | 2021-03-29 | 2021-06-22 | 华南理工大学 | Pulse wave multi-feature fusion-based respiration rate extraction method |
-
2021
- 2021-09-26 CN CN202111130709.XA patent/CN113892939A/en active Pending
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB0807039D0 (en) * | 2008-04-17 | 2008-05-21 | Oxford Biosignals Ltd | Method and apparatus for measuring breathing rate |
| US20150105632A1 (en) * | 2013-10-11 | 2015-04-16 | Xhale, Inc. | Fusion of data from multiple sources for non-invasive detection of respiratory parameters |
| CN107194138A (en) * | 2016-01-31 | 2017-09-22 | 青岛睿帮信息技术有限公司 | A kind of fasting blood-glucose Forecasting Methodology based on physical examination data modeling |
| CN106022477A (en) * | 2016-05-18 | 2016-10-12 | 国网信通亿力科技有限责任公司 | Intelligent analysis decision system and method |
| CN106897776A (en) * | 2017-01-17 | 2017-06-27 | 华南理工大学 | A kind of continuous type latent structure method based on nominal attribute |
| CN107122643A (en) * | 2017-04-07 | 2017-09-01 | 西安电子科技大学 | Personal identification method based on PPG signals and breath signal Fusion Features |
| EP3479758A1 (en) * | 2017-11-03 | 2019-05-08 | Tata Consultancy Services Limited | System and method for breathing pattern extraction from ppg signals |
| CN108446676A (en) * | 2018-05-03 | 2018-08-24 | 南京信息工程大学 | Facial image age method of discrimination based on orderly coding and multilayer accidental projection |
| US20200093425A1 (en) * | 2018-09-21 | 2020-03-26 | Tata Consultancy Services Limited | System and method for non-apnea sleep arousal detection |
| CN109394188A (en) * | 2018-11-27 | 2019-03-01 | 中山大学 | A kind of adnormal respiration detection method, device and equipment based on heart rate variability |
| US20210000356A1 (en) * | 2019-07-02 | 2021-01-07 | Tata Consultancy Services Limited | Method and system for screening and monitoring of cardiac diseases |
| CN110929644A (en) * | 2019-11-22 | 2020-03-27 | 南京甄视智能科技有限公司 | Heuristic algorithm-based multi-model fusion face recognition method and device, computer system and readable medium |
| CN111084618A (en) * | 2019-12-13 | 2020-05-01 | 安徽通灵仿生科技有限公司 | Wearable multifunctional respiration cycle detection system and method |
| CN111568400A (en) * | 2020-05-20 | 2020-08-25 | 山东大学 | Human body sign information monitoring method and system |
| CN111714105A (en) * | 2020-07-24 | 2020-09-29 | 长春理工大学 | Human vital sign perception system based on IPPG |
| CN112101326A (en) * | 2020-11-18 | 2020-12-18 | 北京健康有益科技有限公司 | Multi-person posture recognition method and device |
| CN112998690A (en) * | 2021-03-29 | 2021-06-22 | 华南理工大学 | Pulse wave multi-feature fusion-based respiration rate extraction method |
Non-Patent Citations (3)
| Title |
|---|
| LE TRANG T,FU WEIXUAN,MOORE JASON H;: "Scaling tree-based automated machine learning to biomedical big data with a feature set selector", 《BIOINFORMATICS》 * |
| MD. NAZMUL ISLAM SHUZAN; MOAJJEM HOSSAIN CHOWDHURY ; MD. SHAFAYE: "A Novel Non-Invasive Estimation of Respiration Rate From Motion Corrupted Photoplethysmograph Signal Using Machine Learning Model", 《IEEE ACCESS》 * |
| 邓宝芸，潘燕，孟延: "基于心电和脉搏波数据融合的呼吸率估计", 《中国生物医学工程学报》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110420019B (en) | Deep regression heart rate estimation method for ballistocardiogram signals | |
| CN113349752B (en) | Wearable device real-time heart rate monitoring method based on sensing fusion | |
| Belgacem et al. | ECG based human authentication using wavelets and random forests | |
| CN108498089A (en) | A kind of noninvasive continuous BP measurement method based on deep neural network | |
| CN111466876A (en) | Alzheimer's disease auxiliary diagnosis system based on fNIRS and graph neural network | |
| CN113095302B (en) | Depth model for arrhythmia classification, method and device using same | |
| CN111887858B (en) | Ballistocardiogram signal heart rate estimation method based on cross-modal mapping | |
| CN108577830A (en) | A kind of user oriented sign information dynamic monitor method and dynamic monitor system | |
| CN112806977B (en) | Physiological parameter measuring method based on multi-scale fusion network | |
| Zhao et al. | An IoT-based wearable system using accelerometers and machine learning for fetal movement monitoring | |
| Chen et al. | Machine learning method for continuous noninvasive blood pressure detection based on random forest | |
| CN109288508A (en) | A kind of pressure value intelligent measurement method based on CRNN-BP | |
| CN107890339A (en) | A kind of sleep stage detection method and wearable sleep stage detection means | |
| Gupta et al. | Higher order derivative-based integrated model for cuff-less blood pressure estimation and stratification using PPG signals | |
| CN110200619A (en) | It is a kind of to establish major cardiovascular disease risks Early-warning Model method and device | |
| CN110313900A (en) | It is a kind of based on deep learning to the classification method of knee joint swinging signal | |
| JP5544365B2 (en) | Improvements in multi-parameter monitoring or improvements related to multi-parameter monitoring | |
| CN113171080A (en) | Energy metabolism evaluation method and system based on wearable sensing information fusion | |
| Bashar et al. | Extraction of heart rate from PPG Signal: a machine learning approach using decision tree regression algorithm | |
| CN115089145A (en) | Intelligent blood pressure prediction method based on multi-scale residual error network and PPG signal | |
| CN113576475B (en) | Deep learning-based contactless blood glucose measurement method | |
| CN115089179A (en) | Psychological emotion insights analysis method and system | |
| CN110693510A (en) | Attention deficit hyperactivity disorder auxiliary diagnosis device and using method thereof | |
| CN113033358A (en) | Cuff-free blood pressure modeling method based on deep neural network | |
| CN117100239A (en) | Blood pressure detection system based on CNN+LSTM and PPG signals |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220107 |
|
| RJ01 | Rejection of invention patent application after publication |