CN112270240A - Signal processing method and device, electronic equipment and storage medium - Google Patents
Signal processing method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN112270240A CN112270240A CN202011139874.7A CN202011139874A CN112270240A CN 112270240 A CN112270240 A CN 112270240A CN 202011139874 A CN202011139874 A CN 202011139874A CN 112270240 A CN112270240 A CN 112270240A
- Authority
- CN
- China
- Prior art keywords
- signal
- initial
- signals
- segmented
- determining
- 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
- 238000003672 processing method Methods 0.000 title claims abstract description 23
- 239000013598 vector Substances 0.000 claims abstract description 153
- 238000012545 processing Methods 0.000 claims abstract description 36
- 238000000034 method Methods 0.000 claims abstract description 22
- 230000011218 segmentation Effects 0.000 claims abstract description 19
- 230000015654 memory Effects 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 description 8
- 201000010099 disease Diseases 0.000 description 8
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 8
- 230000006870 function Effects 0.000 description 6
- 230000003287 optical effect Effects 0.000 description 6
- 208000019622 heart disease Diseases 0.000 description 5
- 238000004891 communication Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000007787 long-term memory Effects 0.000 description 3
- 230000006403 short-term memory Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 101000860173 Myxococcus xanthus C-factor Proteins 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012800 visualization Methods 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
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- 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
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Pathology (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Psychiatry (AREA)
- Veterinary Medicine (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Mathematical Physics (AREA)
- Heart & Thoracic Surgery (AREA)
- General Engineering & Computer Science (AREA)
- Physiology (AREA)
- General Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Fuzzy Systems (AREA)
- Software Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The embodiment of the invention discloses a signal processing method, a signal processing device, electronic equipment and a storage medium, wherein the signal identification method comprises the following steps: acquiring multi-channel initial signals, and respectively carrying out segmentation processing on each initial signal to obtain a segmented signal of each initial signal; extracting initial characteristic vectors and time interval characteristics of each segmented signal, and determining intermediate characteristic vectors based on the initial characteristic vectors and the time interval characteristics; combining intermediate eigenvectors corresponding to a first preset number of segmented signals in the multichannel initial signals based on the time sequence of each segmented signal to form combined eigenvectors; adding the time sequence identification and the corresponding channel identification of each segmented signal to the combined characteristic vector to obtain a target characteristic vector; and inputting the target feature vector into a self-attention model, and determining probability values of the multichannel initial signals belonging to each preset classification based on the output of the self-attention model. The space-time correlation between different signal segments can be established, the signal identification is enhanced, and the identification precision is improved.
Description
Technical Field
The embodiment of the invention relates to the field of medical health, in particular to a signal processing method, a signal processing device, electronic equipment and a storage medium.
Background
As heart diseases are receiving more and more attention, diagnosis of heart diseases is becoming important. Electrocardiographic recognition is a major method for diagnosing heart diseases, and the objective of electrocardiographic recognition is to judge the health condition of the heart of a patient or predict the type of a possible heart disease from an electrocardiogram signal.
In the prior art, an end-to-end model such as a Convolutional Neural Network (CNN) or a Recurrent Neural Network (RNN) is usually used, one is to input a signal of each lead into the corresponding model, output a characteristic value, combine characteristic values of a plurality of leads to judge and directly identify the signal of the electrocardiogram, and the other is to use all lead channels as a plurality of characteristics of an input data, enter a model and directly output a final identification result.
The prior art has at least the following disadvantages: although the ability of the CNN or RNN model to remember information at all positions decreases as the length of the signal increases, symptoms are easily recognized when they are expressed on adjacent signals, and are not easily recognized when they are expressed over time, and spatial-temporal correlation between different signal segments cannot be established, which weakens the recognition of the signal and reduces the recognition accuracy.
Disclosure of Invention
Embodiments of the present invention provide a signal processing method, an apparatus, an electronic device, and a storage medium, which can establish a time-space association between different signal segments of a signal, enhance signal identification, and improve identification accuracy.
In a first aspect, an embodiment of the present invention provides a signal processing method, including:
acquiring multi-channel initial signals, and respectively carrying out segmentation processing on each initial signal to obtain segmented signals of each initial signal;
extracting initial feature vectors and time interval features of the segmented signals, and determining intermediate feature vectors based on the initial feature vectors and the time interval features;
combining the intermediate eigenvectors corresponding to a first preset number of segmented signals in the multichannel initial signals based on the time sequence of each segmented signal to form combined eigenvectors;
adding the time sequence identification and the corresponding channel identification of each segmented signal into the combined feature vector to obtain a target feature vector;
and inputting the target feature vector into an attention model, and determining probability values of the multichannel initial signals belonging to each preset classification based on the output of each attention model.
In a second aspect, an embodiment of the present invention further provides a signal processing apparatus, including:
the signal segmentation module is used for acquiring multi-channel initial signals and respectively carrying out segmentation processing on the initial signals to obtain segmented signals of the initial signals;
the intermediate characteristic vector determining module is used for extracting an initial characteristic vector and a time interval characteristic of each segmented signal and determining an intermediate characteristic vector based on the initial characteristic vector and the time interval characteristic;
the feature vector combination module is used for combining the intermediate feature vectors corresponding to a first preset number of segmented signals in the multichannel initial signals based on the time sequence of each segmented signal to form combined feature vectors;
the target characteristic vector determining module is used for adding the time sequence identification and the corresponding channel identification of each segmented signal into the combined characteristic vector to obtain a target characteristic vector;
and the probability value determining module is used for inputting the target feature vector into an attention model and determining the probability value of the multichannel initial signal belonging to each preset classification based on the output of each attention model. .
In a third aspect, an embodiment of the present invention further provides an electronic device, where the electronic device includes:
one or more processors;
a memory for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement the signal processing method steps as provided by any of the embodiments of the invention.
In a fourth aspect, the embodiments of the present invention further provide a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the signal processing method provided in any embodiment of the present invention.
The technical scheme provided by the embodiment of the invention comprises the steps of obtaining multi-channel initial signals, respectively carrying out segmentation processing on each initial signal to obtain segmented signals of each initial signal, extracting initial feature vectors and time interval features of each segmented signal, determining intermediate feature vectors based on the initial feature vectors and the time interval features, and adding the intermediate feature vectors into the multi-channel initial signals based on the time sequence of each segmented signal, combining the intermediate eigenvectors corresponding to a first preset number of segmented signals to form a combined eigenvector, adding the timing identification and the corresponding channel identification of each segmented signal to the combined eigenvector, and obtaining a target feature vector, inputting the target feature vector into an attention model, and determining probability values of the multichannel initial signals belonging to each preset classification based on the output of each attention model. And the space-time correlation between different signal segments of the signal is established, so that the signal identification is enhanced, and the identification precision is improved.
Drawings
In order to illustrate the embodiments of the present invention or the technical solutions in the prior art more clearly, a brief description will be given below to the drawings required for the description of the embodiments or the prior art, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic flowchart of a signal processing method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a segment of an electrocardiogram signal according to an embodiment of the present invention;
fig. 3 is a schematic flowchart of a signal processing method according to a second embodiment of the present invention;
FIG. 4 is a schematic diagram of merging segmented signals of different channels of an electrocardiograph signal according to a second embodiment of the present invention;
FIG. 5 is a schematic flow chart of an ECG signal processing according to a second embodiment of the present invention;
fig. 6 is a block diagram of a signal processing apparatus according to a third embodiment of the present invention;
fig. 7 is a schematic structural diagram of an electronic device according to a fourth embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.
Example one
Fig. 1 is a schematic flow chart of a signal processing method according to an embodiment of the present invention, which is applicable to a case of determining a category to which a signal corresponds according to the signal, and is particularly applicable to a case of determining a disease type corresponding to an electrocardiographic signal according to the electrocardiographic signal. The method may be performed by a signal processing means, which may be implemented by means of software and/or hardware, integrated in a device having a control function, such as a computer. The method specifically comprises the following steps:
s110, acquiring multi-channel initial signals, and respectively performing segmentation processing on each initial signal to obtain segmented signals of each initial signal.
The method includes the steps of acquiring initial signals acquired by different signal acquisition channels, and performing segmentation processing on the initial signals of a certain signal length acquired by each channel according to signal characteristics, where the acquired initial signals of each channel all have multiple segmented signals, where the signal characteristics may be signal waveform characteristics or time interval characteristics corresponding to signal waveforms at different times.
Alternatively, the initial signals may be electrocardiogram signals, and the acquiring of the multichannel initial signals may be acquiring 8-lead or 12-lead electrocardiogram signals, wherein 8-lead and 12-lead electrocardiogram signals are acquired respectively for 8-channel and 12-channel electrocardiogram signals.
Optionally, the segmenting processing is performed on each initial signal respectively, and includes: for any initial signal, identifying a first type of wave in the initial signal;
and determining each section time according to the time of each first type wave and the first time interval, and determining a section signal based on any two adjacent section times.
When the initial signal is segmented, the first type wave in the signal waveform is identified, the time for segmenting the signal is determined according to the first type wave at different times and a first time interval which is separated from the first type wave by a certain time period, and the signal between two adjacent segmented times is taken as a segmented signal according to the determined time for segmenting the signal.
Using the first type of wave of the initial signal and the first time interval, the segmented signal may be determined from signal characteristics of the initial signal.
Illustratively, fig. 2 is a schematic illustration of a segmentation of an electrocardiogram signal, see fig. 2, the electrocardiogram signal being segmented according to heart beat activity: with the R wave crest of each heartbeat of a channel of a lead I (recording the voltage difference between electrodes of a left arm and a right arm) as a reference point, signal data in a time window range from 200ms before to 200ms before the next R wave crest forms a time window, namely a segmented signal, and 1-dimensional signals of all channels are divided according to the time window. Each segment of each channel is an independent unit, so that an electrocardiogram signal is roughly divided into independent units of 'number of leads x number of heart beats', and the length of each segment is variable.
S120, extracting initial feature vectors and time interval features of the segmented signals, and determining intermediate feature vectors based on the initial feature vectors and the time interval features.
The initial feature vector is a feature vector obtained by extracting unstructured features from the segmented signals determined in step S110, wherein the segmented signals of each channel are respectively determined to correspond to the initial feature vector. Alternatively, a neural network model may be used, and the initial feature vector of the segmented signal of each channel may be determined by using the model output by inputting the segmented signal of each channel into the neural network model, respectively. The time interval characteristic may be characteristic data of a time interval determined from a signal waveform characteristic.
And determining a new feature vector, which can be called a first feature vector, by using the extracted initial feature vector of the segmented signal and the time interval feature determined according to the waveform of the segmented signal.
S130, combining the intermediate feature vectors corresponding to the first preset number of segmented signals in the multichannel initial signals based on the time sequence of each segmented signal to form combined feature vectors.
After the initial signals of different channels are segmented to determine segmented signals, according to the time sequence of each segmented signal, combining the intermediate eigenvectors determined by using the initial eigenvectors and the time interval characteristics in step S120 corresponding to the first preset number of segmented signals to form a combined eigenvector, wherein the combination of the eigenvectors retains all data of the original eigenvectors, and may be only formal combination.
S140, adding the time sequence identification and the corresponding channel identification of each segmented signal into the combined feature vector to obtain a target feature vector.
The time sequence identifier is corresponding position information that is determined by the segmented signals in time sequence in the first preset number of segmented signals, and the value can be position information/the first preset number. The channel identification refers to the number of signal channels for acquiring the initial signal of the segmented signal. In order to retain the time position and the signal channel information of each segmented signal, the time identifier and the channel identifier corresponding to each segmented signal are added to the combined feature vector, optionally, the signal identifier and the channel identifier may be added to the first feature vector corresponding to the segmented signal, and then the first feature vectors corresponding to the first preset number of segmented signals of a single channel are combined to determine the combined feature vector.
S150, inputting the target feature vector into the attention model, and determining probability values of the multichannel initial signals belonging to each preset classification based on the output of each attention model.
The preset classification refers to one or more signal types to which the initial signal may belong, which are judged in advance, wherein the initial signal may belong to one of the classes of the preset classification, may belong to multiple classes of the preset classification, or may not belong to any one of the classes of the preset classification. The target feature vectors determined in step S140 are respectively input into the self-attention models, a probability value that the multi-channel initial signal belongs to a preset classification is determined according to the output of each attention model, and whether the initial signal belongs to one of the preset classification categories, or multiple categories, or does not belong to the preset classification category is correspondingly determined according to the probability value.
Illustratively, the first predetermined number is 8, and all segmented signals contained in 8 heartbeats are a "large segment" D as an input from the attention model, so that D is composed of "8 × N _ C" units, where N _ C is the number of channels. And establishing association among the objects in D by adopting a self-attention network model among the characteristics F of each segmented signal of each D to form a time-crossing and space-crossing characteristic information communication network and form high-level unit object characteristics, determining corresponding key, query and value according to model input by the self-attention model, and adjusting related parameters.
Optionally, the training method of the self-attention model includes: acquiring a multi-channel sample signal and at least one label corresponding to the sample signal; determining a target characteristic vector corresponding to the sample signal, inputting the target characteristic vector into a self-attention model to be trained, and determining probability values of sample information belonging to each preset classification; and forming a loss function based on the probability value of the preset classification and at least one label corresponding to the sample signal, and performing parameter adjustment on the self-attention model to be trained based on the loss function until the self-attention model meets the training adjustment to obtain the trained self-attention model.
When the initial signal is an electrocardiosignal, multi-label classification can be carried out on the electrocardiosignal, wherein each label corresponds to a disease, a prediction result of whether the disease exists is output, the prediction of different diseases is independent, information on different time periods and different channels of the electrocardiosignal is related through a signal model, and the accuracy of extracting information of each local part is enhanced through dynamic sharing of the information, so that the overall recognition is enhanced.
The overall model formed by combining all parts can adopt the electrocardio type disease label marked manually to carry out end-to-end learning, the type of the label is not limited, but the disease type meeting the clinical standard is adopted in the ideal situation. The model can be trained with data of different lengths during the learning phase, since it is automatically divided into a preset number of "large segments". Data of different time lengths can also be processed after learning is completed.
The trained model, the attention module of which provides a basis for visualization and interpretability. When a section of test electrocardiogram signals are input, the dynamically changed attention value can reflect the mutual connection between different channels and different heartbeats, and the judgment conclusion of the output can be expressed according to the signals of the space-time positions.
In the technical solution of this embodiment, a multichannel initial signal is obtained, each initial signal is segmented to obtain a segmented signal of each initial signal, an initial feature vector and a time interval feature of each segmented signal are extracted, an intermediate feature vector is determined based on the initial feature vector and the time interval feature, and in the multichannel initial signal, based on a time sequence of each segmented signal, combining the intermediate eigenvectors corresponding to a first preset number of segmented signals to form a combined eigenvector, adding the timing identification and the corresponding channel identification of each segmented signal to the combined eigenvector, and obtaining a target feature vector, inputting the target feature vector into an attention model, and determining probability values of the multichannel initial signals belonging to each preset classification based on the output of each attention model. And the space-time correlation between different signal segments of the signal is established, so that the signal identification is enhanced, and the identification precision is improved.
Example two
Fig. 3 is a schematic flow chart of a signal processing method according to a second embodiment of the present invention, which is further optimized based on the second embodiment, specifically, the steps of determining intermediate feature vectors, determining combined feature vectors, and determining probability values of the multi-channel initial signal belonging to each preset classification based on outputs of respective attention models are refined. Wherein explanations of the same or corresponding terms as those of the above-described embodiments are omitted.
Referring to fig. 3, the signal processing method provided in this embodiment specifically includes the following steps:
s201, acquiring multi-channel initial signals, and respectively performing segmentation processing on each initial signal to obtain segmented signals of each initial signal.
S202, inputting the segmented signals into the long-term and short-term memory model corresponding to the channel to which the segmented signals belong, and determining initial characteristic vectors based on model output.
The segmented signals of each channel determine feature vectors through a Long-Short Term Memory (LSTM) model, the segmented signals are input into the LSTM model corresponding to the channel to which the segmented signals belong, and the feature vectors output by the model are used as initial feature vectors. The segmented signals are input into the LSTM in a signal waveform mode, output are feature vectors with fixed dimensions containing segmented signal features, each channel is provided with an independent LSTM model, and parameters are not shared among the channels.
S203, determining time interval characteristics based on the time interval between at least one type of wave in the segmented signals.
The method comprises the steps that different types of waves can appear in signal waveforms corresponding to different moments of a segmented signal, and time interval characteristics between two waves of at least one type are determined according to the time interval of the two waves in the segmented signal interval, wherein the time interval characteristics refer to characteristic data determined according to the time interval between the waves.
Optionally, the time interval characteristic includes at least one of: r wave-R wave interval characteristics, R wave-P wave interval characteristics and S wave-T wave interval characteristics.
When the initial signal is an electrocardiogram signal, a segmented signal, namely, P waves, Q waves, R waves, S waves, T waves and U waves exist between electrocardiogram signals corresponding to one heartbeat, R wave-R wave time intervals, R wave-P wave time intervals and S wave-T wave time intervals are determined, corresponding time interval characteristics are calculated according to the respectively determined time intervals, and the time interval characteristics can be data for identifying the time intervals.
And S204, adding the time interval features to the initial feature vector, and determining a middle feature vector.
The time interval signature is added after the initial signature vector of the LSTM output, forming an intermediate signature vector of the segmented signal.
Illustratively, for the electrocardiosignal, if the dimension of an initial feature vector corresponding to a segmented signal output by the LSTM is F _ F, after the R wave-R wave interval feature, the R wave-P wave interval feature and the S wave-T wave interval feature of the segmented signal are added to the initial feature vector, the length of an intermediate feature vector corresponding to the determined segmented signal is F _ F + 3.
S205, corresponding to any channel, merging the segmented signals of a first preset number based on the time sequence of the segmented signals to form middle segment signals, and combining the middle characteristic vectors of the merged segmented signals.
For the segmented signals of any channel, according to the corresponding time sequence, combining the segmented signals of a first preset number, forming middle segment signals by combined signal segments, and simultaneously combining the middle characteristic vectors which are determined by the LSTM and correspond to the segmented signals combined in the same middle segment signal.
Optionally, a second preset number of pieces of overlapping segment information exist in the previous middle segment signal and the subsequent middle segment signal, where the second preset number is smaller than the first preset number.
Exemplarily, fig. 4 is a schematic diagram of merging segmented signals of different channels of an electrocardiograph signal, referring to fig. 4, where a first preset number is 8, that is, all segmented signals F included in 8 heartbeats are a "large segment" D, and optionally, a second preset number may be half of the first preset number, that is, the second preset number is 4, that is, 4 heartbeats overlap exists between every 8 heartbeats, that is, a scheme of overlapping adjacent D and F in time by 50% is adopted for dividing the adjacent D and F.
The signals of the adjacent middle sections have overlapping areas, so that the space-time correlation between different signal sections is established conveniently.
And S206, carrying out secondary combination on the combined intermediate eigenvectors corresponding to the middle section signals of the multiple channels to obtain a combined eigenvector.
After the intermediate eigenvectors of each segmented signal of the intermediate section signal of each channel are combined, the intermediate eigenvector corresponding to the intermediate section signal of each channel is determined, and then the intermediate eigenvectors of each channel are also combined to obtain the combined eigenvector corresponding to the intermediate section signal of a plurality of channels.
And S207, adding the time sequence identification and the corresponding channel identification of each segmented signal into the combined feature vector to obtain a target feature vector.
And S208, inputting the target feature vector into the attention model, acquiring feature results output by the attention model, and calling weight information and calculation strategies corresponding to preset classifications.
The preset classification refers to the type of the signal, optionally, the signal is an electrocardiosignal, and the preset classification is an electrocardio condition type, and whether the heart disease corresponding to the electrocardio condition type exists or not can be judged according to the electrocardio condition type.
Inputting the determined target feature vector into the attention model, and obtaining an output feature result as H, wherein H is a vector of a dimension F _ H multiplied by 1, whether the signal belongs to different preset classifications is determined by calling different weight information and calculation strategies from the attention model, and when the feature output H corresponding to each preset classification is determined, the model is output by calling the weight information and the calculation strategies corresponding to the attention model.
For example, when the target feature vector is input into the attention model, a sequence of 8 × N _ C units rearranged into 1 dimension is regarded as each input token of the attention model, where N _ C signal channel number is, but in order to preserve the time position and channel information of the token, 2 pieces of indication information are additionally added to the initial feature vector F of each token: a single scalar quantity is used for representing the position in the sequence with the length of 8, and the index is obtained by using index/8, wherein the index represents the position information of the segmented signals corresponding to the time sequence in the first preset number of the segmented signals; and another one-hot coding adopting NcX 1 is adopted to represent the channel, and then an intermediate feature vector is determined, wherein the length of the intermediate feature vector is F _ F + 3. Thus, the target vector length of the true input token is F _ F +3+1+ N _ C.
S209, determining the initial probability values of the feature results output by the respective attention models belonging to the preset classification based on the weight information and the calculation strategy.
Determining the initial probability r corresponding to the preset classification c by the characteristic result H output from the attention model through the following formulac：
rc＝sigmoid(Wc,2relu(Wc,1H))
Wherein, W _ c,2, W _ c,1 are weight matrixes of 1 × 10 and 10 × F _ H, respectively, and subscript c represents a preset type to which the signal belongs, which may be a specific electrocardiogram status type.
For a plurality of preset classifications, the initial probability expressions of all the corresponding preset classifications are as follows:
r＝[rc]C
r is formed by C rcWhere C represents the total number of types to which the signal belongs, e.g., C may be the total number of types of electrocardiographic conditions detected.
Optionally, all the high-level features H assemble the final multi-label classification result through an additional multi-layer fully-connected network, and the number of categories of the final result is determined according to the setting of the task.
S210, obtaining probability values of the multichannel initial signals belonging to each preset classification based on probability statistical strategies and probability values corresponding to the characteristic results.
Obtaining an output r of a middle segment signal composed of a first preset number of segment signals on the basis of a feature result output from the attention model1、r2… …, which are all vectors of length C, are merged into a data R of shape C N D, where N D is the number of signals in all intermediate segments, and for signal path i, the combined final output RiIs also a vector with the length of C, and the calculation formula is as follows:
namely, the maximum probability value of the initial signal for the corresponding preset type is determined among the N _ D intermediate segment signals, and the probability value of the initial signal belonging to each preset classification is determined.
Optionally, strategies such as calculating a mean value, a mean square error and the like of each intermediate segment signal for a preset classification may be adopted according to the output from the attention model.
Exemplarily, fig. 5 is a schematic flow chart of the electrocardiograph signal processing, and referring to fig. 5, the electrocardiograph signal processing process includes:
collecting electrocardiogram data of different channels, wherein the common medical electrocardiogram is 8-lead or 12-lead;
the acquired electrocardiogram is led into a processing module, signal segmentation is carried out by utilizing time windows, a segmented signal in each time window is determined, and the total length of the signal generally exceeds one large segment;
sequentially calculating the results of each large section by a window moving method, and combining a plurality of results into one by a multi-window combination method;
inputting the segmented signals in the windows corresponding to different time intervals into an LSTM model to determine initial characteristic vectors, and extracting corresponding characteristic data according to the segmented signals in the American window;
performing total segmentation on the electrocardiosignal, taking all segmented signals F contained in 8 heartbeats as a 'large segment' D as the input of a self-attention model, wherein D consists of '8 multiplied by N _ C' units, and N _ C is the number of lead channels;
the feature after self-attention model processing is H, and the shape is: the vector of F _ H multiplied by 1 becomes output r by the formula;
all the high-level features H are combined into a final multi-label classification result through an additional multi-layer full-connection network.
End-to-end electrocardio disease type identification is realized through a new network model, and the network consists of a plurality of parts, so that the length of a signal processed by each part is limited in a small range, the length of a heartbeat interval is about 1s, and the identification precision is improved. And outputting LSTM models of different processing units, establishing space-time association between different channels and different signal sections based on a self-attention model, and ensuring the associated dynamics by the attention model to adapt to different types of changes. The segmentation of the signal is dynamic, and the time interval characteristic calculated by the traditional method of the signal is added into the characteristic by taking each heartbeat activity as a reference time window, so that the signal identification is enhanced. In the electrocardiosignal processing process, the unit space-time relationship is quantitative, and has better interpretability compared with a common model.
The technical solution of this embodiment is to obtain multi-channel initial signals, respectively perform segment processing on each initial signal to obtain segment signals of each initial signal, input the segment signals into a long-term and short-term memory model corresponding to a channel to which the initial signals belong, determine an initial feature vector based on model output, determine a time interval feature based on a time interval between at least one type of waves in the segment signals, add the time interval feature to the initial feature vector, determine an intermediate feature vector, combine a first preset number of segment signals based on a time sequence of the segment signals corresponding to any channel to form an intermediate segment signal, combine the intermediate feature vectors of the combined segment signals, perform secondary combination on the combined intermediate feature vectors corresponding to the multi-channel intermediate segment signals to obtain a combined feature vector, add a time sequence identifier and a corresponding channel identifier of each segment signal to the combined feature vector, the method comprises the steps of obtaining target feature vectors, inputting the target feature vectors into attention models, obtaining feature results output by the attention models, calling weight information and a calculation strategy corresponding to preset classifications, determining initial probability values of the feature results output by the attention models belonging to the preset classifications based on the weight information and the calculation strategy, obtaining probability values of multichannel initial signals belonging to the preset classifications based on probability statistics strategies and probability values corresponding to the feature results, establishing space-time association among different signal segments of the signals, enhancing the identification of the signals and improving the identification precision.
EXAMPLE III
Fig. 6 is a block diagram of a data processing apparatus according to a third embodiment of the present invention, which is applicable to a case where a category to which a signal corresponds is determined based on the signal, and in particular, a case where a disease type corresponding to an electrocardiographic signal is determined based on the electrocardiographic signal. The signal processing method provided by any embodiment of the invention can be realized by applying a signal processing device. As shown in fig. 6, the signal processing apparatus includes:
the signal segmentation module 310 is configured to obtain multi-channel initial signals, and perform segmentation processing on each initial signal to obtain a segmented signal of each initial signal;
an intermediate feature vector determining module 320, configured to extract an initial feature vector and a time interval feature of each segmented signal, and determine an intermediate feature vector based on the initial feature vector and the time interval feature;
the feature vector combination module 330 is configured to combine intermediate feature vectors corresponding to a first preset number of segmented signals in the multichannel initial signal based on a time sequence of each segmented signal to form a combined feature vector;
a target feature vector determining module 340, configured to add the timing identifier and the corresponding channel identifier of each segmented signal to the combined feature vector to obtain a target feature vector;
a probability value determining module 350, configured to input the target feature vector into the attention model, and determine, based on an output of each attention model, a probability value that the multichannel initial signal belongs to each preset classification. .
Optionally, the segmenting processing is performed on each initial signal respectively, and includes:
for any initial signal, identifying a first type of wave in the initial signal;
and determining each section time according to the time of each first type wave and the first time interval, and determining a section signal based on any two adjacent section times.
Specifically, the intermediate feature vector determining module 320 includes:
the initial characteristic vector determining unit is used for inputting the segmented signals into the long-term and short-term memory model corresponding to the channel to which the segmented signals belong and determining the initial characteristic vector based on the model output;
a time interval characteristic determination unit, which is used for determining the time interval characteristic based on the time interval between at least one type of wave in the segmented signal;
and the intermediate feature vector determining unit is used for adding the time interval features to the initial feature vector and determining an intermediate feature vector.
And the positioning mode determining unit is used for determining the RTK positioning mode as the target positioning mode if the communication quality of the RTK in the drop point setting area meets the setting condition, and otherwise, determining the target positioning mode according to the light intensity in the drop point setting area.
Optionally, the time interval characteristic includes at least one of: r wave-R wave interval characteristics, R wave-P wave interval characteristics and S wave-T wave interval characteristics.
Specifically, the feature vector combining module 330 includes:
the first feature vector combination unit is used for corresponding to any channel, combining the segmented signals of a first preset number based on the time sequence of the segmented signals to form middle segment signals, and combining the middle feature vectors of the combined segmented signals.
And the second feature vector combination unit is used for carrying out secondary combination on the combined intermediate feature vectors corresponding to the intermediate section signals of the multiple channels to obtain a combined feature vector.
Optionally, a second preset number of pieces of overlapping segment information exist in the previous middle segment signal and the subsequent middle segment signal, where the second preset number is smaller than the first preset number.
Specifically, the probability value determining module 350 includes:
and the calling unit is used for acquiring the feature results output by the respective attention models and calling the weight information and the calculation strategy corresponding to each preset classification.
The initial probability determining unit is used for determining the initial probability value of the feature result output by each attention model belonging to the preset classification based on the weight information and the calculation strategy;
and the preset classification probability value determining unit is used for obtaining the probability value of the multichannel initial signal belonging to each preset classification based on the probability statistical strategy and the probability value corresponding to each characteristic result.
The signal processing device provided by the embodiment of the invention can execute the signal processing method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects for executing the signal processing method. For details of the signal processing method provided in any embodiment of the present invention, reference may be made to the following description.
EXAMPLE seven
Fig. 7 is a schematic structural diagram of an electronic device according to a seventh embodiment of the present invention. FIG. 7 illustrates a block diagram of an exemplary electronic device 12 suitable for use in implementing embodiments of the present invention. The electronic device 12 shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiment of the present invention.
As shown in FIG. 7, electronic device 12 is embodied in the form of a general purpose computing device. The components of electronic device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that couples various system components including the system memory 28 and the processing unit 16.
The system memory 28 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM)30 and/or cache memory 32. The electronic device 12 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 7, and commonly referred to as a "hard drive"). Although not shown in FIG. 7, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 by one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the invention.
A program/utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. Program modules 42 generally carry out the functions and/or methodologies of the described embodiments of the invention.
The processing unit 16 executes various functional applications by executing programs stored in the system memory 28, for example, to implement a signal processing method step provided by the embodiment of the present invention, the method including:
acquiring multi-channel initial signals, and respectively carrying out segmentation processing on each initial signal to obtain segmented signals of each initial signal;
extracting initial feature vectors and time interval features of the segmented signals, and determining intermediate feature vectors based on the initial feature vectors and the time interval features;
combining the intermediate eigenvectors corresponding to a first preset number of segmented signals in the multichannel initial signals based on the time sequence of each segmented signal to form combined eigenvectors;
adding the time sequence identification and the corresponding channel identification of each segmented signal into the combined feature vector to obtain a target feature vector;
and inputting the target feature vector into an attention model, and determining probability values of the multichannel initial signals belonging to each preset classification based on the output of each attention model.
Of course, those skilled in the art can understand that the processor can also implement the technical solution of the signal processing method provided by any embodiment of the present invention.
Example four
A fourth embodiment provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps of a signal processing method as provided by any of the embodiments of the invention, the method comprising:
acquiring multi-channel initial signals, and respectively carrying out segmentation processing on each initial signal to obtain segmented signals of each initial signal;
extracting initial feature vectors and time interval features of the segmented signals, and determining intermediate feature vectors based on the initial feature vectors and the time interval features;
combining the intermediate eigenvectors corresponding to a first preset number of segmented signals in the multichannel initial signals based on the time sequence of each segmented signal to form combined eigenvectors;
adding the time sequence identification and the corresponding channel identification of each segmented signal into the combined feature vector to obtain a target feature vector;
and inputting the target feature vector into an attention model, and determining probability values of the multichannel initial signals belonging to each preset classification based on the output of each attention model.
Computer storage media for embodiments of the invention may employ any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. The computer-readable storage medium may be, for example but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
It will be understood by those skilled in the art that the modules or steps of the invention described above may be implemented by a general purpose computing device, they may be centralized on a single computing device or distributed across a network of computing devices, and optionally they may be implemented by program code executable by a computing device, such that it may be stored in a memory device and executed by a computing device, or it may be separately fabricated into various integrated circuit modules, or it may be fabricated by fabricating a plurality of modules or steps thereof into a single integrated circuit module. Thus, the present invention is not limited to any specific combination of hardware and software.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (10)
1. A signal processing method, comprising:
acquiring multi-channel initial signals, and respectively carrying out segmentation processing on each initial signal to obtain segmented signals of each initial signal;
extracting initial feature vectors and time interval features of the segmented signals, and determining intermediate feature vectors based on the initial feature vectors and the time interval features;
combining the intermediate eigenvectors corresponding to a first preset number of segmented signals in the multichannel initial signals based on the time sequence of each segmented signal to form combined eigenvectors;
adding the time sequence identification and the corresponding channel identification of each segmented signal into the combined feature vector to obtain a target feature vector;
and inputting the target feature vector into an attention model, and determining probability values of the multichannel initial signals belonging to each preset classification based on the output of each attention model.
2. The method of claim 1, wherein the separately segmenting each of the initial signals comprises:
for any initial signal, identifying a first type of wave in the initial signal;
and determining each section time according to the time of each first type wave and the first time interval, and determining a section signal based on any two adjacent section times.
3. The method of claim 1, wherein the extracting initial feature vectors and time interval features of each of the segmented signals, and determining intermediate feature vectors based on the first feature vectors and the time interval features comprises:
inputting the segmented signals into a long-short term memory model corresponding to the channel to which the segmented signals belong, and determining the initial characteristic vector based on model output;
determining the time interval characteristic based on the time interval between at least one type of wave in the segmented signal;
adding the time interval features to the initial feature vector, determining the intermediate feature vector.
4. The method of claim 3, wherein the time interval characteristic comprises at least one of: r wave-R wave interval characteristics, R wave-P wave interval characteristics and S wave-T wave interval characteristics.
5. The method of claim 1, wherein the combining the intermediate eigenvectors corresponding to a first preset number of segmented signals in the multichannel initial signal based on the timing of each segmented signal to form a combined eigenvector comprises:
corresponding to any channel, combining a preset number of segmented signals based on the time sequence of the segmented signals to form middle segment signals, and combining the middle eigenvectors of the combined segmented signals;
and carrying out secondary combination on the combined intermediate eigenvectors corresponding to the intermediate section signals of the multiple channels to obtain a combined eigenvector.
6. The method of claim 5, wherein there is a second predetermined number of overlapping segment information for the previous mid-segment signal and the subsequent mid-segment signal, wherein the second predetermined number is less than the first predetermined number.
7. The method of claim 1, wherein determining probability values for the multi-channel initial signal belonging to respective preset classifications based on outputs of respective attention models comprises:
acquiring feature results output by the self-attention models, and calling weight information and calculation strategies corresponding to preset classifications;
determining initial probability values of feature results output by the respective attention models, which belong to the preset classification, based on the weight information and a calculation strategy;
and obtaining the probability value of the multichannel initial signal belonging to each preset classification based on the probability statistical strategy and the probability value corresponding to each characteristic result.
8. A signal processing apparatus, characterized by comprising:
the signal segmentation module is used for acquiring multi-channel initial signals and respectively carrying out segmentation processing on the initial signals to obtain segmented signals of the initial signals;
the intermediate characteristic vector determining module is used for extracting an initial characteristic vector and a time interval characteristic of each segmented signal and determining an intermediate characteristic vector based on the initial characteristic vector and the time interval characteristic;
the feature vector combination module is used for combining the intermediate feature vectors corresponding to a first preset number of segmented signals in the multichannel initial signals based on the time sequence of each segmented signal to form combined feature vectors;
the target characteristic vector determining module is used for adding the time sequence identification and the corresponding channel identification of each segmented signal into the combined characteristic vector to obtain a target characteristic vector;
and the probability value determining module is used for inputting the target feature vector into an attention model and determining the probability value of the multichannel initial signal belonging to each preset classification based on the output of each attention model.
9. An electronic device, characterized in that the electronic device comprises:
one or more processors;
a memory for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement the signal processing method steps of any of claims 1-7.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the signal processing method steps of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011139874.7A CN112270240B (en) | 2020-10-22 | 2020-10-22 | Signal processing method, device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011139874.7A CN112270240B (en) | 2020-10-22 | 2020-10-22 | Signal processing method, device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112270240A true CN112270240A (en) | 2021-01-26 |
| CN112270240B CN112270240B (en) | 2024-02-09 |
Family
ID=74341744
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011139874.7A Active CN112270240B (en) | 2020-10-22 | 2020-10-22 | Signal processing method, device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112270240B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113707309A (en) * | 2021-08-31 | 2021-11-26 | 平安科技（深圳）有限公司 | Disease prediction method and device based on machine learning |
| CN114330431A (en) * | 2021-12-23 | 2022-04-12 | 天津工业大学 | Carrier roller fault diagnosis method based on dynamic self-attention and related equipment |
| CN116361668A (en) * | 2023-06-02 | 2023-06-30 | 北京安天网络安全技术有限公司 | Monitoring method, device, equipment and medium for multiple SDR equipment |
| WO2022179645A3 (en) * | 2022-06-13 | 2023-09-21 | 合肥心之声健康科技有限公司 | Electrocardiogram analysis method and apparatus, electronic device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110125043A1 (en) * | 2009-11-09 | 2011-05-26 | Biotronik Crm Patent Ag | Device, Method, and Computer-Readable Storage Medium for Detecting Events in Cardiac Signals |
| US10045710B2 (en) * | 2016-03-30 | 2018-08-14 | Medtronic, Inc. | Atrial arrhythmia episode detection in a cardiac medical device |
| CN109871742A (en) * | 2018-12-29 | 2019-06-11 | 安徽心之声医疗科技有限公司 | A kind of electrocardiosignal localization method based on attention Recognition with Recurrent Neural Network |
| CN111184508A (en) * | 2020-01-19 | 2020-05-22 | 武汉大学 | Electrocardiosignal detection device and analysis method based on joint neural network |
| CN111358459A (en) * | 2020-02-11 | 2020-07-03 | 广州视源电子科技股份有限公司 | Arrhythmia identification method, device, equipment and storage medium |
-
2020
- 2020-10-22 CN CN202011139874.7A patent/CN112270240B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110125043A1 (en) * | 2009-11-09 | 2011-05-26 | Biotronik Crm Patent Ag | Device, Method, and Computer-Readable Storage Medium for Detecting Events in Cardiac Signals |
| US10045710B2 (en) * | 2016-03-30 | 2018-08-14 | Medtronic, Inc. | Atrial arrhythmia episode detection in a cardiac medical device |
| CN109871742A (en) * | 2018-12-29 | 2019-06-11 | 安徽心之声医疗科技有限公司 | A kind of electrocardiosignal localization method based on attention Recognition with Recurrent Neural Network |
| CN111184508A (en) * | 2020-01-19 | 2020-05-22 | 武汉大学 | Electrocardiosignal detection device and analysis method based on joint neural network |
| CN111358459A (en) * | 2020-02-11 | 2020-07-03 | 广州视源电子科技股份有限公司 | Arrhythmia identification method, device, equipment and storage medium |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113707309A (en) * | 2021-08-31 | 2021-11-26 | 平安科技（深圳）有限公司 | Disease prediction method and device based on machine learning |
| CN114330431A (en) * | 2021-12-23 | 2022-04-12 | 天津工业大学 | Carrier roller fault diagnosis method based on dynamic self-attention and related equipment |
| WO2022179645A3 (en) * | 2022-06-13 | 2023-09-21 | 合肥心之声健康科技有限公司 | Electrocardiogram analysis method and apparatus, electronic device and storage medium |
| CN116361668A (en) * | 2023-06-02 | 2023-06-30 | 北京安天网络安全技术有限公司 | Monitoring method, device, equipment and medium for multiple SDR equipment |
| CN116361668B (en) * | 2023-06-02 | 2023-08-11 | 北京安天网络安全技术有限公司 | Monitoring method, device, equipment and medium for multiple SDR equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112270240B (en) | 2024-02-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112270240B (en) | Signal processing method, device, electronic equipment and storage medium | |
| CN108875575B (en) | Automatic arrhythmia analysis method based on multi-channel signal fusion deep neural network | |
| EP3777674A1 (en) | Time series data learning and analyzing method using artificial intelligence | |
| US20200211706A1 (en) | Intelligent traditional chinese medicine diagnosis method, system and traditional chinese medicine system | |
| CN108937912B (en) | Automatic arrhythmia analysis method based on deep neural network | |
| CN111772619B (en) | Heart beat identification method based on deep learning, terminal equipment and storage medium | |
| CN111160139A (en) | Electrocardiosignal processing method and device and terminal equipment | |
| CN111956212B (en) | Inter-group atrial fibrillation recognition method based on frequency domain filtering-multi-mode deep neural network | |
| US20230352164A1 (en) | Method for generating prediction result for predicting occurrence of fatal symptoms of subject in advance and device using same | |
| CN108766557B (en) | Automatic arrhythmia analysis method based on channel signal fusion neural network | |
| CN109431492B (en) | Electrocardiogram lead data simulation reconstruction method based on neural network algorithm | |
| CN111345816B (en) | Multi-lead QRS complex detection method, device, equipment and storage medium | |
| KR102483693B1 (en) | Method and apparatus of explainable multi electrocardiogram arrhythmia diagnosis | |
| CN115470828A (en) | Multi-lead electrocardiogram classification and identification method based on convolution and self-attention mechanism | |
| CN112690802B (en) | Method, device, terminal and storage medium for detecting electrocardiosignals | |
| CN111652135A (en) | Electrocardiogram data generation method and device and electronic equipment | |
| CN111358460A (en) | Arrhythmia identification method and device and electronic equipment | |
| CN113693611B (en) | Machine learning-based electrocardiogram data classification method and device | |
| CN112450944B (en) | Label correlation guide feature fusion electrocardiogram multi-classification prediction system and method | |
| CN117238510B (en) | Sepsis prediction method and system based on deep learning | |
| Tao et al. | A resource-efficient ECG diagnosis model for mobile health devices | |
| CN110786847B (en) | Electrocardiogram signal library building method and analysis method | |
| CN112989971B (en) | Electrocardiogram data fusion method and device for different data sources | |
| CN114190889A (en) | Electrocardiosignal classification method and system, electronic equipment and readable storage medium | |
| CN117547283B (en) | Electrocardiosignal data classification method, device, computer equipment and medium |
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 |