CN108694409B - Reconstruction method and device of electrocardiogram data - Google Patents
Reconstruction method and device of electrocardiogram data Download PDFInfo
- Publication number
- CN108694409B CN108694409B CN201710236488.1A CN201710236488A CN108694409B CN 108694409 B CN108694409 B CN 108694409B CN 201710236488 A CN201710236488 A CN 201710236488A CN 108694409 B CN108694409 B CN 108694409B
- Authority
- CN
- China
- Prior art keywords
- dictionary
- matrix
- training
- overcomplete
- initialized
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2136—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on sparsity criteria, e.g. with an overcomplete basis
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention provides a reconstruction method and a device of electrocardiogram data, which are characterized in that after a personalized over-complete dictionary is judged to be generated, the received compressed electrocardiogram data of a user is subjected to data reconstruction by using the personalized over-complete dictionary and the existing OMP algorithm, so that first reconstructed electrocardiogram data can be obtained, and medical care personnel can conveniently use the reconstructed electrocardiogram data to realize remote electrocardiogram diagnosis on a patient.
Description
Technical Field
The invention relates to the technical field of biomedical engineering, in particular to a method and a device for reconstructing electrocardio data.
Background
In recent years, as the incidence of heart diseases increases year by year, the traditional face-to-face diagnosis mode between patients and medical care personnel cannot meet the requirements of people on health, so that the method for remotely monitoring the electrocardiographic data of the patients by using a wireless sensing network is gradually becoming a favored electrocardiographic monitoring means.
At present, collected electrocardiographic data of a patient is compressed and then transmitted to a terminal device held by a medical worker through a wireless sensing network, so that the data volume of the electrocardiographic data to be transmitted can be reduced, the phenomena of delay and packet loss are avoided, and then the terminal device held by the medical worker decodes the received electrocardiographic data, namely data reconstruction is carried out, so that original electrocardiographic data before being compressed is obtained, and the medical worker can diagnose the electrocardiographic data conveniently. However, the conventional method for reconstructing data of electrocardiographic data by using a terminal device held by medical staff mainly completes reconstruction operation by directly using an overcomplete dictionary preset in the terminal device and a conventional OMP algorithm, but because the conventional overcomplete dictionary is obtained by training based on historical electrocardiographic data of a large number of different users, the electrocardiographic data reconstructed by using the overcomplete dictionary is different from the original electrocardiographic data of a single user, so that the accuracy of the reconstructed electrocardiographic data is reduced, and misdiagnosis of the medical staff is easily caused.
Disclosure of Invention
In view of this, the present invention provides a method and an apparatus for reconstructing electrocardiographic data, so as to improve the accuracy of the electrocardiographic data obtained after reconstruction, thereby effectively avoiding occurrence of misdiagnosis of medical care personnel.
In order to achieve the purpose, the invention provides the following technical scheme:
a reconstruction method of electrocardiogram data comprises the following steps:
receiving compressed electrocardiogram data of a user;
judging whether a personalized over-complete dictionary is generated, wherein the personalized over-complete dictionary is an over-complete dictionary generated by training according to original electrocardiogram data of the user, and the original electrocardiogram data are not compressed electrocardiogram data;
and if the personalized overcomplete dictionary is generated, reconstructing the compressed electrocardiogram data by using the personalized overcomplete dictionary and an OMP algorithm to obtain first reconstructed electrocardiogram data.
Preferably, the process of generating the personalized overcomplete dictionary comprises:
receiving original electrocardiogram data of the user within a preset time;
grouping the original electrocardiogram data according to receiving time to obtain a first initialized over-complete dictionary and a first training sample;
and training the first initialized overcomplete dictionary by using the first training sample to generate the personalized overcomplete dictionary.
Preferably, the training the first initialized overcomplete dictionary by using the first training sample to generate the personalized overcomplete dictionary includes:
performing matrix transformation on the first initialized overcomplete dictionary to generate a second initialized overcomplete dictionary;
generating a target iteration number corresponding to the second initialized overcomplete dictionary according to the matrix attribute of the second initialized overcomplete dictionary, wherein the matrix attribute is the number of rows and the number of columns of the second initialized overcomplete dictionary;
generating a required atom number parameter, wherein the required atom number parameter is the number of columns in the second initialized overcomplete dictionary required by the original electrocardiogram data;
performing matrix transformation on the first training sample to generate a second training sample;
taking the second initialized overcomplete dictionary as a sensing matrix, taking the required atom number parameter as a first sparsity, and calculating a sparse coefficient matrix of the second training sample by using the OMP algorithm, wherein the sensing matrix and the first sparsity are the sensing matrix and sparsity adopted in the OMP algorithm;
taking the needed atom number parameter as a second sparsity, and training the second initialized overcomplete dictionary, the sparse coefficient matrix of the second training sample and the second training sample by using a K-SVD algorithm to generate a training update dictionary, wherein the second sparsity is the sparsity adopted in the K-SVD algorithm;
recording the generation times of the training updated dictionary;
judging whether the generation times reach the target iteration times or not;
if the generation times reach the target iteration times, taking the training updated dictionary as the personalized overcomplete dictionary;
and if the generation times do not reach the target iteration times, taking the training updated dictionary as the second initialization overcomplete dictionary, returning the number parameter of the needed atoms as a second sparsity, and training the second initialization overcomplete dictionary, the sparse coefficient matrix of the second training sample and the second training sample by utilizing a K-SVD algorithm to generate the training updated dictionary.
Preferably, the reconstructing the compressed electrocardiographic data by using the personalized overcomplete dictionary and the OMP algorithm to obtain first reconstructed electrocardiographic data includes:
multiplying the personalized overcomplete dictionary by a compression matrix to obtain a multiplication matrix, wherein the compression matrix is used for compressing the original electrocardiogram data of the user;
taking the multiplication matrix as the sensing matrix, taking the required atom number parameter as the first sparsity, taking the compressed electrocardiogram data as a sampling vector, and calculating a sparse coefficient matrix of the electrocardiogram data by utilizing the OMP algorithm;
and multiplying the personalized over-complete dictionary by the sparse coefficient matrix of the electrocardiogram data to obtain the first reconstructed electrocardiogram data.
Preferably, after determining that the personalized overcomplete dictionary is not generated, the method further includes:
and reconstructing the compressed electrocardiogram data by using the alternative over-complete dictionary to obtain second reconstructed electrocardiogram data, wherein the alternative over-complete dictionary is obtained in advance.
An apparatus for reconstructing electrocardiographic data, comprising:
the compressed data receiving module is used for receiving compressed electrocardio data of a user;
the first judgment module is used for judging whether a personalized overcomplete dictionary is generated or not, wherein the personalized overcomplete dictionary is an overcomplete dictionary generated by training according to original electrocardiogram data of the user, and the original electrocardiogram data are uncompressed electrocardiogram data;
and the first reconstruction module is used for reconstructing the compressed electrocardiogram data by utilizing the personalized overcomplete dictionary and an OMP algorithm to obtain first reconstructed electrocardiogram data if the personalized overcomplete dictionary is generated.
Preferably, the apparatus further comprises:
the original data receiving module is used for receiving original electrocardio data of the user within preset time;
the grouping module is used for grouping the original electrocardiogram data according to receiving time to obtain a first initialized over-complete dictionary and a first training sample;
and the first training module is used for training the first initialized overcomplete dictionary by using the first training sample to generate the personalized overcomplete dictionary.
Preferably, the first training module comprises:
the first matrix conversion module is used for performing matrix conversion on the first initialized overcomplete dictionary to generate a second initialized overcomplete dictionary;
the iteration number generation module is used for generating a target iteration number corresponding to the second initialized overcomplete dictionary according to the matrix attribute of the second initialized overcomplete dictionary, wherein the matrix attribute is the row number and the column number of the second initialized overcomplete dictionary;
a required atom number parameter generating module, configured to generate a required atom number parameter, where the required atom number parameter is a number of columns in the second initialized overcomplete dictionary required for representing the original electrocardiographic data;
the second matrix conversion module is used for performing matrix conversion on the first training sample to generate a second training sample;
a first calculation module, configured to use the second initialized overcomplete dictionary as a sensing matrix, use the required atom number parameter as a first sparsity, and calculate a sparse coefficient matrix of the second training sample by using the OMP algorithm, where the sensing matrix and the first sparsity are the sensing matrix and sparsity adopted in the OMP algorithm;
the second training module is used for taking the needed atom number parameter as a second sparsity, training the second initialized overcomplete dictionary, the sparse coefficient matrix of the second training sample and the second training sample by utilizing a K-SVD algorithm, and generating a training updating dictionary, wherein the second sparsity is the sparsity adopted in the K-SVD algorithm;
the recording module is used for recording the generation times of the training updated dictionary;
the second judgment module is used for judging whether the generation times reach the target iteration times;
the selection module is used for taking the training updated dictionary as the personalized overcomplete dictionary if the generation times reach the target iteration times;
the updating module is used for taking the training updated dictionary as the second initialized overcomplete dictionary if the generation times do not reach the target iteration times;
the second training module is further configured to, after the updating module uses the training updated dictionary as the second initialized overcomplete dictionary, use the required atom number parameter as a second sparsity, and train the second initialized overcomplete dictionary, the sparse coefficient matrix of the second training sample, and the second training sample by using a K-SVD algorithm to generate the training updated dictionary, where the second sparsity is a sparsity adopted in the K-SVD algorithm.
Preferably, the first reconstruction module comprises:
the second calculation module is used for multiplying the personalized over-complete dictionary by a compression matrix to obtain a multiplication matrix, wherein the compression matrix is used for compressing the original electrocardiogram data of the user;
the third calculation module is used for taking the multiplication matrix as the sensing matrix, taking the required atom number parameter as the first sparsity, taking the compressed electrocardio data as a sampling vector, and calculating a sparse coefficient matrix of the electrocardio data by utilizing the OMP algorithm;
and the fourth calculation module is used for multiplying the personalized over-complete dictionary by the sparse coefficient matrix of the electrocardiographic data to obtain the first reconstructed electrocardiographic data.
Preferably, the apparatus further comprises:
and the second reconstruction module is used for reconstructing the compressed electrocardiogram data by using the alternative overcomplete dictionary to obtain second reconstructed electrocardiogram data if the first judgment module judges that the personalized overcomplete dictionary is not generated, wherein the alternative overcomplete dictionary is obtained in advance.
According to the technical scheme, compared with the prior art, the method and the device for reconstructing the electrocardiogram data provided by the invention have the advantages that the received compressed electrocardiogram data of the user is subjected to data reconstruction by utilizing the personalized overcomplete dictionary and the existing OMP algorithm after the generation of the personalized overcomplete dictionary is judged, the first reconstructed electrocardiogram data can be obtained, so that medical care personnel can conveniently use the reconstructed electrocardiogram data to realize the remote electrocardiogram diagnosis of the patient, and therefore, the reconstruction operation of the compressed electrocardiogram data of the user is completed by utilizing the personalized overcomplete dictionary obtained by training according to the original electrocardiogram data of the user, the accuracy of the reconstructed electrocardiogram data can be improved, and the misdiagnosis probability of the medical care personnel is effectively reduced.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a flowchart of a method for reconstructing electrocardiographic data according to an embodiment of the present invention;
fig. 2 is a flowchart of a method for generating a personalized overcomplete dictionary according to an embodiment of the present invention;
FIG. 3 is a flowchart of another method for reconstructing electrocardiographic data according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of an apparatus for reconstructing electrocardiographic data according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a device for generating a personalized overcomplete dictionary according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of another electrocardiographic data reconstruction apparatus according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment of the invention discloses a reconstruction method of electrocardiographic data, please refer to the attached figure 1, and the method specifically comprises the following steps:
s101: receiving compressed electrocardiogram data of a user;
specifically, when the medical staff uses the terminal device held by the medical staff to perform remote electrocardiographic monitoring on the user, the compressed electrocardiographic data of the corresponding user, which is transmitted from the terminal device held by each user, can be received through the wireless sensing network, that is, the compressed electrocardiographic data of the user is received, so that basic data is provided for the subsequent reconstruction operation. In order to reduce the occurrence probability of delay or packet loss in the process of transmitting the electrocardiographic data of the user by the wireless sensor network, the acquired original electrocardiographic data of each user can be compressed first, and then the compressed electrocardiographic data is transmitted, so that correspondingly, the electrocardiographic data initially received by the terminal device held by the medical staff is the compressed electrocardiographic data.
The terminal device held by the medical staff may include: desktop computers, smart phones, tablet computers, and the like.
The terminal device held by each user may include: smart phones, wearable electronics, and the like.
S102: judging whether a personalized over-complete dictionary is generated, wherein the personalized over-complete dictionary is an over-complete dictionary generated according to original electrocardiogram data training of the user, the original electrocardiogram data is uncompressed electrocardiogram data, and if yes, executing S103;
specifically, the personalized over-complete dictionary is generated according to the original electrocardiogram data of the user, so that the electrocardiogram data of the user can be matched more accurately, the accuracy of the electrocardiogram data obtained by subsequent reconstruction is improved, and the condition that medical staff misdiagnoses due to inaccurate reconstructed electrocardiogram data is effectively avoided.
Each user has a respective personalized overcomplete dictionary, and the generation mode of each personalized overcomplete dictionary can be generated on terminal equipment held by medical staff in advance so as to reconstruct the electrocardiogram data of the user in the following process.
S103: reconstructing the compressed electrocardiogram data by utilizing the personalized overcomplete dictionary and the OMP algorithm to obtain first reconstructed electrocardiogram data;
specifically, the generated personalized overcomplete dictionary is utilized, and the existing OMP algorithm is combined to reconstruct the received compressed electrocardiogram data, so that the original electrocardiogram data of the user, which is initially acquired by the terminal device held by the user, can be accurately restored, and the remote electrocardiogram diagnosis of the user can be completed on the basis of improving the diagnosis accuracy of medical staff.
According to the reconstruction method of the electrocardiographic data disclosed by the embodiment of the invention, after the generation of the personalized over-complete dictionary is judged, the received compressed electrocardiographic data of the user is subjected to data reconstruction by using the personalized over-complete dictionary and the existing OMP algorithm, so that the first reconstructed electrocardiographic data can be obtained, and medical care personnel can conveniently use the reconstructed electrocardiographic data to realize remote electrocardiographic diagnosis on the patient.
After the original electrocardiogram data of the user is received, how to accurately generate the personalized overcomplete dictionary is an important step in the process of subsequently completing the reconstruction operation by utilizing the personalized overcomplete dictionary. Therefore, how to generate the personalized over-complete dictionary quickly and accurately belongs to a key point concerned by the scheme.
Therefore, as shown in fig. 2, for S102 in the embodiment corresponding to fig. 1, an embodiment of the present invention discloses a method for generating a personalized overcomplete dictionary, which specifically includes the following steps:
s201: receiving original electrocardiogram data of the user within a preset time;
specifically, because the personalized overcomplete dictionary is generated by training according to the original electrocardiographic data of the user, the acquired original electrocardiographic data needs to be acquired from the terminal device held by the user; secondly, in order to increase the speed of generating the personalized over-complete dictionary, the original electrocardiographic data collected within a period of time, namely the original electrocardiographic data of the user within a preset time, can be received, wherein the preset time can be any time period such as 5 hours, one day, three days and the like.
S202: grouping the original electrocardiogram data according to receiving time to obtain a first initialized over-complete dictionary and a first training sample;
for example, if a terminal device held by a medical care worker receives original electrocardiographic data of a user in 48 hours transmitted by a terminal device held by the user, the original electrocardiographic data in 48 hours are sequentially grouped according to time, so that the original electrocardiographic data in the first 24 hours and the original electrocardiographic data in the last 24 hours, that is, the first initialized overcomplete dictionary and the first training sample, are obtained, so that related training operations are performed subsequently.
S203: performing matrix transformation on the first initialized overcomplete dictionary to generate a second initialized overcomplete dictionary, and executing S204;
specifically, since the first initialized overcomplete dictionary is a set of a plurality of original electrocardiographic data within a time period, a matrix conversion process needs to be performed on the first initialized overcomplete dictionary, so that a matrix is obtained and used as a second initialized overcomplete dictionary for subsequent correlation calculation; the method for matrix transformation of the first initialized overcomplete dictionary may be that a window with a length of K is adopted to randomly select the original electrocardiographic data in the N first initialized overcomplete dictionaries, so as to obtain an N x K matrix, that is, the second initialized overcomplete dictionary, and values of N and K may be arbitrarily set according to actual requirements, which is not limited herein.
S204: generating a target iteration number corresponding to the second initialized overcomplete dictionary according to the matrix attribute of the second initialized overcomplete dictionary, wherein the matrix attribute is the number of rows and the number of columns of the second initialized overcomplete dictionary, and executing S205;
specifically, because the generated second initialized overcomplete dictionary has different matrix sizes, in order to ensure that after multiple iterative computations, a personalized overcomplete dictionary more conforming to the change rule of the original electrocardiographic data of the user can be generated by training, after the second initialized overcomplete dictionary is obtained, the corresponding times of iteration, namely the target iteration times, which are required to be performed are generated according to the matrix size, namely the matrix attribute, of the second initialized overcomplete dictionary.
S205: generating a required atom number parameter, wherein the required atom number parameter is the number of columns in the second initialized overcomplete dictionary required for representing the original electrocardiogram data, and executing S206;
specifically, the generated second initialized overcomplete dictionary is a matrix, the matrix includes a plurality of rows and columns, and the length of each original electrocardiographic data of the user needs to be represented by a plurality of columns in the matrix.
For example, the generated second initialized overcomplete dictionary is a 5 × 10 matrix, where 10 is the total number of columns in the matrix, and the length of each piece of raw electrocardiographic data of the user is 6, that is, the length of each piece of raw electrocardiographic data needs to select 6 columns from 10 columns in the matrix for representation.
S206: performing matrix transformation on the first training sample to generate a second training sample, and executing S207;
specifically, because the first training sample is a set of a plurality of original electrocardiographic data within a time period, a matrix conversion process needs to be performed on the first training sample, so as to obtain a matrix, which is used as a second training sample for subsequent correlation calculation; the method for performing matrix transformation on the first training sample may be to randomly select original electrocardiographic data in M first training samples by using a window with a length of K, so as to obtain an M × K matrix, that is, a second training sample, where the length of the window used may be the same as the window used when performing matrix transformation on the first initialized overcomplete dictionary, and a value of M may be arbitrarily set according to actual requirements, which is not limited herein.
S207: taking the second initialized overcomplete dictionary as a sensing matrix, taking the required atom number parameter as a first sparsity, calculating a sparse coefficient matrix of the second training sample by using the OMP algorithm, wherein the sensing matrix and the first sparsity are the sensing matrix and sparsity adopted in the OMP algorithm, and executing S208;
specifically, the generated second initialized overcomplete dictionary is used as a calculation variable-sensing matrix required in the existing OMP algorithm, the required atom number parameter is used as another calculation variable-sparsity required in the OMP algorithm, and the sparse coefficient matrix of the second training sample is calculated by combining the OMP algorithm.
S208: taking the needed atom number parameter as a second sparsity, training the second initialized overcomplete dictionary, the sparse coefficient matrix of the second training sample and the second training sample by using a K-SVD algorithm to generate a training update dictionary, wherein the second sparsity is the sparsity adopted in the K-SVD algorithm, and executing S209;
specifically, after the sparse coefficient matrix of the second training sample is calculated by using the existing OMP algorithm, the generated second initialized overcomplete dictionary, the sparse coefficient matrix of the second training sample and the second training sample are trained by using the existing K-SVD algorithm to obtain a training updated dictionary, and basic data are provided for subsequently obtaining the personalized overcomplete dictionary.
The generated required atom number parameter is used as a calculation variable-sparsity required in the conventional K-SVD algorithm, so that the training updated dictionary generated by training is more close to the jump rule of the original electrocardiogram data of the user, and the accuracy of representing the original electrocardiogram data by the personalized overcomplete dictionary is improved.
S209: recording the generation times of the training updated dictionary, and executing S2010;
specifically, in order to complete the iterative training of the training updated dictionary within the specified number of iterations, the cumulative number of current iterations, that is, the number of times of generation of the training updated dictionary, needs to be recorded after each iteration is completed, so as to avoid performing excessive iterations, thereby increasing the training time.
S2010: judging whether the generation times reach the target iteration times, if so, executing S2011, and if not, executing S2012;
s2011: taking the training updated dictionary as the personalized overcomplete dictionary;
specifically, after the recorded generation times reach the preset target iteration times, it is proved that the updated training dictionary generated by training in the current iteration operation reaches the training standard, and the updated training dictionary can be used as an individualized over-complete dictionary for subsequent electrocardiographic data reconstruction operation.
S2012: taking the training updated dictionary as the second initialized overcomplete dictionary, and returning to execute S208;
specifically, after the recorded generation times do not reach the preset target iteration times, it is proved that the training updated dictionary generated by training in the current iteration operation does not reach the training standard, the training updated dictionary generated by the training still needs to be used as a second initialized overcomplete dictionary required in the next iteration operation, and the iteration updating is continued to obtain the training updated dictionary closer to the jump rule of the original electrocardiogram data of the user.
The above steps S203 to S2012 are only one preferable implementation manner of the process of training the first initialized overcomplete dictionary by using the first training sample and generating the personalized overcomplete dictionary, which is disclosed in the embodiment of the present invention, and a specific implementation manner related to the process may be arbitrarily set according to an actual requirement, which is not limited herein.
The embodiment of the invention discloses a method for generating an individualized overcomplete dictionary, which comprises the steps of respectively carrying out matrix transformation on a first initialized overcomplete dictionary and a first training sample which are obtained by grouping according to original electrocardiogram data received within preset time to obtain a second initialized overcomplete dictionary and a second training sample, correspondingly generating target iteration times and required atomic number parameters, and then sequentially combining the existing OMP algorithm and K-SVD algorithm to calculate so as to obtain the individualized overcomplete dictionary, wherein the training is carried out by utilizing the original electrocardiogram data within the preset time, so that the training time of the individualized overcomplete dictionary can be effectively shortened, the first initialized overcomplete dictionary and the first training sample which are generated according to the original electrocardiogram data of a user are combined with the existing OMP algorithm and K-SVD algorithm to calculate so as to generate the individualized overcomplete dictionary, the matching degree of the personalized over-complete dictionary and the original electrocardiogram data of the user can be improved, and the accuracy of the original electrocardiogram data representation in the electrocardiogram data reconstruction process is further improved.
On the basis of the embodiment corresponding to fig. 1, another method for reconstructing electrocardiographic data is disclosed in the embodiment of the present invention, please refer to fig. 3, in which the method specifically includes the following steps:
s301: and receiving the compressed electrocardiogram data of the user.
S302: and judging whether a personalized over-complete dictionary is generated, wherein the personalized over-complete dictionary is an over-complete dictionary generated according to original electrocardiogram data training of the user, the original electrocardiogram data is uncompressed electrocardiogram data, if so, executing S303, and if not, executing S304.
S303: multiplying the personalized overcomplete dictionary by a compression matrix to obtain a multiplication matrix, wherein the compression matrix is used for compressing the original electrocardiogram data of the user, and S305 is executed;
specifically, the compression matrix adopted in the process of obtaining the multiplication matrix is the same as the compression data adopted by the terminal device held by each patient user when compressing the acquired original electrocardiographic data of the user, so as to further improve the accuracy of the restored original electrocardiographic data.
S304: reconstructing the compressed electrocardiogram data by using an alternative over-complete dictionary to obtain second reconstructed electrocardiogram data, wherein the alternative over-complete dictionary is obtained in advance;
specifically, since it takes a certain time to generate the personalized overcomplete dictionary on the terminal device held by the medical care personnel in advance, when the terminal device held by the medical care personnel does not generate the personalized overcomplete dictionary yet and the compressed electrocardiographic data is received from the terminal device held by the user, the reconstruction operation of the compressed electrocardiographic data can be completed by using the preset alternative overcomplete dictionary, so as to avoid the problem of reconstruction delay caused by untrained generation of the personalized overcomplete dictionary.
In order to ensure the accuracy of the reconstructed second electrocardiographic data, the overcomplete candidate dictionary according to the embodiment of the present invention may be a Discrete Cosine Transform (DCT) redundant dictionary.
S305: taking the multiplication matrix as the sensing matrix, taking the required atom number parameter as the first sparsity, taking the compressed electrocardiogram data as a sampling vector, calculating a sparse coefficient matrix of the electrocardiogram data by using the OMP algorithm, and executing S306;
specifically, the generated multiplication matrix is used as a calculation variable-sensing matrix required in the existing OMP algorithm, the required atom number parameter is used as another calculation variable-sparsity required in the OMP algorithm, the compressed electrocardio data is used as a third calculation variable-sampling vector required in the OMP algorithm, and the sparse coefficient matrix of the electrocardio data is calculated by combining the OMP algorithm and is used for accurately restoring the original electrocardio data in the follow-up process.
S306: and multiplying the personalized over-complete dictionary by the sparse coefficient matrix of the electrocardiogram data to obtain the first reconstructed electrocardiogram data.
According to the reconstruction method of the electrocardiographic data, disclosed by the embodiment of the invention, a multiplication matrix obtained by multiplying the personalized overcomplete dictionary by the compression matrix is used as a sensing matrix of an OMP algorithm, the needed atom number parameter generated in the generation process of the personalized overcomplete dictionary is used as the first sparsity of the OMP algorithm, the OMP algorithm is combined to calculate the sparse coefficient matrix of the electrocardiographic data, and then the sparse coefficient matrix is multiplied by the personalized overcomplete dictionary to obtain the restored first reconstructed electrocardiographic data, so that the personalized overcomplete dictionary is introduced into the OMP algorithm in the reconstruction process of the electrocardiographic data to improve the reconstruction precision, and meanwhile, when the personalized overcomplete dictionary is not generated, the received compressed electrocardiographic data is reconstructed by using the preset alternative overcomplete dictionary, and the time required for reconstruction can be effectively shortened.
The embodiment of the invention discloses an electrocardiographic data reconstruction device, please refer to the attached figure 4, the device comprises:
a compressed data receiving module 401, configured to receive compressed electrocardiographic data of a user;
a first judging module 402, configured to judge whether a personalized overcomplete dictionary is generated, where the personalized overcomplete dictionary is an overcomplete dictionary generated according to training of original electrocardiographic data of the user, and the original electrocardiographic data is electrocardiographic data that is not compressed;
a first reconstructing module 403, configured to reconstruct the compressed electrocardiographic data by using the personalized overcomplete dictionary and an OMP algorithm if the personalized overcomplete dictionary is generated, so as to obtain first reconstructed electrocardiographic data.
According to the reconstruction device for the electrocardiographic data disclosed by the embodiment of the invention, after the first judgment module 402 judges that the personalized overcomplete dictionary is generated, the first reconstruction module 403 performs data reconstruction on the received compressed electrocardiographic data of the user by using the personalized overcomplete dictionary and the existing OMP algorithm, and can acquire the first reconstructed electrocardiographic data, so that medical care personnel can use the reconstructed electrocardiographic data to realize remote electrocardiographic diagnosis on a patient.
Please refer to a method flowchart corresponding to fig. 1 for the working process of each module provided in the embodiment of the present invention, and detailed description of the working process is omitted.
As shown in fig. 5, an embodiment of the present invention discloses a device for generating a personalized overcomplete dictionary, which is directed to the personalized overcomplete dictionary in the embodiment corresponding to fig. 4, and includes:
an original data receiving module 501, configured to receive original electrocardiographic data of the user within a preset time;
a grouping module 502, configured to group the original electrocardiographic data according to receiving time to obtain a first initialized overcomplete dictionary and a first training sample;
a first training module 503, configured to train the first initialized overcomplete dictionary by using the first training sample, and generate the personalized overcomplete dictionary.
Wherein the first training module 503 comprises:
a first matrix conversion module 5031, configured to perform matrix conversion on the first initialized overcomplete dictionary to generate a second initialized overcomplete dictionary;
an iteration number generating module 5032, configured to generate a target iteration number corresponding to the second initialized overcomplete dictionary according to a matrix attribute of the second initialized overcomplete dictionary, where the matrix attribute is a number of rows and a number of columns of the second initialized overcomplete dictionary;
a required atom number parameter generating module 5033, configured to generate a required atom number parameter, where the required atom number parameter is a number of columns in the second initialized overcomplete dictionary required for representing the original electrocardiographic data;
a second matrix conversion module 5034, configured to perform matrix conversion on the first training sample to generate a second training sample;
a first calculating module 5035, configured to use the second initialized overcomplete dictionary as a sensing matrix, use the required atom number parameter as a first sparsity, and calculate a sparse coefficient matrix of the second training sample by using the OMP algorithm, where the sensing matrix and the first sparsity are a sensing matrix and a sparsity adopted in the OMP algorithm;
a second training module 5036, configured to use the required atom number parameter as a second sparsity, and train the second initialized overcomplete dictionary, the sparse coefficient matrix of the second training sample, and the second training sample by using a K-SVD algorithm to generate a training update dictionary, where the second sparsity is a sparsity adopted in the K-SVD algorithm;
a recording module 5037, configured to record the number of times of generation of the training updated dictionary;
a second determining module 5038, configured to determine whether the generation number reaches the target iteration number;
a selecting module 5039, configured to, if the generation number reaches the target iteration number, use the training updated dictionary as the personalized overcomplete dictionary;
an updating module 50310, configured to, if the generation number does not reach the target iteration number, use the training updated dictionary as the second initialized overcomplete dictionary;
the second training module 5036 is further configured to, after the updating module 50310 uses the training updated dictionary as the second initialized overcomplete dictionary, use the required atom number parameter as a second sparsity, and train the second initialized overcomplete dictionary, the sparse coefficient matrix of the second training sample, and the second training sample by using a K-SVD algorithm to generate a training updated dictionary, where the second sparsity is a sparsity adopted in the K-SVD algorithm.
The device for generating a personalized overcomplete dictionary disclosed in the embodiment of the present invention performs matrix transformation on a first initialized overcomplete dictionary and a first training sample obtained by grouping the grouping module 502 according to the original electrocardiographic data received within a preset time through the first matrix conversion module 5031 and the second matrix conversion module 5034, respectively, to obtain a second initialized overcomplete dictionary and a second training sample, generates a target iteration number and a required atomic number parameter by the iteration number generation module 5032 and the required atomic number parameter generation module 5033, and calculates by combining the existing OMP algorithm and the existing K-SVD algorithm in sequence, and then determines the personalized overcomplete dictionary through the selection module 5039, thus training by using the original electrocardiographic data within the preset time can effectively reduce the training time of the personalized overcomplete dictionary, and combining the first initialized overcomplete dictionary and the first training sample generated according to the original electrocardiographic data of the user The existing OMP algorithm and K-SVD algorithm are used for calculating to generate the personalized over-complete dictionary, so that the matching degree of the personalized over-complete dictionary and the original electrocardiogram data of the user can be improved, and the accuracy of the original electrocardiogram data representation in the electrocardiogram data reconstruction process is further improved.
Please refer to a method flowchart corresponding to fig. 2 for the working process of each module provided in the embodiment of the present invention, and detailed description of the working process is omitted.
The embodiment of the invention discloses another reconstruction device of electrocardiographic data, please refer to fig. 6, which comprises:
a compressed data receiving module 401, a first judging module 402, a first reconstructing module 403, a second reconstructing module 404;
wherein the first reconstructing module 403 comprises:
a second calculation module 4031, configured to multiply the personalized overcomplete dictionary with a compression matrix to obtain a multiplication matrix, where the compression matrix is used to compress the original electrocardiographic data of the user;
a third calculation module 4032, configured to use the multiplication matrix as the sensing matrix, use the required number of atoms parameter as the first sparsity, use the compressed electrocardiographic data as a sampling vector, and calculate a sparse coefficient matrix of the electrocardiographic data by using the OMP algorithm;
a fourth calculation module 4033, configured to multiply the personalized overcomplete dictionary with the sparse coefficient matrix of the electrocardiographic data, so as to obtain the first reconstructed electrocardiographic data.
The second reconstructing module 404 is configured to, if the first determining module 402 determines that the personalized overcomplete dictionary is not generated, reconstruct the compressed electrocardiographic data by using an alternative overcomplete dictionary to obtain second reconstructed electrocardiographic data, where the alternative overcomplete dictionary is obtained in advance.
In the electrocardiographic data reconstruction device disclosed by the embodiment of the invention, the third calculation module 4032 is used for taking a multiplication matrix obtained by multiplying the personalized overcomplete dictionary by the compression matrix as a sensing matrix of the OMP algorithm, the number parameter of atoms required in the generation process of the personalized overcomplete dictionary is taken as the first sparsity of the OMP algorithm, and calculates a sparse coefficient matrix of the electrocardiographic data by combining with an OMP algorithm, then a fourth calculation module 4033 multiplies the sparse coefficient matrix with the personalized over-complete dictionary to obtain the restored first reconstructed electrocardiographic data, thereby introducing the personalized over-complete dictionary into the OMP algorithm in the reconstruction process of the electrocardiographic data to improve the reconstruction precision, meanwhile, when the second reconstruction module 404 is selected and the personalized overcomplete dictionary is not generated, the received compressed electrocardiographic data is reconstructed by using the preset alternative overcomplete dictionary, so that the time required by reconstruction can be effectively shortened.
Please refer to a method flowchart corresponding to fig. 3 for the working process of each module provided in the embodiment of the present invention, and detailed description of the working process is omitted.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (6)
1. A method for reconstructing electrocardiographic data, comprising:
receiving compressed electrocardiogram data of a user;
judging whether a personalized over-complete dictionary is generated, wherein the personalized over-complete dictionary is an over-complete dictionary generated by training according to original electrocardiogram data of the user, and the original electrocardiogram data are not compressed electrocardiogram data;
if the personalized overcomplete dictionary is generated, reconstructing the compressed electrocardiogram data by using the personalized overcomplete dictionary and an OMP algorithm to obtain first reconstructed electrocardiogram data;
the process of personalized overcomplete dictionary generation comprises:
receiving original electrocardiogram data of the user within a preset time;
grouping the original electrocardiogram data according to receiving time to obtain a first initialized over-complete dictionary and a first training sample;
performing matrix transformation on the first initialized overcomplete dictionary to generate a second initialized overcomplete dictionary;
generating a target iteration number corresponding to the second initialized overcomplete dictionary according to the matrix attribute of the second initialized overcomplete dictionary, wherein the matrix attribute is the number of rows and the number of columns of the second initialized overcomplete dictionary;
generating a required atom number parameter, wherein the required atom number parameter is the number of columns in the second initialized overcomplete dictionary required by the original electrocardiogram data;
performing matrix transformation on the first training sample to generate a second training sample;
taking the second initialized overcomplete dictionary as a sensing matrix, taking the required atom number parameter as a first sparsity, and calculating a sparse coefficient matrix of the second training sample by using the OMP algorithm, wherein the sensing matrix and the first sparsity are the sensing matrix and sparsity adopted in the OMP algorithm;
taking the needed atom number parameter as a second sparsity, and training the second initialized overcomplete dictionary, the sparse coefficient matrix of the second training sample and the second training sample by using a K-SVD algorithm to generate a training update dictionary, wherein the second sparsity is the sparsity adopted in the K-SVD algorithm;
recording the generation times of the training updated dictionary;
judging whether the generation times reach the target iteration times or not;
if the generation times reach the target iteration times, taking the training updated dictionary as the personalized overcomplete dictionary;
and if the generation times do not reach the target iteration times, taking the training updated dictionary as the second initialization overcomplete dictionary, returning the number parameter of the needed atoms as a second sparsity, and training the second initialization overcomplete dictionary, the sparse coefficient matrix of the second training sample and the second training sample by utilizing a K-SVD algorithm to generate the training updated dictionary.
2. The method of claim 1, wherein reconstructing the compressed electrocardiographic data using the personalized overcomplete dictionary and the OMP algorithm to obtain first reconstructed electrocardiographic data comprises:
multiplying the personalized overcomplete dictionary by a compression matrix to obtain a multiplication matrix, wherein the compression matrix is used for compressing the original electrocardiogram data of the user;
taking the multiplication matrix as the sensing matrix, taking the required atom number parameter as the first sparsity, taking the compressed electrocardiogram data as a sampling vector, and calculating a sparse coefficient matrix of the electrocardiogram data by utilizing the OMP algorithm;
and multiplying the personalized over-complete dictionary by the sparse coefficient matrix of the electrocardiogram data to obtain the first reconstructed electrocardiogram data.
3. The method of claim 1, further comprising, after determining that the personalized overcomplete dictionary is not generated:
and reconstructing the compressed electrocardiogram data by using the alternative over-complete dictionary to obtain second reconstructed electrocardiogram data, wherein the alternative over-complete dictionary is obtained in advance.
4. An apparatus for reconstructing electrocardiographic data, comprising:
the compressed data receiving module is used for receiving compressed electrocardio data of a user;
the first judgment module is used for judging whether a personalized overcomplete dictionary is generated or not, wherein the personalized overcomplete dictionary is an overcomplete dictionary generated by training according to original electrocardiogram data of the user, and the original electrocardiogram data are uncompressed electrocardiogram data;
the first reconstruction module is used for reconstructing the compressed electrocardiogram data by utilizing the personalized overcomplete dictionary and an OMP algorithm to obtain first reconstructed electrocardiogram data if the personalized overcomplete dictionary is generated;
the original data receiving module is used for receiving original electrocardio data of the user within preset time;
the grouping module is used for grouping the original electrocardiogram data according to receiving time to obtain a first initialized over-complete dictionary and a first training sample;
the first training module is used for training the first initialized overcomplete dictionary by using the first training sample to generate the personalized overcomplete dictionary;
the first training module comprises:
the first matrix conversion module is used for performing matrix conversion on the first initialized overcomplete dictionary to generate a second initialized overcomplete dictionary;
the iteration number generation module is used for generating a target iteration number corresponding to the second initialized overcomplete dictionary according to the matrix attribute of the second initialized overcomplete dictionary, wherein the matrix attribute is the row number and the column number of the second initialized overcomplete dictionary;
a required atom number parameter generating module, configured to generate a required atom number parameter, where the required atom number parameter is a number of columns in the second initialized overcomplete dictionary required for representing the original electrocardiographic data;
the second matrix conversion module is used for performing matrix conversion on the first training sample to generate a second training sample;
a first calculation module, configured to use the second initialized overcomplete dictionary as a sensing matrix, use the required atom number parameter as a first sparsity, and calculate a sparse coefficient matrix of the second training sample by using the OMP algorithm, where the sensing matrix and the first sparsity are the sensing matrix and sparsity adopted in the OMP algorithm;
the second training module is used for taking the needed atom number parameter as a second sparsity, training the second initialized overcomplete dictionary, the sparse coefficient matrix of the second training sample and the second training sample by utilizing a K-SVD algorithm, and generating a training updating dictionary, wherein the second sparsity is the sparsity adopted in the K-SVD algorithm;
the recording module is used for recording the generation times of the training updated dictionary;
the second judgment module is used for judging whether the generation times reach the target iteration times;
the selection module is used for taking the training updated dictionary as the personalized overcomplete dictionary if the generation times reach the target iteration times;
the updating module is used for taking the training updated dictionary as the second initialized overcomplete dictionary if the generation times do not reach the target iteration times;
the second training module is further configured to, after the updating module uses the training updated dictionary as the second initialized overcomplete dictionary, use the required atom number parameter as a second sparsity, and train the second initialized overcomplete dictionary, the sparse coefficient matrix of the second training sample, and the second training sample by using a K-SVD algorithm to generate the training updated dictionary, where the second sparsity is a sparsity adopted in the K-SVD algorithm.
5. The apparatus of claim 4, wherein the first reconstruction module comprises:
the second calculation module is used for multiplying the personalized over-complete dictionary by a compression matrix to obtain a multiplication matrix, wherein the compression matrix is used for compressing the original electrocardiogram data of the user;
the third calculation module is used for taking the multiplication matrix as the sensing matrix, taking the required atom number parameter as the first sparsity, taking the compressed electrocardio data as a sampling vector, and calculating a sparse coefficient matrix of the electrocardio data by utilizing the OMP algorithm;
and the fourth calculation module is used for multiplying the personalized over-complete dictionary by the sparse coefficient matrix of the electrocardiographic data to obtain the first reconstructed electrocardiographic data.
6. The apparatus of claim 4, further comprising:
and the second reconstruction module is used for reconstructing the compressed electrocardiogram data by using the alternative overcomplete dictionary to obtain second reconstructed electrocardiogram data if the first judgment module judges that the personalized overcomplete dictionary is not generated, wherein the alternative overcomplete dictionary is obtained in advance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710236488.1A CN108694409B (en) | 2017-04-12 | 2017-04-12 | Reconstruction method and device of electrocardiogram data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710236488.1A CN108694409B (en) | 2017-04-12 | 2017-04-12 | Reconstruction method and device of electrocardiogram data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108694409A CN108694409A (en) | 2018-10-23 |
| CN108694409B true CN108694409B (en) | 2022-04-19 |
Family
ID=63843650
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710236488.1A Active CN108694409B (en) | 2017-04-12 | 2017-04-12 | Reconstruction method and device of electrocardiogram data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108694409B (en) |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140037199A1 (en) * | 2005-04-04 | 2014-02-06 | Michal Aharon | System and method for designing of dictionaries for sparse representation |
| CN102867413A (en) * | 2012-07-18 | 2013-01-09 | 浙江工业大学 | Compressed sensing acquiring method for vehicle sensing data under vehicle-road coordination environment |
| CN103247034B (en) * | 2013-05-08 | 2016-01-20 | 中国科学院光电研究院 | A kind of compressed sensing high spectrum image reconstructing method based on sparse spectrum dictionary |
| CN103679762A (en) * | 2013-12-13 | 2014-03-26 | 江苏大学 | Ultrasonic signal reestablishing method based on sparse data |
| CN103944579B (en) * | 2014-04-10 | 2017-06-20 | 东华大学 | A kind of coding/decoding system of compressed sensing reconstruct |
| CN105631839A (en) * | 2014-09-08 | 2016-06-01 | 西南科技大学 | Super-resolution measuring method for particle thickness |
| CN104504015A (en) * | 2014-12-11 | 2015-04-08 | 中国科学院遥感与数字地球研究所 | Learning algorithm based on dynamic incremental dictionary update |
| US9667456B2 (en) * | 2015-08-05 | 2017-05-30 | National Taiwan University | Compressive sensing system based on personalized basis and method thereof |
| CN105827250B (en) * | 2016-03-16 | 2019-10-01 | 江苏大学 | A kind of power quality data compression reconfiguration method based on self-adapting dictionary study |
-
2017
- 2017-04-12 CN CN201710236488.1A patent/CN108694409B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108694409A (en) | 2018-10-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Rezaii et al. | ECG signal compression and denoising via optimum sparsity order selection in compressed sensing framework | |
| CN107647860B (en) | Heart rate detection method and device, electronic equipment and storage medium | |
| KR102240256B1 (en) | Signal sampling recovery method, apparatus and OvXDM system applied to OvXDM system | |
| Akhter et al. | ECG comptression using run length encoding | |
| WO2009094357A1 (en) | Parallel and adaptive signal processing | |
| CN105574322B (en) | Physiological parameter index operation system and method | |
| RU162110U1 (en) | ECG STORAGE AND ANALYSIS | |
| CN109284761A (en) | A kind of image characteristic extracting method, device, equipment and readable storage medium storing program for executing | |
| CN108685570B (en) | Method, device and system for processing over-complete dictionary | |
| CN108694409B (en) | Reconstruction method and device of electrocardiogram data | |
| CN110851317A (en) | Method, device, equipment and storage medium for predicting IOPS performance data of storage equipment | |
| Lai et al. | An efficient DCT-IV-based ECG compression algorithm and its hardware accelerator design | |
| CN116712048A (en) | Blood pressure measurement method and device and related equipment | |
| Petrellis et al. | Compressing and filtering medical data in a low cost health monitoring system | |
| Lai et al. | An efficient method of ECG signal compression by using a DCT-IV spectrum | |
| US11436205B2 (en) | Method and apparatus for predicting parameter values of data entries in data set | |
| Cao et al. | Real-Time Deep Compressed Sensing Reconstruction for Electrocardiogram Signals | |
| CN113693610A (en) | Few-lead electrocardiogram data processing method and device, storage medium and computer equipment | |
| Bhadoria et al. | Comparative analysis of basis & measurement matrices for non-speech audio signal using compressive sensing | |
| Moses et al. | A novel lossless ECG compression technique for transmission in GSM networks | |
| Olenko et al. | Average sampling restoration of harmonizable processes | |
| RU2595929C2 (en) | Method and apparatus for compressing data depending on time signal | |
| Premalatha et al. | Implementation of VLSI based efficient lossless EEG compression architecture using verilog HDL | |
| JP2021115429A (en) | Signal measuring system, measurement signal processing device and program | |
| CN212186496U (en) | Simulation system for real-time compression sampling reconstruction of electromyographic signals |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |