CN114041801A - Electrocardiosignal reconstruction method and system based on PSA-EWT and DCGAN - Google Patents
Electrocardiosignal reconstruction method and system based on PSA-EWT and DCGAN Download PDFInfo
- Publication number
- CN114041801A CN114041801A CN202111349502.1A CN202111349502A CN114041801A CN 114041801 A CN114041801 A CN 114041801A CN 202111349502 A CN202111349502 A CN 202111349502A CN 114041801 A CN114041801 A CN 114041801A
- Authority
- CN
- China
- Prior art keywords
- layer
- signal
- ewt
- dcgan
- psa
- 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
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 14
- 230000008569 process Effects 0.000 claims abstract description 11
- 230000029058 respiratory gaseous exchange Effects 0.000 claims abstract description 11
- 230000002194 synthesizing effect Effects 0.000 claims abstract description 4
- 230000004913 activation Effects 0.000 claims description 30
- 230000006870 function Effects 0.000 claims description 18
- 238000001228 spectrum Methods 0.000 claims description 9
- 238000000354 decomposition reaction Methods 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 238000005259 measurement Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 7
- 230000011218 segmentation Effects 0.000 description 5
- 230000003042 antagnostic effect Effects 0.000 description 4
- 238000005457 optimization Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011946 reduction process Methods 0.000 description 1
- 230000000241 respiratory effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—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]
- A61B5/346—Analysis of electrocardiograms
-
- 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]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
- A61B5/35—Detecting specific parameters of the electrocardiograph cycle by template matching
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Cardiology (AREA)
- Heart & Thoracic Surgery (AREA)
- Molecular Biology (AREA)
- Pathology (AREA)
- Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention relates to an electrocardiosignal reconstruction method and an electrocardiosignal reconstruction system based on PSA-EWT and DCGAN. The invention aims to solve the problem that the conventional electrocardiogram signal is inconvenient to operate by adopting contact measurement. The process is as follows: firstly, the method comprises the following steps: decomposing the original BCG signal, dividing the original BCG signal into m intervals, constructing a band-pass filter in each interval, and reconstructing m components by m filter groups; II, secondly: optimizing and combining the m intervals to obtain heartbeat components; thirdly, the method comprises the following steps: establishing a countermeasure network; fourthly, the method comprises the following steps: acquiring an ECG signal and a respiration signal; synthesizing a BCG signal; taking the synthesized BCG signal as the input of a generator and the ECG signal as the input of a discriminator, and training until the maximum iteration number is reached to obtain a well-trained confrontation network; fifthly: and inputting the two obtained heartbeat components into a trained countermeasure network to reconstruct an ECG signal. The method is used for the field of electrocardiosignal reconstruction.
Description
Technical Field
The invention relates to an electrocardiosignal reconstruction method and an electrocardiosignal reconstruction system.
Background
The Ballistocardiogram (BCG) signals contain physiological parameters such as heartbeats in a sleep period, non-contact measurement is adopted, the BCG can be used for continuously monitoring vital sign parameters of a human body in real time without influencing the life of a user, but the non-contact BCG acquisition equipment is easily interfered by noise in the body movement and environment of the user, so that the application is limited. An Electrocardiogram (ECG) signal has a huge database and is widely applied to sleep analysis and diagnosis, but the operation is inconvenient by adopting contact measurement. The current techniques related to BCG signal processing are mostly focused on heart rate detection, and the potential association between BCG signal and ECG signal is not studied deeply.
Disclosure of Invention
The invention aims to solve the problems that the conventional Electrocardiogram (ECG) signal is measured in a contact mode and is inconvenient to operate, and provides an electrocardiosignal reconstruction method and an electrocardiosignal reconstruction system based on PSA-EWT and DCGAN.
The electrocardiosignal reconstruction method based on PSA-EWT and DCGAN comprises the following specific processes:
the method comprises the following steps: decomposing the original BCG signal by using an adaptive parameterless EWT method, and dividing the Fourier spectrum of the original BCG signal into m intervals, wherein the boundaries of the m intervals are omega ═ omega { (omega) }0,ω1,…,ωm-1,ωmAnd constructing a band-pass filter in each interval, constructing m filter groups in m intervals, and reconstructing m components u ═ u by the m filter groups1,…,um}；
Step two: obtaining m components u ═ { u ═ based on step one1,…,umUsing wavelet filter bank to reconstruct m-l new components u ═ u between each boundary1,…,um-l}；
Calculating new component u ═ { u ═ u1,…,um-lComputing the kurtosis of each component in u ═ u1,…,um-lSetting a threshold value alpha for the correlation coefficient of each component and the real ECG signal, and keeping the component of which the correlation coefficient is greater than the threshold value alpha and the kurtosis is the heartbeat component obtained by final decomposition;
step three: establishing a countermeasure network, wherein the countermeasure network comprises a generator and a discriminator;
step four: acquiring an ECG signal and a respiration signal of the MIT-BIH polysomnography database in the same time period;
synthesizing a BCG signal based on the acquired ECG signal and the respiration signal and 20dB white Gaussian noise;
taking the synthesized BCG signal as the input of a generator and the ECG signal as the input of a discriminator, and training until the maximum iteration number is reached to obtain a well-trained confrontation network;
step five: inputting the heartbeat component obtained in the second step into a trained countermeasure network, and reconstructing an ECG signal;
the PSA is a parameter-free scale space method; EWT is an empirical wavelet transform; the DCGAN generates a countermeasure network for deep convolution.
The electrocardiosignal reconstruction system based on PSA-EWT and DCGAN is used for executing the electrocardiosignal reconstruction method based on PSA-EWT and DCGAN.
The invention has the beneficial effects that:
according to the electrocardiosignal reconstruction method based on the ballistocardiogram decomposition of the improved empirical wavelet transform and the antagonistic network generated based on the improved depth convolution, the BCG signal is decomposed by using the empirical wavelet change introduced into the parameter-free scale space method to obtain the heartbeat component, and the ECG signal is reconstructed by the antagonistic network generated through the improved depth convolution on the basis. The invention can effectively separate the heartbeat component from the Ballistocardiogram (BCG) signal measured in a non-contact way, and reconstruct and recover the ECG signal, and the root mean square error is-16.8422 dB; the problem of current heart Electrograph (ECG) signal adopt contact measurement, inconvenient operation is solved.
General DCGAN: when the discriminator stops learning, the optimization purpose of the generator G is mainly to make the output false data infinitely close to the real data x, and when the discriminator is used for discrimination, the result that the false data is discriminated as the real data is obtained, namely D (G (z) ═ D (x), and the loss function of G is:
z is random noise input into G, G (z) is false data output from G, x is real data, D (x) is the result of D judging x, and D (G (z)) is the result of D judging (G (z)). Pdata(x) Is a distribution of x, Pz(z) is the prior distribution of z.
The improved DCGAN of the invention: the purpose of generating the network is to make the output data impossible to distinguish whether the network is true or false, but if the data output by the generating network is not close to the real data but the network is difficult to distinguish, this will result in that the data generated by the generating countermeasure network is not ideal data.
The invention improves the loss function of the generator by adding L which is the difference between the generated data G (z) and the real data x1Norm when it is difficult to discriminate a network even if generated data is not close to real data, L1The norm value is large and does not accord with the condition that the generator stops learning, so that the network is promoted to continue working, and the condition that the data generated by the generation countermeasure network is not ideal data is avoided. The loss function is:
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a waveform diagram of a raw ECG signal, a respiration signal, a BCG signal synthesized based on the raw ECG signal and the respiration signal;
FIG. 3a is a diagram of the original spectrum segmentation result;
FIG. 3b is a diagram of the result of spectrum segmentation after optimized merging;
FIG. 4a is a waveform diagram of components 1-4 in a heart beat signal decomposed and reconstructed using a modified EWT;
FIG. 4b is a waveform diagram of components 5-7 in a heart beat signal decomposed and reconstructed using the modified EWT;
FIG. 4c is a graph comparing a raw ECG with a component h (t) in a signal decomposed and reconstructed using a modified EWT;
FIG. 5a is a graph of a loss function in reconstructing an ECG signal using the modified DCGAN;
FIG. 5b is a plot of score values in an ECG signal reconstructed using modified DCGAN;
FIG. 5c is a graph of relationship numbers in an ECG signal reconstructed using a modified DCGAN;
FIG. 5d is a graph of the root mean square error in the reconstructed ECG signal using the modified DCGAN;
FIG. 5e is a comparison graph of the waveform reconstruction result of the heartbeat signal;
FIG. 6 is a diagram of a generator network architecture;
fig. 7 is a diagram of a discriminator network.
Detailed Description
The first embodiment is as follows: the present embodiment is described with reference to fig. 1, and the detailed procedure of the electrocardiographic signal reconstruction method based on PSA-EWT and DCGAN in the present embodiment is as follows:
the method comprises the following steps: decomposing the original BCG signal by using an adaptive parameterless EWT (empirical wavelet transform) method, and dividing the Fourier spectrum of the original BCG signal into m intervals, wherein the boundary of the m intervals is omega ═ omega0,ω1,…,ωm-1,ωmAnd constructing a band-pass filter in each interval, constructing m filter groups in m intervals, and reconstructing m components u ═ u by the m filter groups1,…,um}；
Step two: obtaining m components u ═ { u ═ based on step one1,…,umUsing wavelet filter bank to reconstruct m-l new components u ═ u between each boundary1,…,um-l}；
Calculating new component u ═ { u ═ u1,…,um-lComputing the kurtosis of each component in u ═ u1,…,um-lSetting a threshold value alpha for the correlation coefficient of each component and the real ECG signal, and keeping the component of which the correlation coefficient is greater than the threshold value alpha and the kurtosis is the heartbeat component obtained by final decomposition;
step three: establishing a countermeasure network, wherein the countermeasure network comprises a generator and a discriminator;
step four: acquiring an ECG signal and a respiration signal of the MIT-BIH polysomnography database in the same time period;
synthesizing a BCG signal based on the acquired ECG signal and the respiration signal and 20dB white Gaussian noise;
taking the synthesized BCG signal as the input of a generator and the ECG signal as the input of a discriminator, and training until the maximum iteration number is reached to obtain a well-trained confrontation network;
step five: inputting the heartbeat component obtained in the second step into a trained countermeasure network, and reconstructing an ECG signal;
the PSA is a parameter-free scale space method; EWT is an empirical wavelet transform; the DCGAN generates a countermeasure network for deep convolution.
The second embodiment is as follows: in this embodiment, the difference between the first embodiment and the second embodiment is that m components u ═ u { u } obtained in the first step in the second step1,…,umUsing wavelet filter bank to reconstruct m-l new components u ═ u between each boundary1,…,um-l}; the specific process is as follows:
merging the m section boundaries (m +1) obtained in the step one, setting the number of merging section types l as 2, … and m-1 (when the number of merging section types m-1 is, the corresponding boundary is m), defining i as 1 and … for each merging type, and obtaining a new section boundary of a Fourier spectrum after the types are merged, wherein the number of the merging section types l as 2, … and m-1 is m, and the corresponding boundary is m
Reconstructing m-l new components u ═ u between new inter boundaries using a wavelet filter bank1,…,um-l}。
Other steps and parameters are the same as those in the first embodiment.
The third concrete implementation mode: the difference between this embodiment and the first or second embodiment is that the kurtosis calculation formula in the second step is:
wherein N is the original BCG signal length, sigmaiIs the standard deviation, x, of the ith component of the original BCG signalijIs the jth component of the ith component of the original BCG signalThe signal point is a point of the signal,
is the average of the i-th component of the original BCG signal.
Other steps and parameters are the same as those in the first or second embodiment.
The fourth concrete implementation mode: the difference between this embodiment and one of the first to third embodiments is that, in the third step, an antagonistic network is established, and the antagonistic network includes a generator and a discriminator; the specific process is as follows:
the generator sequentially comprises a generator input layer, a matrix conversion layer, a first deconvolution layer, a first batch of normalized BN layers, a first activation layer, a second deconvolution layer, a second batch of normalized BN layers, a second activation layer, a third deconvolution layer, a third batch of normalized BN layers, a third activation layer, a fourth deconvolution layer, a fourth batch of normalized BN layers, a fourth activation layer, a fifth deconvolution layer and a generator output layer;
the discriminator comprises a discriminator input layer, a first convolution layer, a fifth active layer, a second convolution layer, a sixth active layer, a third convolution layer, a seventh active layer, a fourth convolution layer, an eighth active layer, a full-link layer and a discriminator output layer in sequence.
The generator aims to output a reconstructed ECG which is difficult to distinguish from an original (real) ECG signal, the discriminator finishes distinguishing operation, and the final state of the generator and the discriminator achieves dynamic balance, namely the generator can output false data which is as true as real data, and the discriminator cannot accurately distinguish whether the data discriminated by the discriminator is the real data or the false data.
The generator designed by the invention is a convolutional neural network structure, the number of the generator is 6, the input of the generator is firstly subjected to matrix transformation to change the dimension, the following layers of networks are all deconvolution layers, the output data dimension is gradually enlarged, the activation function uses LeakyReLU, batch normalization is used, the batch normalization can enable the network to be normalized, and the training speed is accelerated.
The discriminator is also in a 6-layer convolutional neural network structure, the first four layers are convolutional layers, the activation function uses LeakyReLU, the last layer is a full-link layer, and the output dimension is 1 multiplied by 1 to serve as the output value of the discriminator.
Setting the batch size to be 256 and the iteration number to be 80, training the generator and the discriminator by using 1087 training samples, and showing that the DCGAN training process is stable through experimental results.
Other steps and parameters are the same as those in one of the first to third embodiments.
The fifth concrete implementation mode: the difference between this embodiment and one of the first to fourth embodiments is that the activation functions of the first activation layer, the second activation layer, the third activation layer, and the fourth activation layer are all LeakyReLU.
Other steps and parameters are the same as in one of the first to fourth embodiments.
The sixth specific implementation mode: the difference between this embodiment and one of the first to fifth embodiments is that the activation functions of the fifth activation layer, the sixth activation layer, the seventh activation layer, and the eighth activation layer are all LeakyReLU.
Other steps and parameters are the same as those in one of the first to fifth embodiments.
The seventh embodiment: this embodiment differs from one of the first to sixth embodiments in that the loss function of the generator becomes
Wherein z is random noise input into the generator G, G (z) is false data output from the generator G, x is real data, D (x) is the result of the discriminator D for discriminating x, and D (G (z)) is the result of the discriminator D for discriminating (G (z)). Pz(z) is a prior distribution of z,
for averaging, α is a coefficient of the norm L1.
For the deep convolution generation countermeasure network, the loss function of the generation network is changed under the condition of not changing the discrimination network, and L is added on the basis of the loss function of the original generation network1Norm (alpha | G (z) -x | non-volatile memory cell|)。
General DCGAN: when the discriminator stops learning, the optimization purpose of the generator G is mainly to make the output false data infinitely close to the real data x, and when the discriminator is used for discrimination, the result that the false data is discriminated as the real data is obtained, namely D (G (z) ═ D (x), and the loss function of G is:
z is random noise input into G, G (z) is false data output from G, x is real data, D (x) is the result of D judging x, and D (G (z)) is the result of D judging (G (z)). Pdata(x) Is a distribution of x, Pz(z) is the prior distribution of z.
The improved DCGAN of the invention: the purpose of generating the network is to make the output data impossible to distinguish whether the network is true or false, but if the data output by the generating network is not close to the real data but the network is difficult to distinguish, this will result in that the data generated by the generating countermeasure network is not ideal data.
The invention improves the loss function of the generator by adding L which is the difference between the generated data G (z) and the real data x1Norm when it is difficult to discriminate a network even if generated data is not close to real data, L1The norm value is large and does not accord with the condition that the generator stops learning, so that the network is promoted to continue working, and the condition that the data generated by the generation countermeasure network is not ideal data is avoided. The loss function is:
Other steps and parameters are the same as those in one of the first to sixth embodiments.
The specific implementation mode is eight: the difference between this embodiment and one of the first to seventh embodiments is that the sizes of the first, second, third and fourth normalized BN layers are 256.
Other steps and parameters are the same as those in one of the first to seventh embodiments.
The specific implementation method nine: this embodiment differs from one of the first to eighth embodiments in that 0. ltoreq. alpha. ltoreq.1.
Other steps and parameters are the same as those in one to eight of the embodiments.
The detailed implementation mode is ten: the system for reconstructing an electrocardiographic signal based on PSA-EWT and DCGAN according to the present embodiment is characterized by being configured to execute the method for reconstructing an electrocardiographic signal based on PSA-EWT and DCGAN according to any one of the first to ninth embodiments.
The following examples were used to demonstrate the beneficial effects of the present invention:
the first embodiment is as follows:
the experimental BCG signal was synthesized using the ECG signal and the respiratory signal of the same time period from the MIT-BIH polysomnography database plus 20dB gaussian white noise as shown in fig. 2. Performing EWT spectrum segmentation according to the step pair, dividing the spectrum into 78 intervals as shown in figure 3a, performing optimization merging on the segmentation intervals by using the method provided by the invention to obtain 58 new segmentation intervals as shown in figure 3b, and reconstructing between boundaries of the merged new intervals to obtain 58 decomposed signals, wherein the first seven single-component signals are as shown in figures 4a and 4b, selecting the 7 th component with the largest kurtosis index to reconstruct a heartbeat signal and a respiration signal as shown in figure 4c, wherein the BCG signal is decomposed into a heartbeat component h (t) and a respiration component, but h (t) is still different from the original ECG signal. Setting the batch size to be 256 and the iteration number to be 80, training the improved generator and the improved discriminator by using 1087 training samples, and as can be seen from experimental results, after the generator is improved, the DCGAN training process is still stable, the loss function curve and the score value curve of the generator and the discriminator in the training process are shown in fig. 5a and 5b, the score value of the generator and the discriminator completely approaches to 0.5, and calculating the correlation coefficient and the root mean square error of the current generator generation data and the original data in the training process is shown in fig. 5c and 5 d. Along with the increase of training times, the correlation coefficient of the generated data and the original data is gradually close to 1, the root mean square error is reduced to be close to-16 dB, the fluctuation of the numerical value is small in the reduction process of RMSE, and as can be seen from fig. 5e, the fluctuation of the generated data at the end of the data is relieved before the fluctuation is changed by using the improved generator, and the waveform of the generated data is basically the same as that of the original ECG signal. Fig. 6 shows a network structure of a generator used in an experiment, and fig. 7 shows a network structure of a discriminator.
The present invention is capable of other embodiments and its several details are capable of modifications in various obvious respects, all without departing from the spirit and scope of the present invention.
Claims (10)
1. The electrocardiosignal reconstruction method based on PSA-EWT and DCGAN is characterized in that: the method comprises the following specific processes:
the method comprises the following steps: decomposing the original BCG signal by using an adaptive parameterless EWT method, and dividing the Fourier spectrum of the original BCG signal into m intervals, wherein the boundaries of the m intervals are omega ═ omega { (omega) }0,ω1,…,ωm-1,ωmAnd constructing a filter in each interval, constructing m filter groups in m intervals, and reconstructing m components u-by the m filter groups1,…,um}；
Step two: obtaining m components u ═ { u ═ based on step one1,…,umUsing wavelet filter bank to reconstruct m-l new components u ═ u between each boundary1,…,um-l}；
Calculating new component u ═ { u ═ u1,…,um-lComputing the kurtosis of each component in u ═ u1,…,um-lSetting a threshold value alpha for the correlation coefficient of each component and the real ECG signal, and keeping the component of which the correlation coefficient is greater than the threshold value alpha and the kurtosis is the heartbeat component obtained by final decomposition;
step three: establishing a countermeasure network, wherein the countermeasure network comprises a generator and a discriminator;
step four: acquiring an ECG signal and a respiration signal of the MIT-BIH polysomnography database in the same time period;
synthesizing a BCG signal based on the acquired ECG signal and the respiration signal and 20dB white Gaussian noise;
taking the synthesized BCG signal as the input of a generator and the ECG signal as the input of a discriminator, and training until the maximum iteration number is reached to obtain a well-trained confrontation network;
step five: inputting the heartbeat component obtained in the second step into a trained countermeasure network, and reconstructing an ECG signal;
the PSA is a parameter-free scale space method; EWT is an empirical wavelet transform; the DCGAN generates a countermeasure network for deep convolution.
2. The method for reconstructing electrocardiosignals based on PSA-EWT and DCGAN according to claim 1, wherein: in the second step, m components u ═ { u } obtained in the first step are obtained1,…,umUsing wavelet filter bank to reconstruct m-l new components u ═ u between each boundary1,…,um-l}; the specific process is as follows:
merging the m section boundaries obtained in the step one, setting the number l of the merged section types to be 2, … and m-1, defining i to be 1, … and m-l as the index value of the original boundary for each merged type, and obtaining a Fourier spectrum new section boundary after the types are merged as
Reconstructing m-l new components u ═ u between new inter boundaries using a wavelet filter bank1,…,um-l}。
3. The method for reconstructing electrocardiosignals based on PSA-EWT and DCGAN according to claim 2, wherein: the kurtosis calculation formula in the second step is as follows:
wherein N is the original BCG signal length, sigmaiIs the standard deviation, x, of the ith component of the original BCG signalijFor the jth signal point of the ith component of the original BCG signal,
is the average of the i-th component of the original BCG signal.
4. The method for reconstructing electrocardiosignals based on PSA-EWT and DCGAN according to claim 3, wherein: establishing a countermeasure network in the third step, wherein the countermeasure network comprises a generator and a discriminator; the specific process is as follows:
the generator sequentially comprises a generator input layer, a matrix conversion layer, a first deconvolution layer, a first batch of normalized BN layers, a first activation layer, a second deconvolution layer, a second batch of normalized BN layers, a second activation layer, a third deconvolution layer, a third batch of normalized BN layers, a third activation layer, a fourth deconvolution layer, a fourth batch of normalized BN layers, a fourth activation layer, a fifth deconvolution layer and a generator output layer;
the discriminator comprises a discriminator input layer, a first convolution layer, a fifth active layer, a second convolution layer, a sixth active layer, a third convolution layer, a seventh active layer, a fourth convolution layer, an eighth active layer, a full-link layer and a discriminator output layer in sequence.
5. The method for reconstructing electrocardiosignals based on PSA-EWT and DCGAN according to claim 4, wherein: the activation functions of the first activation layer, the second activation layer, the third activation layer and the fourth activation layer are LeakyReLU.
6. The method for reconstructing electrocardiosignals based on PSA-EWT and DCGAN according to claim 5, wherein: the activation functions of the fifth activation layer, the sixth activation layer, the seventh activation layer and the eighth activation layer are LeakyReLU.
7. The method for reconstructing electrocardiosignals based on PSA-EWT and DCGAN according to claim 6, wherein: loss function of the generator becomes
Wherein z is random noise input into the generator G, G (z) is false data output from the generator G, x is real data, D (x) is the result of the discriminator D for discriminating x, and D (G (z)) is the result of the discriminator D for discriminating (G (z)). Pz(z) is a prior distribution of z,
for averaging, α is a coefficient.
8. The method for reconstructing electrocardiosignals based on PSA-EWT and DCGAN according to claim 7, wherein: the sizes of the first batch of normalized BN layers, the second batch of normalized BN layers, the third batch of normalized BN layers and the fourth batch of normalized BN layers are 256.
9. The method for reconstructing electrocardiosignals based on PSA-EWT and DCGAN according to claim 8, wherein: alpha is more than or equal to 0 and less than or equal to 1.
10. The electrocardiosignal reconstruction system based on PSA-EWT and DCGAN is characterized in that: the system is used for executing the electrocardiosignal reconstruction method based on PSA-EWT and DCGAN of one of the claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111349502.1A CN114041801B (en) | 2021-11-15 | 2021-11-15 | Electrocardiosignal reconstruction method and system based on PSA-EWT and DCGAN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111349502.1A CN114041801B (en) | 2021-11-15 | 2021-11-15 | Electrocardiosignal reconstruction method and system based on PSA-EWT and DCGAN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114041801A true CN114041801A (en) | 2022-02-15 |
| CN114041801B CN114041801B (en) | 2024-03-08 |
Family
ID=80209394
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111349502.1A Active CN114041801B (en) | 2021-11-15 | 2021-11-15 | Electrocardiosignal reconstruction method and system based on PSA-EWT and DCGAN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114041801B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116327181A (en) * | 2023-05-30 | 2023-06-27 | 中国科学院苏州生物医学工程技术研究所 | Comprehensive evaluation method and device for real-time noninductive monitoring of heart and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109998500A (en) * | 2019-04-30 | 2019-07-12 | 陕西师范大学 | A kind of pulse signal generation method and system based on production confrontation network |
| CN111419213A (en) * | 2020-03-11 | 2020-07-17 | 哈尔滨工业大学 | ECG electrocardiosignal generation method based on deep learning |
| CN111990989A (en) * | 2020-08-31 | 2020-11-27 | 武汉大学 | Electrocardiosignal identification method based on generation countermeasure and convolution cyclic network |
| CN113080992A (en) * | 2021-03-30 | 2021-07-09 | 北京芯动卫士科技有限公司 | Atrial fibrillation identification method based on DCGAN and deep neural network method |
| CN113180679A (en) * | 2021-05-12 | 2021-07-30 | 武汉大学 | Single-lead electrocardio abnormal signal identification method |
-
2021
- 2021-11-15 CN CN202111349502.1A patent/CN114041801B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109998500A (en) * | 2019-04-30 | 2019-07-12 | 陕西师范大学 | A kind of pulse signal generation method and system based on production confrontation network |
| CN111419213A (en) * | 2020-03-11 | 2020-07-17 | 哈尔滨工业大学 | ECG electrocardiosignal generation method based on deep learning |
| CN111990989A (en) * | 2020-08-31 | 2020-11-27 | 武汉大学 | Electrocardiosignal identification method based on generation countermeasure and convolution cyclic network |
| CN113080992A (en) * | 2021-03-30 | 2021-07-09 | 北京芯动卫士科技有限公司 | Atrial fibrillation identification method based on DCGAN and deep neural network method |
| CN113180679A (en) * | 2021-05-12 | 2021-07-30 | 武汉大学 | Single-lead electrocardio abnormal signal identification method |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116327181A (en) * | 2023-05-30 | 2023-06-27 | 中国科学院苏州生物医学工程技术研究所 | Comprehensive evaluation method and device for real-time noninductive monitoring of heart and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114041801B (en) | 2024-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109171712B (en) | Atrial fibrillation identification method, atrial fibrillation identification device, atrial fibrillation identification equipment and computer readable storage medium | |
| CN108714026B (en) | Fine-grained electrocardiosignal classification method based on deep convolutional neural network and online decision fusion | |
| Übeyli | Combining recurrent neural networks with eigenvector methods for classification of ECG beats | |
| CN111990989A (en) | Electrocardiosignal identification method based on generation countermeasure and convolution cyclic network | |
| Patro et al. | Effective feature extraction of ECG for biometric application | |
| Sepahvand et al. | A novel multi-lead ECG personal recognition based on signals functional and structural dependencies using time-frequency representation and evolutionary morphological CNN | |
| Guan et al. | Low-dimensional denoising embedding transformer for ECG classification | |
| CN115363586A (en) | Psychological stress grade assessment system and method based on pulse wave signals | |
| CN116361688A (en) | Multi-mode feature fusion model construction method for automatic classification of electrocardiographic rhythms | |
| Tan et al. | Best wavelet function identification system for ECG signal denoise applications | |
| Kuila et al. | Feature extraction of electrocardiogram signal using machine learning classification | |
| CN115470832A (en) | Electrocardiosignal data processing method based on block chain | |
| CN114041801A (en) | Electrocardiosignal reconstruction method and system based on PSA-EWT and DCGAN | |
| Phinyomark et al. | Applications of variance fractal dimension: A survey | |
| CN116999063A (en) | Method for realizing electrocardiographic atrial fibrillation detection based on signal decomposition and convolution network | |
| RU2383295C1 (en) | Method of electrocardiosignal processing for diagnostics of myocardial infarction | |
| CN115500807A (en) | Arrhythmia classification detection method and system based on small convolutional neural network | |
| Khandait et al. | ECG signal processing using classifier to analyses cardiovascular disease | |
| CN111543990B (en) | Muscle cooperation extraction method and muscle cooperation-based exercise function analysis and evaluation method | |
| Karimifard et al. | Morphological heart arrhythmia classification using Hermitian model of higher-order statistics | |
| CN111493864B (en) | EEG signal mixed noise processing method, equipment and storage medium | |
| Lee et al. | Multiscale sample entropy based on discrete wavelet transform for clinical heart rate variability recognition | |
| CN116570295B (en) | Electrocardiogram low-voltage T-wave end point positioning method | |
| KRISHNAN et al. | Automated diagnosis of cardiac health using recurrence quantification analysis | |
| CN116777793B (en) | Medical image prompt denoising technology based on logarithmic guide contrast learning |
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 |