CN114246563A - Intelligent heart and lung function monitoring equipment based on millimeter wave radar - Google Patents

Intelligent heart and lung function monitoring equipment based on millimeter wave radar Download PDF

Info

Publication number
CN114246563A
CN114246563A CN202111553397.3A CN202111553397A CN114246563A CN 114246563 A CN114246563 A CN 114246563A CN 202111553397 A CN202111553397 A CN 202111553397A CN 114246563 A CN114246563 A CN 114246563A
Authority
CN
China
Prior art keywords
vital sign
millimeter wave
wave radar
signals
signal
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.)
Granted
Application number
CN202111553397.3A
Other languages
Chinese (zh)
Other versions
CN114246563B (en
Inventor
曹海林
龙凤
王彬宇
孙志伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University
Original Assignee
Chongqing University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chongqing University filed Critical Chongqing University
Priority to CN202111553397.3A priority Critical patent/CN114246563B/en
Publication of CN114246563A publication Critical patent/CN114246563A/en
Application granted granted Critical
Publication of CN114246563B publication Critical patent/CN114246563B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—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
    • A61B5/0205—Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
    • 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
    • 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
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Physiology (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Veterinary Medicine (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Psychiatry (AREA)
  • Cardiology (AREA)
  • Power Engineering (AREA)
  • Pulmonology (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

The invention provides a heart and lung function intelligent monitoring device based on a millimeter wave radar, which comprises: the system comprises a radar vital sign signal extraction module and a disease classification prediction module; the radar vital sign signal extraction module comprises a millimeter wave radar sensor and a vital sign signal extraction module; the millimeter wave radar sensor transmits a millimeter wave radar signal to the abdomen of the emergency patient and receives an echo signal; the millimeter wave radar sensor is electrically connected with the vital sign signal extraction module; extracting vital sign signals of the emergency patient from the echo signals; the disease classification prediction module is in communication connection with the vital sign signal extraction module, and is used for extracting data characteristics of the vital sign signals and performing classification disease prediction on normal or abnormal vital sign data by using the self-adaptive continuous learning network. The invention can solve the technical problem that the disease suffered by the emergency patient can not be pre-judged according to different heartbeat and breathing abnormalities of the emergency patient in the prior art.

Description

Intelligent heart and lung function monitoring equipment based on millimeter wave radar
Technical Field
The invention relates to the technical field of monitoring of human cardio-pulmonary functions, in particular to intelligent monitoring equipment for cardio-pulmonary functions based on a millimeter wave radar.
Background
The parameters of respiration and heartbeat are important judgment bases for judging whether the heart-lung activity of a human body is normal or not, the heart-lung activity of the human body directly influences the activities of various organs and muscles, and different diseases usually cause different abnormalities of the heart-heart and respiration of the human body.
In the prior art, CN110840422A provides a monitor, a cab and a monitoring system, and discloses a non-contact small monitor, a cab and a heartbeat and respiration monitoring system for a driver, wherein the non-contact small monitor comprises a shell, a millimeter wave sensor and the like, and further comprises a liquid crystal display screen or a communication module, wherein the millimeter wave sensor is used for monitoring heartbeat and respiration in real time, a monitoring result is displayed in real time through a liquid crystal display screen, or the monitoring result is sent to the internet through the communication module for remote transmission, so as to realize remote real-time monitoring. The monitor has the beneficial effect of portability, solves the problem that large-scale equipment in hospitals is inconvenient to move, and can be widely popularized and used.
However, when the monitor in the above technical scheme is used in an ambulance to perform non-contact cardiopulmonary function monitoring on an emergency patient, only the heartbeat and respiratory conditions of the emergency patient can be monitored in real time, and the disease suffered by the emergency patient cannot be predicted according to different heartbeat and respiratory abnormalities of the emergency patient; this results in the inability to take symptomatic first aid measures in an ambulance and the inability to bring the hospital to a relevant diagnosis and treatment preparation in advance.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a millimeter wave radar-based heart and lung function intelligent monitoring device, which aims to solve the technical problem that the disease suffered by an emergency patient can not be pre-judged according to different heartbeat and breathing abnormalities of the emergency patient in the prior art.
The technical scheme adopted by the invention is as follows:
in a first implementation manner, an intelligent monitoring device for cardiopulmonary function based on millimeter wave radar is provided, including: the system comprises a radar vital sign signal extraction module and a disease classification prediction module;
the radar vital sign signal extraction module comprises a millimeter wave radar sensor and a vital sign signal extraction module; the millimeter wave radar sensor transmits a millimeter wave radar signal to the abdomen of the emergency patient and receives an echo signal; the millimeter wave radar sensor is electrically connected with the vital sign signal extraction module; extracting vital sign signals of the emergency patient from the echo signals; the vital sign signals comprise respiration signals and heartbeat signals;
the disease classification prediction module is in communication connection with the vital sign signal extraction module, and is used for extracting data characteristics of the vital sign signals and performing disease classification prediction on normal or abnormal vital sign data by using the self-adaptive continuous learning network.
According to the technical scheme of the first implementation mode, the beneficial technical effects of the invention are as follows: the vital sign signals of the first-aid patients monitored at present can be input into the trained self-adaptive continuous learning network model for prediction, and whether the current vital sign signals are abnormal or not and possible diseases are output, so that symptomatic first-aid measures can be taken for the first-aid patients, and relevant diagnosis and treatment preparations can be made in advance by hospitals.
With reference to the first implementable manner, in a second implementable manner, the method for extracting the respiration and heartbeat signals of the emergency patient comprises the following steps:
extracting distance information in the echo signal by using distance fast Fourier transform to obtain a maximum amplitude point in a frequency spectrum;
performing phase extraction and descrambling on the maximum amplitude point to obtain an intermediate frequency signal;
and performing spectrum analysis on the intermediate frequency signal by using fast Fourier transform to obtain respiration and heartbeat signals.
With reference to the second implementation manner, in a third implementation manner, the phase descrambling is calculated according to the following formula:
Figure BDA0003418421820000021
in the above equation, [ phi ] (m) denotes a phase value, unwarp is a phase difference detection function, Q denotes a received in-phase signal, and I denotes a quadrature signal.
With reference to the first implementable manner, in a fourth implementable manner, the performing data feature extraction on the vital sign signal, and performing classification prediction on normal or abnormal vital sign data includes:
constructing an adaptive continuous learning network, wherein the adaptive continuous learning network comprises the following steps: a data buffer, an encoder, a decoder, a predictor; the system comprises a data buffer area, an encoder, a decoder and a predictor which are connected in sequence; the data buffer area is divided into a new data buffer area and an old data buffer area; the number of the encoder and the number of the decoder are respectively 5; the predictor is a long-short term memory neural network;
taking historical data of respiration and heartbeat signals as input, taking historical data of disease classification results as output, training by using a self-adaptive continuous learning network, and storing a model;
newly acquired respiration and heartbeat signals of the emergency patients are sent to a trained self-adaptive continuous learning network model, the model performs result prediction by combining with old knowledge under the condition of learning new knowledge, and whether the signals are abnormal or not and possible follow-up emergency are predicted.
With reference to the fourth implementable manner, in the fifth implementable manner, when performing disease classification prediction, a flexible weight curing algorithm is used at the encoder, the decoder, and the predictor at the same time; the flexible weight curing algorithm formula is as follows:
Figure BDA0003418421820000031
in the above-mentioned formula, the compound of formula,
Figure BDA0003418421820000032
for the optimal weight matrix, P is the posterior probability value, Du,TIn the form of a set of data,
Figure BDA0003418421820000033
is the ith diagonal parameter of the information matrix at the time of the tth update;
Figure BDA0003418421820000034
is that
Figure BDA0003418421820000035
The ith value at the time of the T-1 th update; lambda [ alpha ]tAnd λpriorIs a hyper-parameter.
With reference to the first implementable manner, in a sixth implementable manner, the output of the adaptive continuous learning network is:
Figure BDA0003418421820000036
in the above-mentioned formula, the compound of formula,
Figure BDA0003418421820000037
predicting outcome for disease classification, xnIs a breathing signal and a heartbeat signal,
Figure BDA0003418421820000038
a weight coefficient matrix representing each layer of the network, L representing the number of layers of the neural network, and the output of each layer being a function
Figure BDA0003418421820000039
To indicate.
With reference to the first implementable manner, in a seventh implementable manner, the loss function of the adaptive continuous learning network is:
Figure BDA00034184218200000310
in the above formula, ynThe value is a label value, including abnormal or normal states of heartbeat and respiratory signals;
Figure BDA0003418421820000041
represents and ynCorresponding prediction results; y is the sum of the total weight of the components,
Figure BDA0003418421820000042
are respectively a plurality of ynAnd
Figure BDA0003418421820000043
a matrix of formations;
Figure BDA0003418421820000044
is the solved mean square error; λ R (Θ) is the regularization term.
With reference to the second implementation manner, in an eighth implementation manner, when performing phase descrambling on a maximum amplitude point to obtain an intermediate frequency signal, the method includes:
a phase difference between the forward difference and the backward difference is calculated for each differential phase, and when the phase difference is greater than a threshold value, the differential phase is replaced with an interpolated value.
According to the technical scheme of the eighth implementation mode, the beneficial technical effects of the invention are as follows: errors due to a plurality of noises in calculating the phase value can be reduced.
With reference to the eighth implementable manner, in a ninth implementable manner, the interpolation value is obtained by a differential and newton interpolation method.
In combination with the second implementable manner, in a tenth implementable manner, the band-pass filter is used to perform enhancement processing on the respiration and heartbeat signals, and then discrete wavelet transform processing is performed on the respiration and heartbeat signals passing through the band-pass filter.
According to the technical scheme of the tenth implementation mode, the beneficial technical effects of the invention are as follows: the accuracy of the digital processing of the signals can be further improved.
With reference to the tenth implementable manner, in an eleventh implementable manner, the band-pass filter is a butterworth band-pass filter, a passband cutoff frequency of the butterworth band-pass filter is 0.1Hz, and a stopband start frequency is 0.85 Hz.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. Throughout the drawings, like elements or portions are generally identified by like reference numerals. In the drawings, elements or portions are not necessarily drawn to scale.
Fig. 1 is a schematic flow chart of a radar vital sign signal extraction module extracting respiratory and heartbeat signals according to embodiment 1 of the present invention;
fig. 2 is a schematic diagram of an adaptive continuous learning network model architecture constructed in embodiment 1 of the present invention;
fig. 3 is a schematic diagram of a network layer structure of the adaptive continuous learning network after training in embodiment 1 of the present invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and therefore are only examples, and the protection scope of the present invention is not limited thereby.
It is to be noted that, unless otherwise specified, technical or scientific terms used herein shall have the ordinary meaning as understood by those skilled in the art to which the invention pertains.
Example 1
This embodiment provides a heart lung function intelligent monitoring equipment based on millimeter wave radar, includes: the device comprises a radar vital sign signal extraction module and a disease classification prediction module, wherein the two modules are integrated in the same portable device.
1. Radar vital sign signal extraction module:
the radar vital sign signal extraction module comprises: the millimeter wave radar sensor is electrically connected with the vital sign signal extraction module. The millimeter wave radar sensor transmits a millimeter wave radar signal to the abdomen of the emergency patient and receives an echo signal. Due to the fluctuation of the abdomen of the human body, the obtained radar echo signals are different, and the vital sign signal of the emergency patient is extracted from the echo signals by the vital sign signal extraction module; in this embodiment, the vital sign signals include a respiration signal and a heartbeat signal.
In the embodiment, a millimeter wave radar with high sensitivity is adopted, the working frequency is high, a high-frequency radio frequency signal is not easily interfered by a low-frequency radio frequency signal, and the detection precision of micro-motion is high; meanwhile, the millimeter wave radar can provide more concentrated transmitting beams with smaller size, which is beneficial to improving the detection accuracy and reducing the interference of other moving objects.
When the vital sign signal is extracted, the vital sign signal extraction module performs fast Fourier transform on the intermediate frequency signal of the millimeter wave radar sensor to obtain the frequency spectrum of the time signal at each distance point, finds the maximum amplitude point in the frequency spectrum, the maximum amplitude point is the distance unit where the measured target is located, calculates the phase at the maximum amplitude point, and subtracts the phase of the last sawtooth wave from the phase to obtain the phase difference, so that the vital sign signal can be obtained. In a specific embodiment, the vital sign signal extraction module is implemented in a form without limitation, such as: the hardware carrier is an ARM board loaded with a CPU processor.
In a specific embodiment, when the radar vital sign signal extraction module is used for extracting the respiratory signal and the heartbeat signal of the emergency patient, the emergency patient should lie down in a certain fixed area, such as a stretcher bed and a sickbed on an ambulance; the method comprises the following steps:
the first step is as follows: and extracting distance information in echo signals obtained by the millimeter wave radar sensor by using Range fast Fourier transform (Range FFT) to obtain a maximum amplitude point in a frequency spectrum.
The second step is that: extracting phase of the maximum amplitude point and descrambling to obtain intermediate frequency signal
Specifically, the phase extraction descrambling is calculated according to the following formula:
Figure BDA0003418421820000061
in the above equation, [ phi ] (m) denotes a phase value, an unwarp function is used to detect whether two adjacent phase differences are larger than pi, Q denotes a received in-phase signal, and I is a quadrature signal.
The third step: performing frequency spectrum analysis on the intermediate frequency signal by using fast Fourier transform to obtain respiration and heartbeat signals
Specifically, the respiratory signal is exemplified as: the respiration rate can be estimated using the peak-to-peak distance of its time-domain waveform; two thresholds are defined using the sampling rate and the allowable frequency range, for a respiration rate of 0.1-4.0Hz, i.e. the minimum peak distance PminAnd maximum peak distance Pmax. The breathing frequency is selected according to the frequency of the maximum peak in the breathing region spectrum: when the ratio of the maximum peak value to the signal power of the remaining frequency points in the spectrum of the breathing region is below a certain threshold, an estimate based on the peak-to-peak distance is selected as the breathing frequency.
2. A disease classification prediction module:
the disease classification prediction module is in communication connection with the vital sign signal extraction module and receives the respiration and heartbeat signals extracted by the vital sign signal extraction module.
In this embodiment, the acquired vital sign signals of the emergency patient are a combination of a plurality of signals that are continuous in time, and when the continuous signals are used for classifying and predicting diseases, in order to avoid forgetting information, in this embodiment, an adaptive continuous Learning network (continuous Learning) is used to extract data characteristics of the acquired vital sign signals, and classify and predict normal/abnormal vital sign data, which is specifically as follows:
the first step is as follows: constructing an adaptive continuous learning network, wherein the adaptive continuous learning network comprises the following steps: data buffer, encoder, decoder, predictor.
Specifically, the adaptive continuous learning network takes respiratory and heartbeat signals as input and takes a disease classification prediction result as output, and the following formula is shown:
Figure BDA0003418421820000071
in the above-mentioned formula, the compound of formula,
Figure BDA0003418421820000072
predicting outcome for disease classification, xnIs a breathing signal and a heartbeat signal,
Figure BDA0003418421820000073
a weight coefficient matrix of each layer of the network is represented, L represents the number of layers of the neural network, and the incidence relation between the previous layer and the next layer is embodied through the weight coefficient, so that the effect of feature extraction is achieved; function for output of each layer
Figure BDA0003418421820000074
To indicate.
The loss function of the neural network is shown as follows:
Figure BDA0003418421820000075
in the above formula, ynIs a label value including abnormal or normal states of heartbeat and respiratory signals (in the embodiment, 0 is used for normal, and 1 is used for abnormal);
Figure BDA0003418421820000076
represents and ynCorresponding prediction results; y is the sum of the total weight of the components,
Figure BDA0003418421820000077
are respectively a plurality of ynAnd
Figure BDA0003418421820000078
a matrix is formed.
Figure BDA0003418421820000079
The method is used for solving the mean square error, wherein the mean square error is data representing the difference between a prediction result and a real result, and the smaller the mean square error is, the better the mean square error is. The lambda R (theta) is a regularization term, and the phenomenon of overfitting can be avoided after the lambda R (theta) is introduced.
The self-adaptive continuous learning network is constructed by the following specific parts:
data Buffers (Buffers): in this embodiment, the buffer is divided into a new data buffer and an old data buffer. Let the probability distribution of the input data (i.e. respiration and heartbeat signals) of the neural network be P (X), and the probability distribution of the output result (i.e. disease classification prediction result) of the neural network be P (Y | X), and determine which buffer to store the data into by determining whether there is a distribution change in P (X) and P (Y | X).
Auto encoder (Autoencoder) and auto decoder (Autodecoder): reconstructing the input at the output of the neural network using an automatic encoder, the encoder and decoder being respectively denoted zn＝fΘ(xn) And
Figure BDA0003418421820000081
wherein Θ and Φ are parameter matrices; l (x)n,Φ,Θ)＝MSE(xn,gΦ(f Θ (xn))), the distribution variation of p (x) can be detected by the reconstruction error of the automatic encoder. When in use
Figure BDA0003418421820000082
And when the value is larger than the threshold value 1, storing the data into a new data buffer area, otherwise, storing the data into an old data buffer area. And when the new data buffer is full, the parameters of the automatic encoder are updated, and the process is circulated.
Predictor (Predictor): in this embodiment, the predictor preferably uses a long short-term memory neural network (LSTM) adapted to process and predict events with longer intervals and delays in the time series. Error of predicted value from actual value using predictor
Figure BDA0003418421820000083
To determine whether there is a change in distribution:
when in use
Figure BDA0003418421820000084
And when the value is larger than the threshold value 2, storing the data into a new data buffer area, otherwise, storing the data into an old data buffer area. And when the new data buffer is full, the parameters of the automatic encoder are updated, and the process is circulated.
The second step is that: and taking historical data of respiration and heartbeat signals as input, taking historical data of disease classification results as output, and training and storing the model by using the self-adaptive continuous learning network.
In this embodiment, a training sample data set is constructed by using historical data of respiration and heartbeat signals and historical data of disease classification results, the historical data of the respiration and heartbeat signals is used as input of a neural network, the historical data of the disease classification results is used as output of the neural network, a self-adaptive continuous learning network constructed in the previous step is trained, and a model obtained after training can predict disease classification according to the respiration and heartbeat signals, for example: the model predicts the abdominal pressure value according to the abnormal conditions of the heartbeat and the respiratory signal and outputs the disease classification prediction result of whether the peritonitis is possibly caused.
In this embodiment, as shown in fig. 3, input data first passes through an auto-encoder network layer consisting of 5 encoders and 5 decoders. The network layer of the automatic encoder can perform data compression for extracting the main characteristics of the input signal and reconstructing the input signal, so that the number of sequence units can be reduced, and the calculation time required by subsequent prediction can be shortened.
After the autoencoder network layer, 4 LSTM layers with attenuation coefficients are used to continue the feature extraction work. Finally, the network is tiled and expanded through a full connection layer (FC), and a disease classification prediction result is output.
The third step: newly acquired respiration and heartbeat signals of the emergency patients are sent to a trained self-adaptive continuous learning network model, the model performs result prediction by combining with old knowledge under the condition of learning new knowledge, and whether the signals are abnormal or not and possible follow-up emergency are predicted.
And after the model is stored, newly acquired and extracted respiration and heartbeat signals of the emergency patient by the radar vital sign signal extraction module are sent to the trained adaptive continuous learning network model for prediction.
When newly acquired respiratory and heartbeat signals of an emergency patient are used for disease classification prediction, a flexible weight consolidation algorithm (EWC algorithm) is used at an automatic encoder network layer and an LSTM predictor at the same time, and network parameter adjustment is carried out according to judgment that current data is new data and old data. Specifically, the EWC algorithm performs an approximate bayesian posterior probability on the model parameters of a given task. In an adaptive continuous learning network, data can be divided into two categories,namely the known task (T)k) And unknown tasks (T)u)。TkThe data set is D corresponding to the learned task of the adaptive continuous learning networkk,T-1。Dk,T-1Is the combination of all data stored in the new data buffer and the old data buffer during T-1 updates; at this time, Dk,T-1Has been learned before the Tth update, so Dk,T-1Belongs to a known task. T isuRefers to the content that the network will learn in the T-th update, and the corresponding data set is Du,T。Du,TOnly the data stored in the new data buffer for triggering the T-th update. Chronologically, the T-1 th update occurs before the T-th update. Thus, learning unknown task TuThe posterior probability values obtained can be expressed as:
Figure BDA0003418421820000091
the logarithm is taken at the left side and the right side of the formula simultaneously to obtain:
logP(Θ|Du,T,Dk,T-1)＝log(Du,T|Θ)+logP(Θ|Dk,T-1)-logP(Du,T|Dk,T-1)
when trained, the network can perform well the known task TkThen, the optimum parameters can be obtained
Figure BDA0003418421820000092
At this time-logP (Θ | D)k,T-1) With regard to Θ, it is approximately 0. Thus, -logP (Θ | D)k,T-1) Second order Taylor series expansion can be used to estimate
Figure BDA0003418421820000093
Figure BDA0003418421820000094
At the same time, will
Figure BDA0003418421820000095
Is approximated to
Figure BDA0003418421820000096
NkAs a data set Dk,T-1The total number of data in (a) is,
Figure BDA0003418421820000101
data set Dk,T-1Is determined by the empirical data matrix of (a),
Figure BDA0003418421820000102
is a front approximation
Figure BDA0003418421820000103
The value is obtained.
Estimating the information matrix in the high-dimensional space as a diagonal matrix with off-diagonal elements being zero, wherein the ith value on the diagonal is used
Figure BDA0003418421820000104
Expressed, the formula of the EWC algorithm is:
Figure BDA0003418421820000105
in the above-mentioned formula, the compound of formula,
Figure BDA0003418421820000106
for the optimal weight matrix, P is the posterior probability value, Du,TIn the form of a set of data,
Figure BDA0003418421820000107
is the ith diagonal parameter of the information matrix at the time of the tth update;
Figure BDA0003418421820000108
is that
Figure BDA0003418421820000109
The ith value at the time of the T-1 th update; lambda [ alpha ]tAnd λpriorIs a hyper-parameter.
The model uses an automatic encoder network layer and a predictor to judge whether the current newly input data is new data or not by combining the distribution change of P (X) and the distribution change of P (Y | X), and changes the parameters of a neural network, so that the disease classification prediction can be carried out by combining old knowledge under the condition of learning new knowledge.
By adopting the technical scheme of the embodiment, the vital sign signals of the first-aid patients monitored at present can be input into the trained self-adaptive continuous learning network model for prediction, and whether the current vital sign signals are abnormal or not and possible diseases are output, so that symptomatic first-aid measures can be taken for the first-aid patients, and hospitals can make relevant diagnosis and treatment preparation in advance
Meanwhile, the model modifies/adjusts the relevant weights on the previously trained task model every time, only modifies the weights which have small influence on the previous task, and does not modify the weights which have large influence on the previous task, so as to achieve the purpose of avoiding information forgetting.
Example 2
In the technical solution of embodiment 1, when the phase descrambling is performed on the maximum amplitude point and the intermediate frequency signal is obtained, a phase wrapping error caused by a plurality of noises may cause an inaccurate influence on the descrambled differential phase a (m). In order to solve the technical problems, the following technical scheme is adopted:
when the maximum amplitude point is subjected to phase descrambling to obtain an intermediate frequency signal, the phase difference between the forward difference a (m) -a (m +1) and the backward difference a (m) -a (m-1) of each differential phase a (m) is calculated, and when the phase difference is greater than a critical value, the differential phase a (m) is replaced by an interpolation value.
Specifically, the interpolated values are obtained by difference and newton interpolation, by which the values of the unknown data points can be predicted from the values of the known data points.
By adopting the technical scheme of the embodiment, the error caused by a plurality of noises when the phase value is calculated can be reduced.
Example 3
In the technical solution of embodiment 1, when the intermediate frequency signal is subjected to spectrum analysis by using fast fourier transform to obtain the respiration and heartbeat signals, noise interference is caused because clutter burrs generally exist in the echo signal in the time domain. In order to solve the technical problems, the following technical scheme is adopted:
and (3) enhancing the respiration and heartbeat signals by using a band-pass filter, and then performing discrete wavelet transform processing on the respiration and heartbeat signals passing through the band-pass filter.
Specifically, a Butterworth band-pass filter is added to realize band-pass filtering of respiration and heartbeat signals and filter noise of echo signals; preferably, the butterworth band-pass filter has a passband cut-off frequency of 0.1Hz and a stopband start frequency of 0.85 Hz.
Meanwhile, in order to effectively reflect the mutation degree of the signals in some parts, the respiration and heartbeat signals passing through the band-pass filter are subjected to discrete wavelet transformation, and the high-frequency components of the signals are extracted, so that data compensation is provided for a subsequent self-adaptive continuous learning network.
By adopting the technical scheme of the embodiment, the accuracy of digital processing of the signals can be further improved.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the present invention, and they should be construed as being included in the following claims and description.

Claims (11)

1. The utility model provides a heart lung function intelligent monitoring equipment based on millimeter wave radar which characterized in that includes: the system comprises a radar vital sign signal extraction module and a disease classification prediction module;
the radar vital sign signal extraction module comprises a millimeter wave radar sensor and a vital sign signal extraction module; the millimeter wave radar sensor transmits a millimeter wave radar signal to the abdomen of the emergency patient and receives an echo signal; the millimeter wave radar sensor is electrically connected with the vital sign signal extraction module; extracting vital sign signals of the emergency patient from the echo signals; the vital sign signals comprise respiration signals and heartbeat signals;
the disease classification prediction module is in communication connection with the vital sign signal extraction module, and is used for extracting data characteristics of the vital sign signals and performing disease classification prediction on normal or abnormal vital sign data by using a self-adaptive continuous learning network.
2. The millimeter wave radar-based intelligent heart and lung function monitoring device according to claim 1, wherein the step of extracting respiration and heartbeat signals of the emergency patient comprises the following steps:
extracting distance information in the echo signal by using distance fast Fourier transform to obtain a maximum amplitude point in a frequency spectrum;
performing phase extraction and descrambling on the maximum amplitude point to obtain an intermediate frequency signal;
and performing spectrum analysis on the intermediate frequency signal by using fast Fourier transform to obtain respiration and heartbeat signals.
3. The intelligent millimeter wave radar-based cardiopulmonary function monitoring device of claim 2, wherein the phase descrambling is calculated according to the following formula:
Figure FDA0003418421810000011
in the above equation, [ phi ] (m) denotes a phase value, unwarp is a phase difference detection function, Q denotes a received in-phase signal, and I denotes a quadrature signal.
4. The intelligent millimeter wave radar-based cardiopulmonary function monitoring device of claim 1, wherein performing data feature extraction on the vital sign signals and performing classification prediction on normal or abnormal vital sign data comprises:
constructing an adaptive continuous learning network, wherein the adaptive continuous learning network comprises the following steps: the system comprises a data buffer area, an encoder, a decoder and a predictor which are connected in sequence; the data buffer area is divided into a new data buffer area and an old data buffer area; the number of the encoder and the number of the decoder are respectively 5; the predictor is a long-short term memory neural network;
taking historical data of respiration and heartbeat signals as input, taking historical data of disease classification results as output, training by using a self-adaptive continuous learning network, and storing a model;
newly acquired respiration and heartbeat signals of the emergency patients are sent to a trained self-adaptive continuous learning network model, the model performs result prediction by combining with old knowledge under the condition of learning new knowledge, and whether the signals are abnormal or not and possible follow-up emergency are predicted.
5. The MMW radar-based intelligent heart and lung function monitoring device according to claim 4, wherein a flexible weight curing algorithm is used at the encoder, the decoder and the predictor simultaneously when disease classification prediction is performed; the flexible weight curing algorithm formula is as follows:
Figure FDA0003418421810000021
in the above-mentioned formula, the compound of formula,
Figure FDA0003418421810000022
for the optimal weight matrix, P is the posterior probability value, Du,TIn the form of a set of data,
Figure FDA0003418421810000023
is the ith diagonal parameter of the information matrix at the time of the tth update;
Figure FDA0003418421810000024
is that
Figure FDA0003418421810000025
The ith value at the time of the T-1 th update; lambda [ alpha ]tAnd λpriorIs a hyper-parameter.
6. The intelligent millimeter wave radar-based cardiopulmonary function monitoring device of claim 1, wherein the output of the adaptive continuous learning network is:
Figure FDA0003418421810000026
in the above-mentioned formula, the compound of formula,
Figure FDA0003418421810000027
predicting outcome for disease classification, xnIs a breathing signal and a heartbeat signal,
Figure FDA0003418421810000028
a weight coefficient matrix representing each layer of the network, L representing the number of layers of the neural network, and the output of each layer being a function
Figure FDA0003418421810000029
To indicate.
7. The MMR-based intelligent heart-lung function monitoring device of claim 1, wherein the loss function of the adaptive continuous learning network is:
Figure FDA00034184218100000210
in the above formula, ynThe value is a label value, including abnormal or normal states of heartbeat and respiratory signals;
Figure FDA00034184218100000211
represents and ynCorresponding prediction results; y is the sum of the total weight of the components,
Figure FDA0003418421810000031
are respectively a plurality of ynAnd
Figure FDA0003418421810000032
a matrix of formations;
Figure FDA0003418421810000033
is the solved mean square error; λ R (Θ) is the regularization term.
8. The intelligent monitoring equipment for cardiopulmonary function based on millimeter wave radar according to claim 2, wherein the phase descrambling is performed on the maximum amplitude point, and when the intermediate frequency signal is obtained, the method comprises the following steps:
a phase difference between the forward difference and the backward difference is calculated for each differential phase, and when the phase difference is greater than a threshold value, the differential phase is replaced with an interpolated value.
9. The millimeter wave radar-based intelligent heart and lung function monitoring device of claim 8, wherein the interpolated value is obtained by differential and newton interpolation.
10. The millimeter wave radar-based intelligent heart and lung function monitoring device according to claim 2, wherein a band-pass filter is used to enhance the respiration and heartbeat signals, and then discrete wavelet transform processing is performed on the respiration and heartbeat signals passing through the band-pass filter.
11. The millimeter wave radar-based intelligent cardiopulmonary function monitoring device of claim 10, wherein the band pass filter is a butterworth band pass filter, the passband cutoff frequency of the butterworth band pass filter is 0.1Hz, and the stopband start frequency is 0.85 Hz.
CN202111553397.3A 2021-12-17 2021-12-17 Heart and lung function intelligent monitoring equipment based on millimeter wave radar Active CN114246563B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111553397.3A CN114246563B (en) 2021-12-17 2021-12-17 Heart and lung function intelligent monitoring equipment based on millimeter wave radar

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111553397.3A CN114246563B (en) 2021-12-17 2021-12-17 Heart and lung function intelligent monitoring equipment based on millimeter wave radar

Publications (2)

Publication Number Publication Date
CN114246563A true CN114246563A (en) 2022-03-29
CN114246563B CN114246563B (en) 2023-11-17

Family

ID=80792775

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111553397.3A Active CN114246563B (en) 2021-12-17 2021-12-17 Heart and lung function intelligent monitoring equipment based on millimeter wave radar

Country Status (1)

Country Link
CN (1) CN114246563B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113892931A (en) * 2021-10-14 2022-01-07 重庆大学 Method for extracting and analyzing intra-abdominal pressure by FMCW radar based on deep learning
CN115844355A (en) * 2022-11-22 2023-03-28 哈尔滨理工大学 UWB (ultra wide band) biological radar echo signal heart rate extraction method

Citations (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100152600A1 (en) * 2008-04-03 2010-06-17 Kai Sensors, Inc. Non-contact physiologic motion sensors and methods for use
US20130053653A1 (en) * 2011-08-29 2013-02-28 General Electric Company Radar based systems and methods for monitoring a subject
US20150073266A1 (en) * 2013-09-06 2015-03-12 Covidien Lp System and method for lung visualization using ultasound
CN106073703A (en) * 2016-05-27 2016-11-09 重庆大学 A kind of athlete's life sign monitor system based on LoRa technology
CN107871531A (en) * 2016-09-27 2018-04-03 柯惠有限合伙公司 Crack is assessed and the planning of surgical operation Interventional
CN108983187A (en) * 2018-07-11 2018-12-11 西安电子科技大学 Online radar target identification method based on EWC
CN109965858A (en) * 2019-03-28 2019-07-05 北京邮电大学 Based on ULTRA-WIDEBAND RADAR human body vital sign detection method and device
CN110584631A (en) * 2019-10-10 2019-12-20 重庆邮电大学 Static human heartbeat and respiration signal extraction method based on FMCW radar
CN111481184A (en) * 2020-04-24 2020-08-04 华侨大学 Multi-target respiration heart rate monitoring method and system based on millimeter wave radar technology
CN111603138A (en) * 2020-05-19 2020-09-01 杭州电子科技大学 Sleep apnea monitoring system based on millimeter wave radar
CN111723527A (en) * 2020-06-28 2020-09-29 重庆大学 Gear residual life prediction method based on cocktail long-term and short-term memory neural network
US20210030276A1 (en) * 2019-07-29 2021-02-04 DawnLight Technologies Inc. Remote Health Monitoring Systems and Method
US20210093203A1 (en) * 2019-09-30 2021-04-01 DawnLight Technologies Systems and methods of determining heart-rate and respiratory rate from a radar signal using machine learning methods
CN112754431A (en) * 2020-12-31 2021-05-07 杭州电子科技大学 Respiration and heartbeat monitoring system based on millimeter wave radar and lightweight neural network
CN112754441A (en) * 2021-01-08 2021-05-07 杭州环木信息科技有限责任公司 Millimeter wave-based non-contact heartbeat detection method
CN112914545A (en) * 2019-12-05 2021-06-08 南京理工大学 Non-contact intestinal disease diagnostic equipment
CN112998668A (en) * 2021-02-06 2021-06-22 路晟悠拜(重庆)科技有限公司 Millimeter wave-based non-contact far-field multi-human-body respiration heart rate monitoring method
US20210325509A1 (en) * 2020-04-20 2021-10-21 Infineon Technologies Ag Radar-Based Vital Sign Estimation
CN113729641A (en) * 2021-10-12 2021-12-03 南京润楠医疗电子研究院有限公司 Non-contact sleep staging system based on conditional countermeasure network

Patent Citations (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100152600A1 (en) * 2008-04-03 2010-06-17 Kai Sensors, Inc. Non-contact physiologic motion sensors and methods for use
US20130053653A1 (en) * 2011-08-29 2013-02-28 General Electric Company Radar based systems and methods for monitoring a subject
US20150073266A1 (en) * 2013-09-06 2015-03-12 Covidien Lp System and method for lung visualization using ultasound
CN106073703A (en) * 2016-05-27 2016-11-09 重庆大学 A kind of athlete's life sign monitor system based on LoRa technology
CN107871531A (en) * 2016-09-27 2018-04-03 柯惠有限合伙公司 Crack is assessed and the planning of surgical operation Interventional
CN108983187A (en) * 2018-07-11 2018-12-11 西安电子科技大学 Online radar target identification method based on EWC
CN109965858A (en) * 2019-03-28 2019-07-05 北京邮电大学 Based on ULTRA-WIDEBAND RADAR human body vital sign detection method and device
US20210030276A1 (en) * 2019-07-29 2021-02-04 DawnLight Technologies Inc. Remote Health Monitoring Systems and Method
US20210093203A1 (en) * 2019-09-30 2021-04-01 DawnLight Technologies Systems and methods of determining heart-rate and respiratory rate from a radar signal using machine learning methods
CN110584631A (en) * 2019-10-10 2019-12-20 重庆邮电大学 Static human heartbeat and respiration signal extraction method based on FMCW radar
CN112914545A (en) * 2019-12-05 2021-06-08 南京理工大学 Non-contact intestinal disease diagnostic equipment
US20210325509A1 (en) * 2020-04-20 2021-10-21 Infineon Technologies Ag Radar-Based Vital Sign Estimation
CN111481184A (en) * 2020-04-24 2020-08-04 华侨大学 Multi-target respiration heart rate monitoring method and system based on millimeter wave radar technology
CN111603138A (en) * 2020-05-19 2020-09-01 杭州电子科技大学 Sleep apnea monitoring system based on millimeter wave radar
CN111723527A (en) * 2020-06-28 2020-09-29 重庆大学 Gear residual life prediction method based on cocktail long-term and short-term memory neural network
CN112754431A (en) * 2020-12-31 2021-05-07 杭州电子科技大学 Respiration and heartbeat monitoring system based on millimeter wave radar and lightweight neural network
CN112754441A (en) * 2021-01-08 2021-05-07 杭州环木信息科技有限责任公司 Millimeter wave-based non-contact heartbeat detection method
CN112998668A (en) * 2021-02-06 2021-06-22 路晟悠拜(重庆)科技有限公司 Millimeter wave-based non-contact far-field multi-human-body respiration heart rate monitoring method
CN113729641A (en) * 2021-10-12 2021-12-03 南京润楠医疗电子研究院有限公司 Non-contact sleep staging system based on conditional countermeasure network

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
NAISHADHAM, K., PIOU, J. E., REN, L., & FATHY, A. E.: "Estimation of Cardiopulmonary Parameters From Ultra Wideband Radar Measurements Using the State Space Method", 《IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS》, pages 1037 *
梁璐: "基于UWB雷达系统的人体呼吸心跳信号检测方法", 《中国优秀硕士学位论文全文数据库》, pages 006 - 135 *
罗朗娟, 张洽: "非接触式人体生命信号检测方法研究", 《医疗卫生装备》, pages 8 - 13 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113892931A (en) * 2021-10-14 2022-01-07 重庆大学 Method for extracting and analyzing intra-abdominal pressure by FMCW radar based on deep learning
CN113892931B (en) * 2021-10-14 2023-08-22 重庆大学 Method for extracting and analyzing intra-abdominal pressure by FMCW radar based on deep learning
CN115844355A (en) * 2022-11-22 2023-03-28 哈尔滨理工大学 UWB (ultra wide band) biological radar echo signal heart rate extraction method

Also Published As

Publication number Publication date
CN114246563B (en) 2023-11-17

Similar Documents

Publication Publication Date Title
CN114246563B (en) Heart and lung function intelligent monitoring equipment based on millimeter wave radar
CN110755076A (en) Image acquisition method, apparatus, device and medium for magnetic resonance scanner
CN112754431A (en) Respiration and heartbeat monitoring system based on millimeter wave radar and lightweight neural network
WO2010124034A2 (en) Processing physiological sensor data using a physiological model combined with a probabilistic processor
KR20140139564A (en) Systems and methods for ecg monitoring
WO2019216378A1 (en) Calculation device, detection device, calculation method, and computer program
WO2010100418A1 (en) Systems and methods for monitoring heart rate and blood pressure correlation
WO2020228420A1 (en) Denoising autoencoder training method, electrocardiography signal denoising method, and apparatuses
WO2018149147A1 (en) Method and apparatus for extracting respiration rate
CN103976728A (en) Method for improving electrocardiographic waveform drawing accuracy
WO2021034960A1 (en) Systems and methods for imputing real-time physiological signals
CN110060225A (en) A kind of Medical image fusion method based on rapid finite shearing wave conversion and rarefaction representation
CN115736854A (en) Respiration and heartbeat monitoring system based on millimeter wave radar
CN114711784A (en) Heart transplantation postoperation electrocardiogram abnormity analysis method and system
CN114916940A (en) Arrhythmia signal detection method, system and terminal based on expert knowledge
KR20180109903A (en) Method and apparatus for detecting living tissue using signal analysis
CN113033358A (en) Cuff-free blood pressure modeling method based on deep neural network
EP3499513A1 (en) Determining whether a hypothesis concerning a signal is true
CN115861172A (en) Wall motion estimation method and device based on self-adaptive regularized optical flow model
CN110604549B (en) Image acquisition method, apparatus, device and medium for magnetic resonance scanner
Bodile et al. Adaptive filtering of electrocardiogram signal using hybrid empirical mode decomposition-Jaya algorithm
Eleftheriadis et al. Fast and accurate power spectral analysis of heart rate variability using fast Gaussian gridding
CN109662689A (en) A kind of health early warning system of the hospital based on electrocardiogram
CN117349600B (en) Heart sound and heart electricity combined diagnosis method and system based on dual-mode dual input
Liao et al. AAFA-Net: Adaptive Amplitude-Frequency Attention Network for Fetal Monitoring from Noninvasive Abdominal Recordings

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