CN106344005A - Mobile ECG (electrocardiogram) monitoring system and monitoring method - Google Patents
Mobile ECG (electrocardiogram) monitoring system and monitoring method Download PDFInfo
- Publication number
- CN106344005A CN106344005A CN201610973366.6A CN201610973366A CN106344005A CN 106344005 A CN106344005 A CN 106344005A CN 201610973366 A CN201610973366 A CN 201610973366A CN 106344005 A CN106344005 A CN 106344005A
- Authority
- CN
- China
- Prior art keywords
- signal
- cnn
- uwb
- layer
- ecg
- 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
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—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
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
- A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
- A61B5/683—Means for maintaining contact with the body
- A61B5/6832—Means for maintaining contact with the body using adhesives
- A61B5/6833—Adhesive patches
-
- 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
- A61B5/7235—Details of waveform analysis
-
- 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
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
Abstract
The invention discloses a mobile ECG (electrocardiogram) monitoring system and monitoring method. ECG signals are collected; in addition, IR-UWB (impulse radio-ultra wide band) radar signals are collected; after the collected ECG signals and the IR-UWB radar signals are synchronized and processed, a cascade connection CNN (convolutional neural network) is used for feature extraction, integration analysis and diagnosis classification to obtain monitoring results to be output. According to the embodiment of the invention, the IR-UWB radar signals are introduced to be used as the supplement of the monitored ECG signals during the monitoring; through the synchronization on the two signals, the correlation between the two signals is obtained; the monitoring accuracy is improved. The cascade connection CNN with high applicability and stable performance is used for feature extraction, integration analysis and diagnosis classification; the system performance stability is effectively ensured.
Description
Technical field
The present invention relates to the monitoring technology of human body physiological parameter, particularly to a kind of removable electrocardiogram monitoring system and prison
Survey method.
Background technology
Heart disease seriously threatens the life and health of the mankind, and its M ＆ M all comes first of various diseases and the heart
The sudden of disease of ZANG-organs increased treatment difficulty.Therefore, preventing and treating heart disease just becomes the healthy cardinal task of people's maintenance.
Removable electrocardiogram (ecg) monitoring system is that effectively prevention heart disease provides good solution.On the one hand, it is conveniently taken
The feature of band allows this monitoring system to monitor the ecg signal of cardiomotility for a long time such that it is able to capture in time
Accidental arrhythmia and other sudden heart diseasies.On the other hand, with the fast development of mobile Internet of Things and information technology,
Such ecg monitoring system is obtained in that good platform and technical support, contributes to preferably meeting and carries out whenever and wherever possible
The demand of heart health.
At present, when realizing ecg monitoring system, can be realized based on internet of things system structure, based on Internet of Things net system knot
The ecg monitoring system of structure is representational three-decker, i.e. sensing layer, network layer and application layer.Wherein, sensing layer is mainly born
Duty electrocardiographicdata data acquisition, Internet is responsible for for the ECG data collecting being sent to data center, and application layer has been responsible for
Become analytical calculation and the problem early warning of electrocardiogram (ECG) data.When gathering ECG data, various wearable armarium can be adopted
Coupled together by wireless data communication technology in sensing layer, form a wireless sensor network.Network side can be by existing
The Internet or backbone communication network undertaking.Application layer is then to detect and analyze arrhythmia by applying various algorithms.
It is the Core Feature of whole ecg monitoring system clearly for ARR analyzing and diagnosing.
Carrying out ARR analyzing and diagnosing based on ecg signal is a kind of typical classification application, existing arrhythmia
Sorting technique mainly includes frequency analyses, the mode of support vector machine, wavelet transformation and expert's mixing.In short, these sides
Method all relies on greatly the feature extracting some keys from ecg signal waveform.But, these modes suffer from feature correlation
The restriction of threshold value, and the threshold value of different pieces of information collection is usually different, therefore, the performance of these methods not only can be subject to various making an uproar
The interference of sound and decline, also suffer from data set change impact and unstable.
In order to tackle the problems referred to above, more accurately will monitoring variable ecg Modulation recognition, begin to use have high fault-tolerant
Degree ability and the neutral net of robustness.For example, neutral net can be trained to as a heart beating identification detector, Yi Zhongxin
Restrain not normal grader or a feature extractor, for extracting, before classification, the human body physiological parameter not being manually set
Feature.Just because of neutral net, there is the automatic ability obtaining feature, can be disposed based on the ecg monitoring system of neutral net
In mobile terminal and high in the clouds, by periodically recording the new feature of learning from the high in the clouds ecg of specific user, so as to preferably
It is applied to specific user, and be favorably improved the accuracy of classification.Up to the present, occurred in that many with regard in the dynamic heart
In electrograph monitoring system, specific user is carried out with the research of arrhythmia classification.
Specifically, convolutional neural networks (cnn) are one of study hotspots of deep learning in recent years, have structure letter
The features such as single, training parameter is few and adapts to strong, is widely used in the fields such as pattern recognition and image procossing, achieves well
Effect.Ecg signal analysis are a kind of applications of image steganalysis, therefore also have researcher that cnn is applied to this, and
Illustrate that cnn is applicable for the ecg signal of a peacekeeping two dimension.Disclose one kind at present to lead convolutional neural networks
(lcnn), eight led what ecg signal recorded and be converted into a two-dimensional matrix.Wherein, the sliding scale of convolution kernel is filtered
The rule restriction that the ecg signal record that ripple device can not be led by difference is shared, this method not only demonstrates cnn in one-dimensional signal
In availability, but also adapt in different resolution, although this method can improve the accuracy of ecg classification, so
And the application of this method is limited, the multi-lead input to ecg signal requires,
While the sorting algorithm research of ecg signal constantly advances, another just improves the striving direction of systematic function
It is the data acquisition end in ecg signal.It is true that why ecg monitoring system can reduce cardiopathic mortality rate, at all former
Because being that it can record to long-term dynamics ecg signal.In this case, even some rare arrhythmia phenomenons
Also can be caught in such that it is able to timely treat, save sufferer.But, ecg signal record is inevitably subjected to people
The impact of the motion artifactses that body muscular movement produces, leads to sorting algorithm accuracy to decline.Although having been proposed for some motions
Artifact inhibition algorithm, such as adaptive filter method and blind source separation method, but remove motion artifactses completely and remain highly difficult
's.
To sum up, based on the inaccuracy when monitoring ecg signal and the follow-up inaccuracy in Klassifizierung cg signal, cause
Current ecg signal monitoring system is when obtaining monitoring result based on the ecg signal record monitored inaccurate, leads to people
The erroneous judgement of the arrhythmia phenomenon of body.
Content of the invention
In view of this, embodiments provide a kind of removable electrocardiogram monitoring system, this system enables to base
The monitoring result obtaining in the ecg signal record monitored is accurate.
The embodiment of the present invention also provides a kind of removable electrocardiogram monitoring method, and the method enables to based on being monitored
The monitoring result that ecg signal record obtains is accurate.
As can be seen from the above scheme, the present invention is implemented as:
A kind of removable electrocardiogram ecg monitoring system, comprising: acquisition module and cascade cnn module, wherein,
Acquisition module, for monitoring impulse radio ultra-wideband ir-uwb radar signal, synchronize and integrated process after
Obtain ir-uwb heartbeat signal；Monitoring ecg signal, obtains ecg signal data after synchronizing process and unbalanced data process；
Concatenated convolutional neutral net cnn module, for from the ir-uwb heartbeat signal that acquisition module receives and integrated from
Ecg signal data after reason carries out feature extraction, after integrated and diagnostic classification, obtains monitoring result.
A kind of ecg monitoring method, comprising:
Monitoring ir-uwb radar signal, synchronize and integrated process after obtain ir-uwb heartbeat signal；
Monitoring ecg signal, obtains ecg signal data after synchronizing process and unbalanced data process；
Carry out feature to the ecg signal data after the ir-uwb heartbeat signal receiving from acquisition module and integrated process to take out
Take, after integrated and diagnostic classification, obtain monitoring result.
From such scheme as can be seen that the embodiment of the present invention not only gathers ecg signal, and acquisition pulse radio ultra-wide
Band (ir-uwb) radar signal, after the ecg being gathered signal and ir-uwb radar signal are synchronized and process, using cascade
Cnn carry out feature extraction, integrated analysis and diagnostic classification, obtain monitoring result output.Because the embodiment of the present invention is in monitoring
When introduce ir-uwb radar signal, as the supplement of monitoring ecg signal, and by the synchronization to both signals, obtain this
Dependency between two kinds of signals, improves monitoring accuracy, because the embodiment of the present invention employs strong adaptability and stable performance
Cascade cnn carry out feature extraction, integrated analysis and diagnostic classification, stable system performance is effectively ensured.Therefore, using the present invention
The system and method that embodiment provides makes the monitoring result obtaining based on the ecg signal record monitored accurate.
Brief description
Fig. 1 is a kind of removable electrocardiogram monitoring system structure diagram provided in an embodiment of the present invention；
Fig. 2 is that the structure that synchronous submodule provided in an embodiment of the present invention carries out integrated process to ir-uwb radar signal is shown
It is intended to；
Fig. 3 is the structural representation of a cnn or the 2nd cnn provided in an embodiment of the present invention；
Fig. 4 is the structural representation of the 3rd cnn provided in an embodiment of the present invention；
Fig. 5 is a kind of removable electrocardiogram monitoring method flow diagram provided in an embodiment of the present invention；
Fig. 6 is the specific example structure chart of a cnn provided in an embodiment of the present invention；
Fig. 7 is the specific example structure chart of the 2nd cnn provided in an embodiment of the present invention；
Fig. 8 is the specific example structure chart of the 3rd cnn provided in an embodiment of the present invention.
Specific embodiment
For making the objects, technical solutions and advantages of the present invention become more apparent, develop simultaneously embodiment referring to the drawings, right
The present invention is described in further detail.
The embodiment of the present invention in order to accurately monitor arrhythmia under kinestate, by ir-uwb radar signal
Monitoring technology be incorporated into the present invention offer ecg monitoring system in, as auxiliary monitoring instrument.Ir-uwb radar not only provides
A kind of measurement of contactless ir-uwb radar signal, it also avoid electrical interference simultaneously.The clinical practice bag of ir-uwb radar
Include the measurement to heart rate, breathing rate and blood pressure etc..Using the embodiment of the present invention, compared with background technology, mainly from two angles
Carry out the robustness to arrhythmia monitoring function under kinestate for the strengthening system.First, introduce ir-uwb radar signal, as
Effective supplement to ecg signal record, and, by synchronizing and integrated analysis to both signals, can also excavate
Potentially relevant property between both data, to provide more strong for diagnostic classification ARR in dynamic monitoring system
Information foundation, improve monitoring accuracy.Second, using the cascade cnn of strong adaptability and stable performance carry out feature extraction,
Integrated analysis and diagnostic classification, help to maintain stable system performance.
Fig. 1 be ecg monitoring system structural representation provided in an embodiment of the present invention, as illustrated, include acquisition module and
Cascade cnn module, wherein,
Acquisition module, for monitoring ir-uwb radar signal, synchronize and integrated process after obtain ir-uwb heart beating letter
Number；Monitoring ecg signal, obtains ecg signal data after synchronizing process and unbalanced data process；
Cascade cnn module, for the ecg signal after the ir-uwb heartbeat signal receiving from acquisition module and integrated process
Data carries out feature extraction, after integrated and diagnostic classification, obtains monitoring result.
In the structure shown here, described acquisition module also includes ecg signals collecting submodule, for monitoring ecg signal.Ecg believes
Number carry the biological electrical feature of heart beating, such as ventricular depolarization and the interval polarized again.Ecg signals collecting submodule can be based on
Ecg sensor chip bmd101 and supporting bluetooth submodule are built.Although bmd101 chip can only obtain ecg signal one leads
Connection, but the diagnostic classification for subsequently normal/abnormal heart beating is enough.
In the structure shown here, described acquisition module also includes ir-uwb radar signal collection submodule, for monitoring ir-uwb
Radar signal.Ir-uwb radar collection submodule can be constructed using nva-r661 radar module, due to the transmitting work(of radar
Rate is also lower than 1/10th of wireless network transmission power, so injury will not be brought to human body.
Although ecg signal and the ir-uwb radar signal from radar have different features, they can be total to
With description heart beating parameter, can complement each other, be to have very much in the case that particularly a kind of data is heavily disturbed wherein
Meaning.Cascade cnn resume module follow-up for convenience, has also carried out synchronization and unbalanced data is processed.Generally, ecg adopts
The sample rate of collection submodule is 512 hertz (hz), and when radar covering radius is set to 2 meters, ir-uwb radar signal gathers
The sample rate of submodule only has 38hz.Due to the sample rate of both signals different so that the ecg signal after processing and ir-uwb
The length of heartbeat signal is also different, carries out integrated analysis in cascade cnn module, during extraction feature, will be respectively both signals
Different cnn are set.
Because ecg signal and ir-uwb radar signal are to be acquired by different acquisition channels, when collection is initial
Between and transmission time may all can be different.Even so, even if bmd101 chip and nva-r661 radar module open simultaneously
Dynamic, before their life's works, each required time is also different.Therefore, in order to subsequently more accurately be collected
Become analysis it is necessary to synchronize to both signals, that is, time calibration.In embodiments of the present invention, when have selected reception
Between stamp be used as the standard of time calibration.
In the structure shown here, acquisition module also includes integrated process submodule, for entering to the ir-uwb radar signal monitored
The integrated process of row.
Fig. 2 is that the structure that synchronous submodule provided in an embodiment of the present invention carries out integrated process to ir-uwb radar signal is shown
It is intended to, comprising: go direct current component, band-pass filter unit, principal component analysiss (pca) to remove noise signal unit, extract mainly
Signal element and integrated empirical mode decomposition unit, wherein,
Remove direct current component, for removing the DC component in ir-uwb radar signal；
Band-pass filter unit, for carrying out bandpass filtering by ir-uwb radar signal；
Pca removes noise signal unit, for removing the clutter in ir-uwb radar signal；
Extract main signal unit, for extracting the ir-uwb heartbeat signal in ir-uwb radar signal；
Integrated empirical mode decomposition unit, for the ir-uwb heartbeat signal rule of thumb Mode Decomposition obtaining extraction,
Remove breath signal therein, finally give the ir-uwb heartbeat signal of needs.
Because ir-uwb radar has high spatial resolution and strong penetrance, this makes it be highly suitable at monitoring
Thoracic cavity under kinestate rises and falls.Through monitoring in the ir-uwb radar signal obtaining, except the ir-uwb heart beating letter needing
Extra, wherein also mixes DC component, noise signal and breath signal etc., needs to apply following specific signal processing methods
Therefrom obtain ir-uwb heartbeat signal.
In the ir-uwb radar signal matrix that monitoring obtains, the τ when time of every a line is referred to as fast, the time of every string
T when being slow.Every a line in matrix is the sampling of signal waveform, and every string then represents the different positions apart from radar for the echo
Put.Monitor the ir-uwb radar signal that obtains and contain DC component, be static constant, will not with slow when the change of t and change
Become.Therefore, it can, by asking row average the often capable of signal matrix, then deduct row averagely with the often row of primary signal, thus
Remove the DC component in ir-uwb radar signal matrix.
Can be by the frequency range receiving signal be controlled to arranging of parameter in nva-r661 radar module.Known
In the case of parameter setting, it is known that frequency range, then the embodiment of the present invention can using a band-pass filter unit
To obtain the target ir-uwb radar signal of needs further.
Pca mode is commonly used for extracting the principal character composition of data.And noise signal is because static background environment is anti-
Penetrate what radar signal caused, so noise signal, in the echo of same position, changes very in the same row of radar matrix
Little.Therefore, the energy of pca noise signal is typically greater than the energy of the echo-signal caused by the fluctuating of thoracic cavity.Then, by strange
Different value is decomposed (svd) formula and signal matrix is decomposed:
R=usvt(formula 1)
Wherein, s is to be made up of the non-negative singular value of r, the element in s, except the element on diagonal is all 0, diagonally
Line element is then the non-negative singular value according to descending, can be expressed as:
s1,1＞ s2,2＞ ... ＞ sN, n
These values have corresponded to the size of energy, and therefore, it is several that the energy of ir-uwb radar signal is mainly distributed on above
In singular value.For a specific energy ratio, the quantity of corresponding singular value also determines that.If we are these singular values
It is set to 0, then the energy of corresponding proportion is just removed.Then, rebuild after svd, most of energy of noise signal
Just it is removed, the ir-uwb radar signal matrix of reconstruct will have higher signal to noise ratio.
In ir-uwb radar signal matrix at this moment, all row correspond to the energy of different rows position.Because environment is
Static, fluctuating thoracic cavity that string corresponding just has the variance of maximum.Thus, the embodiment of the present invention calculates ir-uwb radar letter
The variance of each row of number matrix, finds that maximum string of variance, has just obtained ir-uwb heartbeat signal.
After determining ir-uwb heartbeat signal column, need the mixing from breath signal and ir-uwb heartbeat signal
In signal, ir-uwb heartbeat signal is separated.Empirical mode decomposition (emd) is according to signal self-characteristic, the ir- that will determine
Uwb heartbeat signal is decomposed into a series of intrinsic mode functions (imf), and each imf embodies different frequency yardstick in signal
Concussion characteristic.Integrated empirical mode decomposition (eemd) is the improved method based on emd.Eemd is in the ir-uwb heartbeat signal determining
Add random white noise, then repeatedly use emd process signal, average as final to all decomposing the imf obtaining
Imf, that is, the basic function of the ir-uwb heartbeat signal determining.The imf obtaining component is arranged from high frequency to low frequency order, sequence is low
Imf component represent and shake faster pattern, the high imf component that sorts represents slower concussion pattern.To each imf, count
Calculate its gross energy (etotal) etotal of whole frequency range and in heart beating frequency range ([0.8hz, 2.5hz]) energy
(eheart).According to the energy ratio of eheart and etotal, select to have the imf of ceiling capacity ratio to reconstruct ir-uwb heart beating letter
Number.
In the structure shown here, acquisition module also includes unbalanced data process submodule, is additionally operable to carry out injustice to ecg signal
After weighing apparatus data processing, increase the abnormal heartbeats number in ecg signal.Here, unbalanced data uses after processing and ecg is believed
Number carry out over-sampling mode and add white noise mode.
Under normal conditions, normal heartbeat often quantitatively substantially exceeds abnormal heartbeats, has thereby resulted in ecg signal
Imbalance, and further cause big class accuracy also above group accuracy.In order to process this unbalanced signal collection,
Through proposing two methods for suppressing the impact to overall accuracy of classifying for the big class.A kind of method is to increase in signal collection side
Group or minimizing big class, another kind of method is then to add weights to grader.In the case of using cnn, over-sampling mode ratio adds
Power classifier methods perform better than.Therefore, the embodiment of the present invention employs over-sampling mode.For the periodic signal of ecg signal,
A kind of over-sampling mode is to increase number of samples by moving the starting point of sampling in one cycle.Additionally, in ecg signal
Middle interpolation noise can also improve cnn classification accuracy.These newly-increased noises, similar to baseline drift and Hz noise, meet
Can removed principle.The embodiment of the present invention also uses both approaches to increase the sample number of abnormal heartbeats.
In the structure shown here, cascade cnn module is responsible for the ecg signal data after the integrated process of integrated analysis and ir-uwb heart beating
Signal, so that diagnosis normal heartbeat/abnormal heartbeats.The structure of cascade cnn module mainly has two-layer, and ground floor includes two differences
Cnn, i.e. a cnn, for ir-uwb heartbeat signal is carried out with feature extraction, extract feature.And the 2nd cnn, for right
Ecg signal data after integrated process carries out feature extraction.The embodiment of the present invention adopts this set, as it was previously stated, be due to
The ecg signal data obtaining and the ir-uwb heartbeat data obtaining from radar have different sample rate data structures.Special
Not, feature here is not artificially pre-designed, such as wave amplitude, area, average etc., but is excavated by cnn oneself
Data in the further feature hidden, also exactly this feature is so that cnn is best in the performance of image processing field at present
's.It is all one-dimensional for being additionally, since ecg signal data and ir-uwb heartbeat signal, and a cnn and the 2nd cnn also accordingly divides
Do not carry out feature extraction using using one-dimensional convolution kernel.The second layer of cascade cnn is the 3rd cnn, and it is responsible for a cnn
And the 2nd the feature that extracts of cnn carry out mark of classifying, label has normal heartbeat and two kinds of abnormal heartbeats.Can be instructed by study
Practice and to lift the accuracy of the 3rd cnn classification.After training, you can real time execution, Dynamic Display goes out monitoring result.?
One layer and the second layer between, the embodiment of the present invention arranges the integrated submodule of feature, for taking out a cnn and the 2nd cnn
The two kinds of features taken out carry out integrated, and core here is Integrated Strategy.A two dimensional character is features defines after integrated
Collection, is input to the 3rd cnn, the convolution kernel of the therefore the 3rd cnn is two-dimentional.The classification results of the 3rd cnn may also pass through assessment
Feed back to the integrated submodule of feature, in order to adjust Integrated Strategy, optimize systematic function.
Fig. 3 is the structural representation of a cnn or the 2nd cnn provided in an embodiment of the present invention, due to a cnn and second
The structure of cnn is identical, so directly explanation, comprising: signals layer, convolutional layer 1, pond layer 1, convolutional layer 2, pond layer 2, convolutional layer
3 and pond layer 3, process is all one-dimensional case.This structure has 3 convolutional layers, a then pond after each convolutional layer
Layer.Input data is all one-dimensional, and all of convolution kernel is also all one-dimensional.During propagated forward, often one-dimensional convolutional layer
Computing formula as follows:
Wherein,It is that the kth of l layer opens characteristic pattern,It is the biasing that in l layer, kth opens characteristic pattern,It is l-1
I-th characteristic pattern of layer,It is the convolution kernel mapping to k-th characteristic pattern of l layer from all features of l-1 layer,
N is the total characteristic number in l-1 layer.Symbol conv represents Vector convolution, and f () represents excitation function.
Can have multiple convolution kernels, each convolution kernel corresponds to an one-dimensional filter, in order to extract a kind of without
The feature of design, generates a characteristic pattern.In fact, due to ir-uwb heartbeat signal scale unlike ecg signal data so
Greatly, the neuron number in each convolutional layer of a cnn and the 2nd cnn, convolution kernel size can differ, but they
Output characteristic will can jointly be input in the second layer, so, they need to export the one-dimensional characteristic of identical scale.
Fig. 4 is the structural representation of the 3rd cnn provided in an embodiment of the present invention, including input layer, convolutional layer, pond layer,
Hidden layer and output layer, specifically, a cnn and the 2nd cnn is by the feature extracting output and integrated, forms the 3rd cnn's
Two-dimensional matrix is input to input layer, after a convolutional layer having two-dimensional convolution core, connects a pond layer, then through complete
Connect and enter a hidden layer, be finally the result that output layer provides classification.In fact, two-layer one below is classical
Multi-layer perception (MLP) (mlp).Here, carry out integrated study two by using two-dimensional convolution core to be given birth to by a cnn and the 2nd cnn
Potentially relevant property between the one-dimensional characteristic becoming, contributes to the biomedical record of multi-platform channel more accurately to assign to difference
In classification.Even if sometimes ecg signal data record or ir-uwb heartbeat signal have one kind destroyed, final output is tied
Fruit still may rely on both one-dimensional characteristics and the potentially relevant property between them, thus keeping preferable stability.
Especially, Chinese cardiovascular disease data base (ccdd) provides some available ecg signal resources, and the present invention is real
Apply the pre-training that example can carry out having supervision to the 2nd cnn for ecg signal processing.The embodiment of the present invention employs 150,000
Ecg signal record in ccdd, the time of each record, from 10 seconds to 20 seconds, comprises 12 signals leading.Through
After data nonbalance is processed, normal heartbeat and abnormal heartbeats number roughly equal it is possible to use as training dataset.
In the structure shown here, the extraction feature of a cnn and the 2nd cnn, before being input to the 3rd cnn, first passes through feature set
Become, so that the 3rd cnn can not only be using ir-uwb heartbeat signal and the respective feature of ecg signal data additionally it is possible to excavate
Therebetween potentially relevant property.In embodiments of the present invention, this dependency exactly improves holistic diagnosis classification accuracy
Critical support, is also that the embodiment of the present invention introduces ir-uwb heartbeat signal to supplement the main purpose of ecg signal data shortcoming.When
So, cnn itself has had preferable integrated analysis ability, if the embodiment of the present invention can improve its input signal further
Integrated advantage, then overall performance will get a promotion further.Therefore, increase the integrated submodule of feature purport should be as
What by ir-uwb heartbeat signal and appropriate the integrating of ecg signal data, can preferably embody the complementarity of the two and
Dependency.
Increase the integrated submodule of feature and mainly consider both sides factor in cnn module.First, ir-uwb heartbeat signal
Processed and feature extraction in each independent passage with ecg signal data, obtained result value scope may phase
Difference ratio is larger, and whether this gap can produce considerable influence to result？The need of being standardized？Second, how accurately to sentence
Complementary presence being used between disconnected two kinds of signals？If ir-uwb heartbeat signal and ecg signal data one of which
Impaired, now integrated result is inevitable takes into account potentially relevant property between the two based on int one kind, this exactly present invention
Embodiment expects the situation that performance can get a promotion.But if two kinds of signals are all impaired, how this is processed？If two kinds of signals
All fine, but assume negative correlation, and how this is processed？
Based on above-mentioned analysis, the integrated submodule of feature that the embodiment of the present invention proposes is integrated using weighting, is expressed as follows:
Y=[a (1- θ1)x1, b (1- θ2)x2]
Wherein, y represents integrated 2D signal, x1Represent ir-uwb heartbeat signal, x2Represent ecg signal, a represents heart beating
The normalization factor of signal, b represents the normalization factor of ecg signal, θ1∈ [0,1] represents the impaired system of ir-uwb heartbeat signal
Number, θ2∈ [0,1] represents the impaired coefficient of ecg signal, therefore, (1- θ1) represent the intact degree of ir-uwb heartbeat signal, (1-
θ2) represent the intact degree of ecg signal.It can be seen that, the size between two kinds of signal dimension can be controlled to put down by normalization factor
Weighing apparatus, can adjust a kind of contribution proportion in integrated signal for signal by impaired coefficient.General, normalization factor can root
Set according to truthful data situation, impaired coefficient can set according to the working condition of data acquisition platform.Obtaining performance evaluation
It is also possible to adjust above-mentioned parameter after feedback.
Fig. 5 is a kind of removable electrocardiogram monitoring method flow diagram provided in an embodiment of the present invention, and it concretely comprises the following steps:
Step 501, monitoring ir-uwb radar signal, synchronize and integrated process after obtain ir-uwb heartbeat signal；
Step 502, monitoring ecg signal, obtain ecg signal data after synchronizing process and unbalanced data process；
Step 503, the ecg signal data after the ir-uwb heartbeat signal receiving from acquisition module and integrated process is carried out
After feature extraction, integrated and diagnostic classification, obtain monitoring result.
Lift a specific example embodiment of the present invention is described in detail.
Ir-uwb radar signal and ecg signal are respectively by radar module nva-r661 and ecg module bmd101 sensor core
Piece gathers.Subject person is sitting in front of ir-uwb radar, wrist is sticked electrode slice and is connected with ecg sensor.Have 20 volunteers
Take part in data acquisition.Process unbalanced data using oversampler method.
Cnn module have input 10 seconds long ir-uwb heartbeat signals, has 360 sampled points.Ecg signal is in input cnn mould
Before block, the resampling of frequency 512hz is 200hz, every 1200 sampled points are designated as one section.Normalization factor and be both configured to 1.
Signals collecting working platform in order, impaired coefficient and be both configured to 0.Experiment porch for training cnn is deep learning
Framework caffe.In experiment, as shown in Fig. 6, Fig. 7 and Fig. 8, Fig. 6 is provided in an embodiment of the present invention to specifically used network parameter
The specific example structure chart of the first cnn, Fig. 7 is the specific example structure chart of the 2nd cnn provided in an embodiment of the present invention, and Fig. 8
Specific example structure chart for the 3rd cnn provided in an embodiment of the present invention.
It can be seen that three convolution kernels that a cnn of the feature for extracting ir-uwb heartbeat signal uses
Size is respectively 1 × 61,1 × 31,1 × 21, and pond layer is all 1 × 2 to carry out average pond.For extracting ecg signal data
2nd cnn of feature employs three convolution kernels, and size is respectively 1 × 201,1 × 141,1 × 141, and each pond layer is adopted
With 1 × 2 average pond method.The characteristic pattern that final both output is 20 1 × 20, is integrated into 2 × 20 feature, input the
Three cnn.The convolution kernel size that 3rd cnn uses is 2 × 5, and pond layer is also 1 × 2 to carry out average pond.Two layer multi-layer senses
Know that the hidden layer of machine has 50 neurons, output layer has 1 neuron.
In order to verify the accuracy for two classification (normal heartbeat, abnormal heartbeats) of this example setting, employ lcnn
Carry out Performance comparision.Table 1 shows the result in different exception records and normal recordings ratio.This example uses following 3
Individual parameter is weighing classification performance.
● accuracy (acc): the ecg record correctly classified and the ratio of radar record
● specificity (sp): the ratio that normal ecg record is correctly classified with normal radar record
● sensitivity (se): the ratio that abnormal ecg record is correctly classified with abnormal radar record
The comparative experimentss of table 1 lcnn and cascade cnn
Table 1 shows, although the specificity of lcnn is more better than cascade cnn, lcnn sensitivity is 0 it is impossible to identification is abnormal
Ecg records, and this also demonstrates the shortcoming of lcnn it is simply that it is impossible to keep in new data set in the case of given training dataset
Stable performance.The sensitivity of cascade cnn is performed well.Lcnn and cascade cnn accuracy be all as exception record with just
The often reduction of ratio of record and improve, but the accuracy cascading cnn is always higher, and fluctuation range is less.It can be seen that, even if number
Yet suffer from disequilibrium according to collection, cascade cnn remains able to show high classification accuracy.
In second experiment, this example compares lcnn and cascade classification performance under kinestate for the cnn.In data
In gatherer process, subject person is allowed to speak or weak vibrations body.Result is as shown in the table.
Table 2 lcnn is compared with cascade classification performance under kinestate for the cnn
It can be seen that, cascade cnn accuracy fluctuation within a narrow range near peak always, and with table 1 in static properties
Quite it was demonstrated that its performance under light exercise state is stable.On the contrary, the accuracy of lcnn but maintain always relatively low
Level, and under having by a relatively large margin compared with the static properties in table 1, show that lcnn is processing static and kinestate combination
Under data when, performance be decline.This result again demonstrates, and cascade cnn uses ir-uwb heartbeat signal and ecg signal number
Arrhythmia classification effect according to integrated analysis is satisfactory, and particularly under some kinestates, its classification accuracy is still
Good level can be reached.
To sum up, the embodiment of the present invention is by being extracted with integrated from ecg signal data with the ir-uwb heart using cascade cnn
Jump the feature of signal, more fully to be analyzed.Also just because of this integrated analysis so that the embodiment of the present invention provides
System can show stable performance in the state of having light exercise.By contrast experiment it was demonstrated that enforcement of the present invention
Example can reach higher level in the accuracy of monitoring, even in the case of light exercise, also has good stability table
Existing.
The object, technical solutions and advantages of the present invention are further described, institute by above act preferred embodiment
It should be understood that the foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all the present invention's
Spirit and principle within, any modification, equivalent and improvement of being made etc., should be included in protection scope of the present invention it
Interior.
Claims (10)
1. a kind of removable electrocardiogram ecg monitoring system is it is characterised in that include: acquisition module and cascade cnn module, wherein,
Acquisition module, for monitoring impulse radio ultra-wideband ir-uwb radar signal, synchronize and integrated process after obtain
Ir-uwb heartbeat signal；Monitoring ecg signal, obtains ecg signal data after synchronizing process and unbalanced data process；
Concatenated convolutional neutral net cnn module, after to the ir-uwb heartbeat signal receiving from acquisition module and integrated process
Ecg signal data carry out feature extraction, after integrated and diagnostic classification, obtain monitoring result.
2. monitoring system as claimed in claim 1 is it is characterised in that described acquisition module also includes ecg signals collecting submodule
Block, for monitoring ecg signal；Described acquisition module also includes ir-uwb radar signal collection submodule, for monitoring ir-uwb
Radar signal.
3. monitoring system as claimed in claim 1, it is characterised in that described acquisition module, is additionally operable to using reception time school
Quasi- mode synchronizes process.
4. monitoring system as claimed in claim 1 is it is characterised in that described acquisition module is believed in integrated process ir-uwb radar
Number when, also include direct current component, band-pass filter unit, principal component analysiss pca remove noise signal unit, extract and main believe
Number unit and integrated empirical mode decomposition unit, wherein,
Remove direct current component, for removing the DC component in ir-uwb radar signal；
Band-pass filter unit, for carrying out bandpass filtering by ir-uwb radar signal；
Pca removes noise signal unit, for removing the clutter in ir-uwb radar signal；
Extract main signal unit, for extracting the ir-uwb heartbeat signal in ir-uwb radar signal；
Integrated empirical mode decomposition unit, for the ir-uwb heartbeat signal rule of thumb Mode Decomposition obtaining extraction, removes
Breath signal therein, finally gives the ir-uwb heartbeat signal of needs.
5. monitoring system as claimed in claim 1 processes son it is characterised in that described acquisition module also includes unbalanced data
Module, is additionally operable to ecg signal is carried out after unbalanced data process using over-sampling mode and interpolation white noise mode, increases
Abnormal heartbeats number in ecg signal, obtains ecg signal data.
6. monitoring system as claimed in claim 1 is it is characterised in that described cascade cnn module includes two-layer:
Ground floor includes two different cnn, a cnn, for ir-uwb heartbeat signal is carried out with feature extraction, extracts special
Levy；And the 2nd cnn, for feature extraction is carried out to the ecg signal data after integrated process；
The second layer is the 3rd cnn, and the feature for extracting to a cnn and the 2nd cnn carries out mark of classifying, and label has normally
Heart beating and two kinds of abnormal heartbeats；
Between ground floor and the second layer include the integrated submodule of feature, for a cnn and the 2nd cnn is extracted two
Kind of feature carry out integrated after, be sent to the 3rd cnn.
7. monitoring system as claimed in claim 6 is it is characterised in that a described cnn or the 2nd cnn includes: 3 convolution
Layer, then 1 pond layer after each convolutional layer, 3 convolutional layers are linked in sequence, and input data is all one-dimensional, convolutional layer
Convolution kernel is one-dimensional, during propagated forward, the often computing formula of one-dimensional convolutional layer:
Wherein,It is that the kth of l layer opens characteristic pattern,It is the biasing that in l layer, kth opens characteristic pattern,It is l-1 layer
I-th characteristic pattern,It is the convolution kernel mapping to k-th characteristic pattern of l layer from all features of l-1 layer, n is
Total characteristic number in l-1 layer.Symbol conv represents Vector convolution, and f () represents excitation function.
8. monitoring system as claimed in claim 6 is it is characterised in that described 3rd cnn, including input layer, convolutional layer, Chi Hua
Layer, hidden layer and output layer, wherein,
First cnn and the 2nd cnn is by the feature extracting output and integrated, and the two-dimensional matrix forming the 3rd cnn is input to input layer,
After a convolutional layer having two-dimensional convolution core, connect a pond layer, then connect one hidden layer of entrance through complete,
After be that output layer provides monitoring result.
9. monitoring system as claimed in claim 6 is it is characterised in that the integrated submodule of described feature is using adopting weight-sets
Become, be expressed as follows:
Y=[a (1- θ1)x1, b (1- θ2)x2]
Wherein, y represents integrated 2D signal, x1Represent ir-uwb heartbeat signal, x2Represent ecg signal, a represents the ir-uwb heart
Jump the normalization factor of signal, b represents the normalization factor of ecg signal, θ1∈ [0,1] represents the impaired of ir-uwb heartbeat signal
Coefficient, θ2∈ [0,1] represents the impaired coefficient of ecg signal.
10. a kind of ecg monitoring method is it is characterised in that include:
Monitoring ir-uwb radar signal, synchronize and integrated process after obtain ir-uwb heartbeat signal；
Monitoring ecg signal, obtains ecg signal data after synchronizing process and unbalanced data process；
Feature extraction, collection are carried out to the ecg signal data after the ir-uwb heartbeat signal receiving from acquisition module and integrated process
After one-tenth and diagnostic classification, obtain monitoring result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610973366.6A CN106344005B (en) | 2016-10-28 | 2016-10-28 | A kind of removable electrocardiogram monitoring system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610973366.6A CN106344005B (en) | 2016-10-28 | 2016-10-28 | A kind of removable electrocardiogram monitoring system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106344005A true CN106344005A (en) | 2017-01-25 |
| CN106344005B CN106344005B (en) | 2019-04-05 |
Family
ID=57864972
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610973366.6A Active CN106344005B (en) | 2016-10-28 | 2016-10-28 | A kind of removable electrocardiogram monitoring system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106344005B (en) |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107184198A (en) * | 2017-06-01 | 2017-09-22 | 广州城市职业学院 | A kind of electrocardiosignal classifying identification method |
| CN107510452A (en) * | 2017-09-30 | 2017-12-26 | 扬美慧普（北京）科技有限公司 | A kind of ECG detecting method based on multiple dimensioned deep learning neutral net |
| CN107529645A (en) * | 2017-06-29 | 2018-01-02 | 重庆邮电大学 | A kind of heart sound intelligent diagnosis system and method based on deep learning |
| CN107811649A (en) * | 2017-12-13 | 2018-03-20 | 四川大学 | A kind of more sorting techniques of heart sound based on depth convolutional neural networks |
| CN108030488A (en) * | 2017-11-30 | 2018-05-15 | 北京医拍智能科技有限公司 | The detecting system of arrhythmia cordis based on convolutional neural networks |
| CN108172291A (en) * | 2017-05-04 | 2018-06-15 | 深圳硅基仿生科技有限公司 | Diabetic retinopathy identifying system based on eye fundus image |
| CN108256452A (en) * | 2018-01-06 | 2018-07-06 | 天津大学 | A kind of method of the ECG signal classification of feature based fusion |
| CN108470158A (en) * | 2018-03-08 | 2018-08-31 | 华南理工大学 | A method of it finding error minimal network for dynamic ECG data and calculates structure |
| CN108577832A (en) * | 2018-02-11 | 2018-09-28 | 东南大学 | A kind of atrial fibrillation method of discrimination |
| CN108618773A (en) * | 2017-03-15 | 2018-10-09 | 深圳市理邦精密仪器股份有限公司 | A kind of Denoising of ECG Signal, device and a kind of ecg signal acquiring equipment |
| CN108771530A (en) * | 2017-05-04 | 2018-11-09 | 深圳硅基仿生科技有限公司 | Eyeground pathological changes screening system based on deep neural network |
| CN108814585A (en) * | 2018-05-03 | 2018-11-16 | 深圳竹信科技有限公司 | ECG's data compression method, apparatus and computer readable storage medium |
| CN108814580A (en) * | 2018-04-18 | 2018-11-16 | 深圳市康年颐华智能科技有限公司 | A kind of method and relevant device of non-contact type heart rate test |
| CN108919218A (en) * | 2018-06-07 | 2018-11-30 | 北京邮电大学 | A kind of contactless number of people in car and the method and device of position judgement |
| CN109044339A (en) * | 2018-08-13 | 2018-12-21 | 重庆工商大学 | A kind of ECG signal diagnostic method of limit convolutional neural networks |
| CN109829408A (en) * | 2019-01-23 | 2019-05-31 | 中国科学技术大学 | Intelligent lightening recognition device based on convolutional neural networks |
| CN109840511A (en) * | 2018-12-29 | 2019-06-04 | 深兰科技（上海）有限公司 | A kind of Modulation recognition method, terminal and storage medium |
| CN109875579A (en) * | 2019-02-28 | 2019-06-14 | 京东方科技集团股份有限公司 | Emotional health management system and emotional health management method |
| 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 |
| CN110070013A (en) * | 2019-04-11 | 2019-07-30 | 杭州电子科技大学 | A kind of ECG signal noise pollution degree and classification intelligent evaluation method |
| CN110236520A (en) * | 2019-05-20 | 2019-09-17 | 上海数创医疗科技有限公司 | ECG type recognition methods and device based on double convolutional neural networks |
| WO2020061987A1 (en) * | 2018-09-28 | 2020-04-02 | 深圳大学 | Multi-lead combined electrocardiogram classification method |
| CN111307881A (en) * | 2020-03-10 | 2020-06-19 | 南京工业大学 | Gas sensor baseline drift compensation method for VOC detection |
| CN111356401A (en) * | 2017-11-29 | 2020-06-30 | 威里利生命科学有限责任公司 | Continuous detection and monitoring of cardiac arrhythmias using wearable sensors and cloud residency analysis |
| CN111366184A (en) * | 2020-04-17 | 2020-07-03 | 中铁隧道局集团有限公司 | Shield tunneling machine multi-sensor performance online monitoring method |
| CN111789574A (en) * | 2020-07-02 | 2020-10-20 | 山东省人工智能研究院 | ECG signal quality evaluation method |
| CN112130118A (en) * | 2020-08-19 | 2020-12-25 | 复旦大学无锡研究院 | SNN-based ultra-wideband radar signal processing system and processing method |
| CN112232256A (en) * | 2020-10-26 | 2021-01-15 | 南京读动信息科技有限公司 | Non-contact motion and body measurement data acquisition system |
| CN112971771A (en) * | 2021-02-23 | 2021-06-18 | 浙江大学计算机创新技术研究院 | Method for reconstructing electrocardiogram based on millimeter waves |
| EP3926361A3 (en) * | 2020-06-18 | 2022-03-30 | Infineon Technologies AG | Parametric cnn for radar processing |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030229289A1 (en) * | 2002-03-18 | 2003-12-11 | Mohler Sailor Hampton | Method and system for generating a likelihood of cardiovascular disease, analyzing cardiovascular sound signals remotely from the location of cardiovascular sound signal acquisition, and determining time and phase information from cardiovascular sound signals |
| US20100179447A1 (en) * | 2007-06-08 | 2010-07-15 | Anthony Charles Hunt | Methods of measurement of drug induced changes in cardiac ion channel function and associated apparatus |
| CN104970789A (en) * | 2014-04-04 | 2015-10-14 | 中国科学院苏州纳米技术与纳米仿生研究所 | Electrocardiogram classification method and system |
| CN105105743A (en) * | 2015-08-21 | 2015-12-02 | 山东省计算中心（国家超级计算济南中心） | Intelligent electrocardiogram diagnosis method based on deep neural network |
| CN105748063A (en) * | 2016-04-25 | 2016-07-13 | 山东大学齐鲁医院 | Intelligent arrhythmia diagnosis method based on multiple-lead and convolutional neural network |
| CN105943021A (en) * | 2016-05-13 | 2016-09-21 | 赵伟 | Wearable heart rhythm monitoring device and heart rhythm monitoring system |
-
2016
- 2016-10-28 CN CN201610973366.6A patent/CN106344005B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030229289A1 (en) * | 2002-03-18 | 2003-12-11 | Mohler Sailor Hampton | Method and system for generating a likelihood of cardiovascular disease, analyzing cardiovascular sound signals remotely from the location of cardiovascular sound signal acquisition, and determining time and phase information from cardiovascular sound signals |
| US20100179447A1 (en) * | 2007-06-08 | 2010-07-15 | Anthony Charles Hunt | Methods of measurement of drug induced changes in cardiac ion channel function and associated apparatus |
| CN104970789A (en) * | 2014-04-04 | 2015-10-14 | 中国科学院苏州纳米技术与纳米仿生研究所 | Electrocardiogram classification method and system |
| CN105105743A (en) * | 2015-08-21 | 2015-12-02 | 山东省计算中心（国家超级计算济南中心） | Intelligent electrocardiogram diagnosis method based on deep neural network |
| CN105748063A (en) * | 2016-04-25 | 2016-07-13 | 山东大学齐鲁医院 | Intelligent arrhythmia diagnosis method based on multiple-lead and convolutional neural network |
| CN105943021A (en) * | 2016-05-13 | 2016-09-21 | 赵伟 | Wearable heart rhythm monitoring device and heart rhythm monitoring system |
Non-Patent Citations (1)
| Title |
|---|
| 王利琴: "《心电信号波形检测与心律失常分类研究》", 《中国博士学位论文全文数据库》 * |
Cited By (49)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108618773A (en) * | 2017-03-15 | 2018-10-09 | 深圳市理邦精密仪器股份有限公司 | A kind of Denoising of ECG Signal, device and a kind of ecg signal acquiring equipment |
| CN111493814A (en) * | 2017-05-04 | 2020-08-07 | 深圳硅基智能科技有限公司 | Recognition system for fundus lesions |
| CN108172291A (en) * | 2017-05-04 | 2018-06-15 | 深圳硅基仿生科技有限公司 | Diabetic retinopathy identifying system based on eye fundus image |
| CN108172291B (en) * | 2017-05-04 | 2020-01-07 | 深圳硅基智能科技有限公司 | Diabetic retinopathy recognition system based on fundus images |
| CN108553079B (en) * | 2017-05-04 | 2020-06-02 | 深圳硅基智能科技有限公司 | Pathological change recognition system based on fundus image |
| CN111481166A (en) * | 2017-05-04 | 2020-08-04 | 深圳硅基智能科技有限公司 | Automatic identification system based on eye ground screening |
| CN111481166B (en) * | 2017-05-04 | 2021-11-26 | 深圳硅基智能科技有限公司 | Automatic identification system based on eye ground screening |
| CN111493814B (en) * | 2017-05-04 | 2021-11-12 | 深圳硅基智能科技有限公司 | Recognition system for fundus lesions |
| CN108771530A (en) * | 2017-05-04 | 2018-11-09 | 深圳硅基仿生科技有限公司 | Eyeground pathological changes screening system based on deep neural network |
| CN108553079A (en) * | 2017-05-04 | 2018-09-21 | 深圳硅基仿生科技有限公司 | Lesion identifying system based on eye fundus image |
| CN107184198A (en) * | 2017-06-01 | 2017-09-22 | 广州城市职业学院 | A kind of electrocardiosignal classifying identification method |
| CN107529645A (en) * | 2017-06-29 | 2018-01-02 | 重庆邮电大学 | A kind of heart sound intelligent diagnosis system and method based on deep learning |
| CN107529645B (en) * | 2017-06-29 | 2019-09-10 | 重庆邮电大学 | A kind of heart sound intelligent diagnosis system and method based on deep learning |
| CN107510452A (en) * | 2017-09-30 | 2017-12-26 | 扬美慧普（北京）科技有限公司 | A kind of ECG detecting method based on multiple dimensioned deep learning neutral net |
| CN107510452B (en) * | 2017-09-30 | 2019-10-08 | 扬美慧普（北京）科技有限公司 | A kind of ECG detecting method based on multiple dimensioned deep learning neural network |
| CN111356401B (en) * | 2017-11-29 | 2024-03-26 | 威里利生命科学有限责任公司 | Continuous detection and monitoring of cardiac arrhythmias using wearable sensors and cloud-resident analysis |
| CN111356401A (en) * | 2017-11-29 | 2020-06-30 | 威里利生命科学有限责任公司 | Continuous detection and monitoring of cardiac arrhythmias using wearable sensors and cloud residency analysis |
| CN108030488A (en) * | 2017-11-30 | 2018-05-15 | 北京医拍智能科技有限公司 | The detecting system of arrhythmia cordis based on convolutional neural networks |
| CN107811649B (en) * | 2017-12-13 | 2020-12-22 | 四川大学 | Heart sound multi-classification method based on deep convolutional neural network |
| CN107811649A (en) * | 2017-12-13 | 2018-03-20 | 四川大学 | A kind of more sorting techniques of heart sound based on depth convolutional neural networks |
| CN108256452A (en) * | 2018-01-06 | 2018-07-06 | 天津大学 | A kind of method of the ECG signal classification of feature based fusion |
| CN108577832A (en) * | 2018-02-11 | 2018-09-28 | 东南大学 | A kind of atrial fibrillation method of discrimination |
| CN108470158A (en) * | 2018-03-08 | 2018-08-31 | 华南理工大学 | A method of it finding error minimal network for dynamic ECG data and calculates structure |
| CN108470158B (en) * | 2018-03-08 | 2020-05-12 | 华南理工大学 | Method for searching error minimum network computing structure for dynamic ECG data |
| CN108814580A (en) * | 2018-04-18 | 2018-11-16 | 深圳市康年颐华智能科技有限公司 | A kind of method and relevant device of non-contact type heart rate test |
| CN108814585A (en) * | 2018-05-03 | 2018-11-16 | 深圳竹信科技有限公司 | ECG's data compression method, apparatus and computer readable storage medium |
| CN108814585B (en) * | 2018-05-03 | 2021-05-28 | 深圳竹信科技有限公司 | Electrocardiosignal processing method and device and computer readable storage medium |
| CN108919218A (en) * | 2018-06-07 | 2018-11-30 | 北京邮电大学 | A kind of contactless number of people in car and the method and device of position judgement |
| CN109044339A (en) * | 2018-08-13 | 2018-12-21 | 重庆工商大学 | A kind of ECG signal diagnostic method of limit convolutional neural networks |
| WO2020061987A1 (en) * | 2018-09-28 | 2020-04-02 | 深圳大学 | Multi-lead combined electrocardiogram classification method |
| CN109840511A (en) * | 2018-12-29 | 2019-06-04 | 深兰科技（上海）有限公司 | A kind of Modulation recognition method, terminal and storage medium |
| CN109829408A (en) * | 2019-01-23 | 2019-05-31 | 中国科学技术大学 | Intelligent lightening recognition device based on convolutional neural networks |
| CN109875579A (en) * | 2019-02-28 | 2019-06-14 | 京东方科技集团股份有限公司 | Emotional health management system and emotional health management method |
| 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 |
| CN109965858B (en) * | 2019-03-28 | 2020-10-30 | 北京邮电大学 | Ultra-wideband radar-based human body vital sign detection method and device |
| CN110070013B (en) * | 2019-04-11 | 2021-06-18 | 杭州电子科技大学 | Intelligent evaluation method for noise pollution degree and category of ECG signal |
| CN110070013A (en) * | 2019-04-11 | 2019-07-30 | 杭州电子科技大学 | A kind of ECG signal noise pollution degree and classification intelligent evaluation method |
| CN110236520B (en) * | 2019-05-20 | 2020-09-29 | 上海数创医疗科技有限公司 | Electrocardiogram type recognition device based on double convolution neural network |
| CN110236520A (en) * | 2019-05-20 | 2019-09-17 | 上海数创医疗科技有限公司 | ECG type recognition methods and device based on double convolutional neural networks |
| CN111307881A (en) * | 2020-03-10 | 2020-06-19 | 南京工业大学 | Gas sensor baseline drift compensation method for VOC detection |
| CN111307881B (en) * | 2020-03-10 | 2022-03-08 | 南京工业大学 | Gas sensor baseline drift compensation method for VOC detection |
| CN111366184A (en) * | 2020-04-17 | 2020-07-03 | 中铁隧道局集团有限公司 | Shield tunneling machine multi-sensor performance online monitoring method |
| EP3926361A3 (en) * | 2020-06-18 | 2022-03-30 | Infineon Technologies AG | Parametric cnn for radar processing |
| CN111789574A (en) * | 2020-07-02 | 2020-10-20 | 山东省人工智能研究院 | ECG signal quality evaluation method |
| CN112130118A (en) * | 2020-08-19 | 2020-12-25 | 复旦大学无锡研究院 | SNN-based ultra-wideband radar signal processing system and processing method |
| CN112130118B (en) * | 2020-08-19 | 2023-11-17 | 复旦大学无锡研究院 | Ultra-wideband radar signal processing system and method based on SNN |
| CN112232256A (en) * | 2020-10-26 | 2021-01-15 | 南京读动信息科技有限公司 | Non-contact motion and body measurement data acquisition system |
| CN112232256B (en) * | 2020-10-26 | 2024-02-02 | 南京读动信息科技有限公司 | Contactless sports and body measurement data acquisition system |
| CN112971771A (en) * | 2021-02-23 | 2021-06-18 | 浙江大学计算机创新技术研究院 | Method for reconstructing electrocardiogram based on millimeter waves |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106344005B (en) | 2019-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106344005B (en) | A kind of removable electrocardiogram monitoring system | |
| Alickovic et al. | Performance evaluation of empirical mode decomposition, discrete wavelet transform, and wavelet packed decomposition for automated epileptic seizure detection and prediction | |
| Yin et al. | ECG monitoring system integrated with IR-UWB radar based on CNN | |
| US11311201B2 (en) | Feature selection for cardiac arrhythmia classification and screening | |
| CN110420019B (en) | Deep regression heart rate estimation method for ballistocardiogram signals | |
| CN107822622A (en) | Electrocardiographic diagnosis method and system based on depth convolutional neural networks | |
| CN106980746A (en) | A kind of general Woundless blood sugar Forecasting Methodology based on Time-Series analysis | |
| US20150374300A1 (en) | Early detection of hemodynamic decompensation using taut-string transformation | |
| CN103034837B (en) | Characteristic parameter is associated with pulse condition key element | |
| CN106821376A (en) | A kind of epileptic attack early warning system and method based on deep learning algorithm | |
| CN105997064A (en) | Method for identifying human lower limb surface EMG signals (electromyographic signals) | |
| CN105125206B (en) | A kind of intelligent cardiac monitoring method and device | |
| CN105997043B (en) | A kind of pulse frequency extracting method based on wrist wearable device | |
| CN113397555A (en) | Arrhythmia classification algorithm of C-LSTM for physiological parameter monitoring | |
| CN111466876A (en) | Alzheimer's disease auxiliary diagnosis system based on fNIRS and graph neural network | |
| CN111887858B (en) | Ballistocardiogram signal heart rate estimation method based on cross-modal mapping | |
| CN109497997A (en) | Based on majority according to the seizure detection and early warning system of acquisition | |
| CN106889981B (en) | A kind of intelligent terminal for being used to extract fetal heart frequency | |
| Zhao et al. | An IoT-based wearable system using accelerometers and machine learning for fetal movement monitoring | |
| CN112806977B (en) | Physiological parameter measuring method based on multi-scale fusion network | |
| WO2022135449A1 (en) | Interictal epileptiform discharge activity detection apparatus and method for epileptic patient | |
| CN113208613B (en) | Multi-mode BCI (binary coded decimal) timing optimization method based on FHLS (FHLS) feature selection | |
| CN114052693A (en) | Heart rate analysis method, device and equipment | |
| CN106419884B (en) | A kind of rate calculation method and system based on wavelet analysis | |
| Pathan et al. | A machine learning based human activity recognition during physical exercise using wavelet packet transform of ppg and inertial sensors data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 |