Signal processing method, electronic equipment and computer readable storage medium
By filtering and dynamic baseline screening of the ECG signals, the noise interference problem is solved, the accurate analysis of the ECG signals is achieved, and the signal quality and classification model processing effect is improved.
Patent Information
- Application Number
- CN202310741849.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-20
- Publication Date
- 2024-12-20
AI Technical Summary
During the acquisition process, electrocardiogram signals are often disturbed by noise caused by violent movement and strong environmental magnetic interference, resulting in inaccurate analysis results.
High-frequency noise is removed by filtering the original ECG signal, dynamic baselines are extracted and effective signals are filtered, and baseline drift is reduced, and an ECG signal classification model is input to obtain accurate analysis results.
Effectively reduce noise in ECG signals, improve signal quality, and enable the ECG signal classification model to process accurate ECG signals, thereby obtaining accurate analysis results.
Smart Images
Figure CN119157550A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of signal processing, and in particular to a signal processing method, an electronic device and a computer-readable storage medium. Background Art
[0002] Electrocardiogram (ECG) is a biomedical signal that reflects detailed information about heart activity in the form of electrical signals.
[0003] During the collection process, ECG signals are often affected by the user's strenuous exercise, strong magnetic interference from the environment, and other uncontrollable interferences. In other words, the collected ECG signals are often accompanied by strong noise.
[0004] If the analysis is performed directly based on the collected ECG signals, accurate analysis results may not be obtained. Summary of the invention
[0005] In view of the above, the embodiments of the present application provide a signal processing method, an electronic device and a computer-readable storage medium, which can reduce noise in electrocardiogram signals and realize accurate analysis of electrocardiogram signals.
[0006] The present application provides a signal processing method, including:
[0007] Get the original ECG signal;
[0008] Performing filtering on the original electrocardiogram signal to obtain a filtered electrocardiogram signal;
[0009] Extracting a dynamic baseline of the filtered ECG signal;
[0010] Based on the dynamic baseline, extracting a valid ECG signal from the filtered ECG signal;
[0011] The effective electrocardiogram signal is input into an electrocardiogram signal classification model to obtain a classification result of the original electrocardiogram signal.
[0012] The embodiment of the present application filters the ECG signal after receiving it. The filtering process can remove high-frequency noise. Then, based on the dynamic baseline, the effective ECG signal is screened, which can reduce the baseline drift of the ECG signal and the noise signal of the ECG signal, so that the ECG signal classification model can process the effective ECG signal, thereby obtaining accurate analysis results of the ECG signal.
[0013] In some embodiments, based on the dynamic baseline, extracting a valid ECG signal from the filtered ECG signal includes:
[0014] Acquiring a baseline range of the effective electrocardiogram signal;
[0015] In the filtered ECG signal, screening a target signal segment, wherein the dynamic baseline of the target signal segment is within the baseline range;
[0016] A valid ECG signal is determined based on the target signal segment.
[0017] This embodiment can remove dirty data with large baseline drift in the ECG signal by screening target signal segments whose dynamic baseline is within the baseline range of the valid ECG signal, thereby facilitating subsequent analysis and processing of the ECG signal classification model.
[0018] In some embodiments, determining a valid ECG signal based on the target signal segment includes:
[0019] If the signal duration of the target signal segment is not less than the preset valid signal duration, the target signal segment is used as a valid ECG signal.
[0020] This embodiment can screen out stable ECG signals through the effective signal duration, and avoid the situation where the ECG signal classification model cannot analyze and process due to the short target signal segment.
[0021] In some embodiments, obtaining the baseline range of the effective ECG signal includes:
[0022] Obtaining the peak value of each R wave in the filtered electrocardiogram signal;
[0023] Calculating the average height of the peaks of the R waves;
[0024] Based on the average height and a preset range coefficient, a baseline range of the effective electrocardiogram signal is obtained.
[0025] In some embodiments, the ECG signal classification model includes a plurality of parallel artificial neural networks, and the inputting of the valid ECG signal into the ECG signal classification model to obtain the classification result of the original ECG signal includes:
[0026] Splitting the effective ECG signal into a plurality of sub-ECG signals;
[0027] Each sub-ECG signal is input into a different artificial neural network in the ECG signal classification model to obtain a classification result of the original ECG signal.
[0028] The embodiment of the present application divides the effective ECG signal into multiple sub-ECG signals, and processes the sub-ECG signals respectively through multiple artificial neural networks, so that the ECG signal classification model can process ECG signals of different lengths, and the multiple artificial neural networks are arranged in parallel, so that the multiple sub-ECG signals can be processed synchronously, and the calculation time will not be prolonged due to the lengthening of the effective ECG signal.
[0029] In some embodiments, the ECG signal classification model further includes a classification layer, each parallel artificial neural network includes a fully connected layer group, and the last fully connected layer of each parallel artificial neural network is combined into a new fully connected layer, and the new fully connected layer is connected to the classification layer;
[0030] The step of inputting each sub-ECG signal into a different artificial neural network in the ECG signal classification model to obtain a classification result of the original ECG signal includes:
[0031] Inputting each sub-ECG signal into the different artificial neural networks respectively to obtain the output result of the new fully connected layer;
[0032] The output result is input into the classification layer to obtain the classification result of the electrocardiogram signal.
[0033] In some embodiments, there are signal segments that overlap in time between the multiple ECG sub-signals.
[0034] The embodiment of the present application can avoid the situation where the cut falls on an important waveform, resulting in an incomplete waveform and making it difficult to classify the ECG signal.
[0035] The present application also provides a signal processing method, including:
[0036] Obtain ECG signals;
[0037] Extract the dynamic baseline of ECG signal;
[0038] Based on the dynamic baseline, a valid ECG signal is extracted from the ECG signal.
[0039] The embodiment of the present application screens effective ECG signals based on a dynamic baseline, which can reduce the baseline drift of the ECG signals and reduce the noise of the ECG signals, so that the ECG signal classification model can process the effective ECG signals, thereby facilitating obtaining accurate analysis results of the ECG signals.
[0040] In some embodiments, based on the dynamic baseline, extracting a valid ECG signal from the ECG signal includes:
[0041] Acquiring a baseline range of the effective electrocardiogram signal;
[0042] In the electrocardiogram signal, a target signal segment is screened, wherein the dynamic baseline of the target signal segment is within the baseline range;
[0043] A valid ECG signal is determined based on the target signal segment.
[0044] This embodiment can remove dirty data with large baseline drift in the ECG signal by screening target signal segments whose dynamic baseline is within the baseline range of the valid ECG signal, thereby facilitating subsequent analysis and processing of the ECG signal classification model.
[0045] In some embodiments, determining a valid ECG signal based on the target signal segment includes:
[0046] If the signal duration of the target signal segment is not less than the preset valid signal duration, the target signal segment is used as a valid ECG signal.
[0047] In some embodiments, obtaining the baseline range of the effective ECG signal includes:
[0048] Obtaining the peak value of each R wave in the electrocardiogram signal;
[0049] Calculating the average height of the peaks of the R waves;
[0050] Based on the average height and a preset range coefficient, a baseline range of the effective electrocardiogram signal is obtained.
[0051] The present application also provides a signal processing method, the signal processing method comprising:
[0052] Obtain ECG signals;
[0053] Splitting the electrocardiogram signal into a plurality of sub-electrocardiogram signals;
[0054] Inputting each sub-ECG signal into a different artificial neural network in an ECG classification model to obtain a classification result of the ECG signal;
[0055] Wherein, the ECG classification model includes multiple parallel artificial neural networks.
[0056] The embodiment of the present application divides the ECG signal into multiple sub-ECG signals, and processes the sub-ECG signals respectively through multiple artificial neural networks, so that the ECG signal classification model can process ECG signals of different lengths, and the multiple artificial neural networks are arranged in parallel, so that the multiple sub-ECG signals can be processed synchronously, and the calculation time will not be prolonged due to the lengthening of the effective ECG signal.
[0057] In some embodiments, the method for segmenting multiple sub-ECG signals includes: segmenting the ECG signal into multiple sub-ECG signals according to a preset signal segmentation duration.
[0058] In some embodiments, the ECG signal classification model further includes a classification layer, each parallel artificial neural network includes a fully connected layer group, and the last fully connected layer of each parallel artificial neural network is combined into a new fully connected layer, and the new fully connected layer is connected to the classification layer;
[0059] The step of inputting each sub-ECG signal into different artificial neural networks in the ECG classification model to obtain the classification result of the ECG signal includes:
[0060] Inputting each sub-ECG signal into the different artificial neural networks to obtain the output result of the new fully connected layer;
[0061] The output result is input into the classification layer to obtain the classification result of the electrocardiogram signal.
[0062] In some embodiments, there are signal segments that overlap in time between the multiple ECG sub-signals.
[0063] An embodiment of the present application also provides an electronic device, which includes a processor and a memory, wherein the memory is used to store instructions, and the processor is used to call the instructions in the memory so that the electronic device executes the above-mentioned signal processing method.
[0064] An embodiment of the present application further provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed on an electronic device, the electronic device executes the above-mentioned signal processing method. BRIEF DESCRIPTION OF THE DRAWINGS
[0065] Figure 1 It is a flowchart of the steps of a signal processing method provided according to an embodiment of the present application.
[0066] Figure 2 It is a flowchart of the steps of extracting effective ECG signals according to an embodiment of the present application.
[0067] Figure 3 It is a sub-step flowchart of step 104 provided according to an embodiment of the present application.
[0068] Figure 4 It is a schematic diagram of the structure of a single convolutional neural network provided according to an embodiment of the present application.
[0069] Figure 5 It is a structural diagram of an electrocardiogram signal classification model provided according to an embodiment of the present application.
[0070] Figure 6 It is a sub-step flowchart of step 105 provided according to an embodiment of the present application.
[0071] Figure 7 It is a signal schematic diagram of ECG signal cutting provided according to an embodiment of the present application.
[0072] Figure 8 It is a schematic diagram of a scenario of ECG signal cutting provided according to an embodiment of the present application.
[0073] Fig. 9 It is a flowchart of the steps of a signal processing method provided according to another embodiment of the present application.
[0074] Fig.10 It is a flowchart of the steps of a signal processing method provided according to another embodiment of the present application.
[0075] Fig.11 It is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION
[0076] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the present application is described in detail below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that the implementation methods of the present application and the features in the implementation methods can be combined with each other without conflict.
[0077] In the following description, many specific details are set forth to facilitate a full understanding of the present application. The described implementations are only part of the implementations of the present application, rather than all the implementations.
[0078] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.
[0079] It should be further noted that, in this article, the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.
[0080] In this application, "at least one" means one or more, and "more than one" means two or more than two. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and drawings of this application are used to distinguish similar objects, rather than to describe a specific order or sequence.
[0081] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.
[0082] An electrocardiogram (ECG) is a biomedical signal that reflects detailed information about cardiac activity in the form of an electrical signal. In this application, if an ECG signal is not specified or specifically described, the ECG signal may be any ECG signal, such as an original ECG signal, an ECG signal to be processed, a pre-processed ECG signal, or an ECG signal after filtering.
[0083] Artificial Neural Network (ANN), also known as Neural Network (NN) or quasi-neural network, is a mathematical model or computational model that imitates the structure and function of biological neural networks (the central nervous system of animals, especially the brain) in the field of machine learning and cognitive science, and is used to estimate or approximate functions.
[0084] Convolutional Neural Network (CNN) is a type of artificial neural network. In the classic model, a convolutional neural network consists of one or more convolutional layers and one or more fully connected layers at the end, as well as associated weights and pooling layers.
[0085] The ECG signal is often affected by the user's strenuous exercise, strong magnetic interference from the environment, and other uncontrollable interferences during the collection process.
[0086] For example, during ECG signal detection, the metal electrodes of the detection device will come into contact with human skin. When the contact between the metal electrodes of the detection device and human skin is unstable, the ECG signal will produce a high-intensity baseline drift.
[0087] Even the ECG signals output by testing equipment that meets the Food and Drug Administration (FDA) standards still have a lot of noise interference, such as high-frequency myoelectric interference and electromagnetic white noise.
[0088] Therefore, in some embodiments, the ECG signal output by the detection device may be filtered, and then the filtered ECG signal may be analyzed manually, or an artificial intelligence model may be used to analyze the filtered ECG signal.
[0089] However, after filtering, the ECG signal still has a lot of interference noise signals. In addition, irregular operations or other factors during the ECG signal measurement process may result in a large amount of dirty data in the ECG signal. In particular, there is a lot of interference in the ECG signal detected by a single lead, which is difficult to directly analyze and detect through manual or artificial intelligence models.
[0090] In view of the above, an embodiment of the present application provides a signal processing method, which may include: acquiring an original ECG signal; filtering the original ECG signal to obtain a filtered ECG signal; extracting a dynamic baseline of the filtered ECG signal; based on the dynamic baseline, extracting a valid ECG signal from the filtered ECG signal; inputting the valid ECG signal into an ECG signal classification model to obtain a classification result of the original ECG signal.
[0091] The signal processing method of the present application can be applied to one or more electronic devices. The electronic device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to processors, microprogrammed control units (MCU), application-specific integrated circuits (ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc. The electronic device can be a portable electronic device (such as a mobile phone, a tablet computer), a personal computer, a server, a wearable electronic device (such as a smart watch, a smart bracelet), etc., but is not limited to this.
[0092] After receiving the ECG signal, the embodiment of the present application filters the ECG signal to remove high-frequency noise. Then, based on the dynamic baseline, the effective ECG signal is screened, which can reduce the baseline drift of the ECG signal and the noise of the ECG signal, so that the ECG signal classification model can process the effective ECG signal, thereby obtaining accurate analysis results of the ECG signal.
[0093] By processing the ECG signal using the signal processing method provided in the embodiment of the present application, the noise of the ECG signal can be effectively reduced, making it easier for the ECG signal classification model to process the ECG signal, thereby obtaining an accurate ECG signal classification result.
[0094] Figure 1This is a flowchart of the steps of an embodiment of the signal processing method of the present application. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted.
[0095] See also Figure 1 As shown, the signal processing method may include the following steps.
[0096] Step 101, obtaining an original electrocardiogram signal.
[0097] The original ECG signal is the ECG signal directly measured from the user.
[0098] The original ECG signal may be an ECG signal measured in a single-lead manner or an ECG signal measured in a multi-lead manner.
[0099] In some embodiments, the ECG measuring instrument can measure the original ECG signal of the user and transmit it to the electronic device of the embodiment of the present application, so that the electronic device of the embodiment of the present application obtains the original ECG signal within a certain time period.
[0100] In other embodiments, the electronic device in the embodiments of the present application is integrated with a function of measuring an electrocardiogram signal, and the electronic device in the embodiments of the present application can directly measure and obtain the original electrocardiogram signal of the user.
[0101] The above-mentioned method of obtaining the original ECG signal is only an example, and the embodiments of the present application are not limited thereto. The following takes the electronic device as a smart bracelet as an example to illustrate the process of the electronic device obtaining the original ECG signal.
[0102] For example, the user can select the ECG signal measurement function on the human-computer interaction interface provided by the smart bracelet, and the smart bracelet can start detecting the user's original ECG signal until the set signal detection time ends. The signal detection time of the original ECG signal can be set according to demand, for example, it can be set to 30 seconds to 60 seconds, or more than 60 seconds, or even more than 24 hours. For example, if the signal detection time is 30 seconds, the smart bracelet can detect 30 seconds of the user's original ECG signal.
[0103] When wearing a smart bracelet, the user will move and exercise, which may cause the smart bracelet to fail to stick closely to the user's skin during the ECG signal measurement process, and thus fail to obtain ECG signals stably and continuously. In addition, smart bracelets are generally single-lead detection devices, and the ECG signals measured by the single-lead detection method also have more interference. Therefore, there will be too many interference signals and dirty data in the original ECG signals detected by the smart bracelet.
[0104] In view of this, after executing step 101, the embodiment of the present application may execute steps 102 to 104, that is, preprocess the original ECG signal, thereby reducing the interference signal in the original ECG signal and removing the dirty data in the original ECG signal.
[0105] Step 102: filter the original ECG signal to obtain a filtered ECG signal.
[0106] In some embodiments, step 102 may include: performing low-pass filtering on the original ECG signal, that is, retaining ECG signals with frequencies below a set critical value in the original ECG signal, and blocking and attenuating ECG signals with frequencies above the set critical value.
[0107] For example, the original ECG signal may include high-frequency noise. The original ECG signal may be input into a low-pass filter, such as inputting the original ECG signal into a 25 Hz low-pass filter to retain ECG signals below 25 Hz and obtain a filtered ECG signal.
[0108] In the above step 102, high-frequency noises such as electromyographic interference and power frequency interference can be filtered out by low-pass filtering, so that effective components such as P wave (atrial depolarization wave), QRS wave (ventricular depolarization wave), and T wave (ventricular repolarization wave) in the electrocardiogram signal can be retained.
[0109] In addition to the interference caused by the above-mentioned high-frequency noise, the original ECG signal may also carry interference from low-frequency noise. For example, breathing or other body activities may cause changes in the impedance between the electrode and the skin, which may cause baseline changes in the ECG signal (also known as low-frequency drift or baseline drift).
[0110] That is to say, human breathing and slight movement of the electrode on the skin will cause baseline drift. Baseline drift can directly affect the upper and lower positions of the ECG signal, causing inconvenience to the detection of the QRS complex of the ECG signal and the direct analysis of the ECG.
[0111] Therefore, after obtaining the filtered ECG signal, the embodiment of the present application can execute steps 103 to 104 to eliminate dirty data in the filtered ECG signal, remove baseline drift, and obtain a valid ECG signal to facilitate subsequent processing of the ECG signal classification model.
[0112] Step 103: extracting the dynamic baseline of the ECG signal after filtering.
[0113] In some embodiments, the dynamic baseline of the filtered ECG signal can be extracted by complex nonlinear calculation methods such as wavelet transform and empirical mode decomposition (EMD).
[0114] However, the above method has large computational complexity, poor real-time performance, and may cause waveform distortion.
[0115] Therefore, reference Figure 2 As shown, in other embodiments, the electronic device may perform median filtering on the filtered ECG signal to obtain a dynamic baseline of the filtered ECG signal.
[0116] For example, the electronic device can input the filtered ECG signal into a median filter, which can be a sliding window median filter. The window length of the sliding window median filter can be determined according to the sampling frequency, etc., and the window length can include at least three sampling points of the ECG signal.
[0117] This embodiment uses median filtering to reduce the amount of calculation for dynamic baseline extraction, improve the real-time performance of the calculation, and avoid waveform distortion.
[0118] After the dynamic baseline of the filtered ECG signal is extracted, step 104 may be executed.
[0119] Step 104 : extracting effective ECG signals from the filtered ECG signals based on the dynamic baseline.
[0120] In some embodiments, reference Figure 3 As shown, step 104 may include:
[0121] Step 1041, obtaining a baseline range of a valid ECG signal.
[0122] The baseline range of the effective ECG signal can be determined based on the filtered ECG signal and the dynamic baseline.
[0123] In some embodiments, the user can refer to the parameters such as the filtered ECG signal and the dynamic baseline, and set the baseline range of the valid ECG signal on the human-computer interaction interface so that the electronic device can obtain the baseline range.
[0124] In some other embodiments, step 1041 may include:
[0125] Step 301, obtaining the peak value of each R wave in the electrocardiographic signal after filtering.
[0126] For example, the peak value of the R wave in the filtered electrocardiographic signal can be obtained by the Pan-Tompkins method.
[0127] For another example, the threshold method and the slope method can be combined to identify the R wave and obtain the peak value of the R wave. Specifically, a differential operation is performed on the filtered ECG signal to obtain a differential threshold to determine the falling edge of the QRS wave, and then the position of the R wave is determined according to a certain amplitude and threshold, thereby obtaining the peak value of the R wave.
[0128] The above method for detecting R waves is only an example and can be set according to actual application requirements during actual application. The embodiments of the present application are not limited to this.
[0129] Step 302, calculating the average height of the peak values of each R wave.
[0130] The height of the peak of the R wave refers to the difference between the peak of the R wave and the dynamic baseline of the R wave.
[0131] For example, the electronic device may obtain the height of each R wave peak in the filtered electrocardiographic signal, calculate the average of the heights of each R wave peak, and obtain the average height Ra.
[0132] Step 303: Obtain a baseline range of a valid ECG signal based on the average height and a preset range coefficient.
[0133] In some embodiments, the range coefficient may include an upper threshold coefficient and a lower threshold coefficient, wherein the upper threshold coefficient is greater than the lower threshold coefficient; the upper threshold coefficient is multiplied by the average height Ra to obtain the upper threshold T1 in the baseline range; the lower threshold coefficient is multiplied by the average height Ra to obtain the lower threshold T2, and the baseline range of the effective ECG signal is [T2, T1].
[0134] For example, if the upper threshold coefficient is a and the lower threshold coefficient is -b, the upper threshold T1 is: T1 = a×Ra, and the lower threshold T2 is as follows: T2 = -b×Ra. The baseline range of the valid ECG signal is [-b×Ra, a×Ra], where a and b are natural numbers between 0 and 10 respectively.
[0135] Step 1042, screening a target signal segment in the filtered ECG signal, wherein the dynamic baseline of the target signal segment is within the baseline range.
[0136] That is, the electronic device can determine a signal segment of the dynamic baseline within the baseline range in the ECG signal after filtering, and use the signal segment as the target signal segment.
[0137] For example, assuming that the ECG signal after filtering is 30 seconds, in the ECG signal after filtering, if the dynamic baseline corresponding to the signal segment from 0 to 21 seconds is within the baseline range of the valid ECG signal, then the signal segment from 0 to 21 seconds is the target signal segment; if the dynamic baseline corresponding to the signal segment from 21 to 25 seconds is not within the baseline range of the valid ECG signal, then the signal segment from 21 to 25 seconds is not the target signal segment; if the dynamic baseline corresponding to the signal segment from 25 to 30 seconds is within the baseline range of the valid ECG signal, then the signal segment from 25 to 30 seconds is the target signal segment.
[0138] Step 1043: determine a valid ECG signal based on the target signal segment.
[0139] In some embodiments, the target signal segment may be used as a valid ECG signal.
[0140] However, when performing subsequent ECG signal analysis, if the target signal segment is too short, the ECG signal will transmit less information, making it difficult to analyze the ECG signal.
[0141] In view of this, in some other embodiments, step 1043 may include: if the signal duration of the target signal segment is not less than a preset valid signal duration, taking the target signal segment as a valid ECG signal.
[0142] That is, in this embodiment, if a signal segment of an ECG signal is a valid ECG signal, the following two conditions must be met:
[0143] 1. The dynamic baseline of the signal segment is within the baseline range of the valid ECG signal.
[0144] 2. The signal duration of the signal segment is not less than the preset effective signal duration.
[0145] Among them, the preset effective signal duration can be set according to the requirements of signal analysis, and the embodiments of the present application are not limited to this.
[0146] For a signal segment that does not meet the above two conditions, the signal segment may be marked as an invalid ECG signal.
[0147] For example, assuming that the preset valid signal duration is 6 seconds, if the signal segment from 0 to 21 seconds is the target signal segment, since the signal duration of the target signal segment is greater than 6 seconds, the target signal segment can be used as a valid ECG signal; if the signal segment from 25 to 30 seconds is the target signal segment, since the signal duration of the target signal segment is less than 6 seconds, the target signal segment is not a valid ECG signal, and the target signal segment can be marked as an invalid ECG signal.
[0148] Step 105: input the valid ECG signal into the ECG signal classification model to obtain the classification result of the original ECG signal.
[0149] The categories and classification results of the ECG signal classification model can be set according to actual needs. For example, the classification results may include whether atrial fibrillation (AF) exists, whether arrhythmia exists, etc., but are not limited thereto.
[0150] The ECG signal classification model is an artificial intelligence model.
[0151] In some embodiments, the ECG signal classification model may be a single artificial neural network, such as a single convolutional neural network, wherein the single convolutional neural network may refer to Figure 4 As shown, in Figure 4 In , n refers to the size of data or feature map, and the size of data or feature map of each layer of a single convolutional neural network can be recorded as 1×n0, 1×n1, 1×n2 and 1×n3 respectively; c refers to the size of the channel, and the size of the channel of each layer of a single convolutional neural network is recorded as 1, c1, c2 and c3 respectively; f refers to the size of the fully connected layer, and the size of the last connected layer of a single convolutional neural network is 1×f0, and A / B refers to the classification category of the original ECG signal, for example, A can refer to the presence of atrial fibrillation, and B refers to the absence of atrial fibrillation. The input of this single convolutional neural network is a valid ECG signal with a certain data length, and the output is the classification result of the original ECG signal.
[0152] For example, the input of the single convolutional neural network is: a valid ECG signal with a time length of 30 seconds, and the sampling frequency of the ECG signal is once every 0.5 seconds, that is, the valid ECG signal has a total of 60 sampling points, that is, the dimension of the valid ECG signal is 60x1, and the input size of the single convolutional neural network is also 60x1.
[0153] However, the data length of the input signal that a single convolutional neural network can process is fixed. If the data length of the effective ECG signal changes, such as the signal time becomes longer or the number of sampling points increases, the ECG signal classification model cannot process the effective ECG signal with changed data length.
[0154] For example, the input size of the above-mentioned single convolutional neural network is 60x1. When the valid ECG signal to be input is 40 seconds (that is, the dimension of the valid ECG signal is 80x1), the ECG signal classification model cannot process the valid ECG signal, and the convolutional neural network needs to be retrained.
[0155] Therefore, in other embodiments, the ECG signal classification model may use a recurrent neural network, which can process ECG signals of varying lengths. However, when the ECG signal input to the recurrent neural network becomes longer, the operation time will become longer.
[0156] In view of the above, the present application embodiment also provides an electrocardiogram signal classification model, referring to Figure 5 As shown, the ECG signal classification model includes multiple parallel artificial neural networks 501.
[0157] Among them, the artificial neural network 501 can be a convolutional neural network, but is not limited to this. In actual application, the specific structure of the artificial neural network can be set according to application requirements.
[0158] Convolutional neural networks can use CNN architectures such as VGG and ResNet.
[0159] The following combination Figure 5The ECG signal classification model shown in FIG. 1 illustrates the sub-steps of step 105. Figure 6 As shown, step 105 may include:
[0160] Step 1051, dividing the effective ECG signal into a plurality of sub-ECG signals.
[0161] For example, the effective ECG signal may be segmented into a plurality of sub-ECG signals according to a preset signal segmentation duration.
[0162] The signal segmentation time length can be set according to the length of the input signals of the multiple parallel artificial neural networks in the ECG classification model, and the length of the input signal of the artificial neural network can be set according to the actual application requirements. For example, if the length of the input signal of the artificial neural network is 4 to 10 seconds, the signal segmentation time length can also be 4 to 10 seconds.
[0163] In some embodiments, the multiple sub-ECG signals can be separated from each other, that is, there is no overlap between the multiple sub-ECG signals. For example, the effective ECG signal is 30 seconds, the signal segmentation time is 10 seconds, and the multiple sub-ECG signals are the sub-ECG signals of the 0th to 10th seconds, the sub-ECG signals of the 10th to 11th seconds, and the sub-ECG signals of the 11th to 30th seconds.
[0164] refer to Figure 7 As shown, in the above segmentation method, if the cutting position falls on an important waveform, the waveform will be incomplete, and the segmented sub-ECG signals will become noise, which is difficult to be processed by the ECG signal classification model.
[0165] Therefore, reference Figure 8 As shown, in some other embodiments, the complete ECG signal 801 may be segmented into a plurality of sub-ECG signals 802 , and there may be signal segments with temporal overlap between the plurality of sub-ECG signals 802 .
[0166] The time of the overlapping part can be set according to requirements, for example, it can be set to 0 to 3 seconds.
[0167] In this embodiment, there are signal segments with temporal overlap between the multiple sub-ECG signals, which can improve the integrity of the segmented sub-ECG signals, thereby facilitating the processing of the ECG signal classification model.
[0168] It is understandable that when the valid ECG signal is segmented to the last segment, if the duration of the ECG signal of the last segment is less than the preset signal segmentation duration, the ECG signal of the last segment can be discarded, or it can be padded forward so that the duration of the ECG signal of the last segment is equal to the preset signal segmentation duration.
[0169] Step 1052: Input each sub-ECG signal into a different artificial neural network in the ECG signal classification model to obtain the classification result of the original ECG signal.
[0170] For example, it is assumed that the ECG signal classification model includes three artificial neural networks, which are respectively denoted as artificial neural network 1, artificial neural network 2 and artificial neural network 3.
[0171] When the effective ECG signal is divided into two N-second sub-ECG signals, sub-ECG signal 1 can be input into artificial neural network 1, and sub-ECG signal 2 can be input into artificial neural network 2, thereby obtaining the classification result of the original ECG signal.
[0172] When the effective ECG signal is divided into three N-second sub-ECG signals, sub-ECG signal 1 can be input into artificial neural network 1, sub-ECG signal 2 can be input into artificial neural network 2, and sub-ECG signal 3 can be input into artificial neural network 3, thereby obtaining the classification result of the original ECG signal.
[0173] This embodiment divides the effective ECG signal into multiple sub-ECG signals, and processes the sub-ECG signals respectively through multiple artificial neural networks, so that the ECG signal classification model can process ECG signals of different lengths, and multiple artificial neural networks are set in parallel, so that multiple sub-ECG signals can be processed synchronously, and the calculation time will not be prolonged due to the lengthening of the effective ECG signal, thereby ensuring the stability of the calculation time.
[0174] Further, refer again Figure 5 As shown, the ECG signal classification model may also include a classification layer, each parallel artificial neural network includes a fully connected layer group, that is, each parallel artificial neural network includes at least one fully connected layer, and the last fully connected layer 5011 of each parallel artificial neural network 501 is combined into a new fully connected layer 502. For example, the last fully connected layer 5011 of each artificial neural network can be combined into a new fully connected layer 502 by concatenation, and the new fully connected layer 502 is connected to the classification layer 503. The classification layer can adopt a binary classification method based on the softmax function, or other classification methods, which is not limited to the embodiments of the present application.
[0175] Among them, each sub-ECG signal is input into a different artificial neural network to obtain the classification result of the original ECG signal, including: inputting each sub-ECG signal into a different artificial neural network to obtain the output result of a new fully connected layer; inputting the output result into the classification layer to obtain the classification result of the ECG signal.
[0176] The ECG signal classification model of this embodiment first analyzes each sub-ECG signal in the effective ECG signal to obtain an output result, and then integrates the output results to obtain the classification result of the ECG signal, which can make the obtained classification result more accurate.
[0177] The ECG signal classification model of the embodiment of the present application can be trained using a back-propagation method.
[0178] For example, the training steps for an ECG signal classification model may include:
[0179] Acquire an ECG signal sample, wherein the ECG signal sample includes a plurality of sub-ECG signals segmented from a complete ECG signal, and a target classification result of the complete ECG signal;
[0180] Inputting the multiple sub-ECG signals into different convolutional neural networks in the ECG signal classification model respectively, and obtaining the predicted classification results output by the ECG signal classification model;
[0181] Calculate the error between the target classification result and the predicted classification result;
[0182] If the error is equal to or less than the preset error expectation value, the training is terminated to obtain a trained ECG signal classification model; if the error is greater than the preset error expectation value, the network parameters in the ECG signal classification model are updated based on the error, and then the training is continued.
[0183] In the ECG signal sample of the embodiment of the present application, only the complete ECG signal needs to be labeled, and there is no need to label the multiple sub-ECG signals that are segmented, thereby reducing the workload of sample labeling.
[0184] Referring to Table 1, Table 1 lists the effects of different ECG signal classification models on the classification results of AF and non-AF tests.
[0185] It can be seen from Table 1 that the accuracy of the ECG signal classification model including multiple parallel convolutional neural networks (i.e., ECG signal classification model 1) proposed in the embodiment of the present application can reach 95%, and the accuracy of the ECG signal classification model including a recurrent neural network (i.e., ECG signal classification model 2) proposed in the embodiment of the present application can reach 93%.
[0186] Table 1
[0187] method Accuracy Sensitivity Specificity ECG signal classification model 1 95％ 93％ 96％ ECG signal classification model 2 93％ 82％ 96％
[0188] The signal processing method of the embodiment of the present application at least includes the following effects:
[0189] 1. The embodiment of the present application adopts low-pass filtering to remove high-frequency noise such as electromyography and power frequency of the ECG signal, thereby highlighting the QRST wave information required by the ECG signal classification model for ECG signal analysis, making it easier for the subsequent ECG signal classification model to process the ECG signal.
[0190] 2. The embodiment of the present application can eliminate dirty data with large baseline drift in the ECG signal by screening the target signal segments. Furthermore, by screening the valid ECG signal in the target signal segments, an effective, stable and low-noise ECG signal can be obtained, thereby improving the quality of the ECG signal input into the ECG signal classification model, and further facilitating the subsequent analysis and processing of the ECG signal classification model.
[0191] 3. The embodiment of the present application divides the effective ECG signal into segments, and then simultaneously inputs the segments into multiple parallel artificial neural networks to obtain classification results, so that the ECG signal classification model can process ECG signals of different lengths, and multiple artificial neural networks are arranged in parallel, so that multiple sub-ECG signals can be processed synchronously, and the calculation time will not be prolonged due to the lengthening of the effective ECG signal. That is, by processing the ECG signal through the ECG signal classification model, the classification result of the complete ECG signal can be quickly obtained.
[0192] The present application also provides a signal processing method, referring to Fig. 9 As shown, the signal processing method includes:
[0193] Step 901, obtaining an electrocardiogram signal.
[0194] The electrocardiogram signal is an electrocardiogram signal to be judged for validity.
[0195] In some embodiments, the ECG signal may be a raw ECG signal.
[0196] However, since high-frequency noise interference may exist in the original ECG signal, in other embodiments, step 901 may include: obtaining the original ECG signal; filtering the original ECG signal to obtain a filtered ECG signal; the filtered ECG signal is the ECG signal obtained in step 901.
[0197] The details of the steps of acquiring the original ECG signal and filtering processing can be referred to the above steps 101 to 102, which are not limited in this embodiment.
[0198] Step 902: extract the dynamic baseline of the ECG signal.
[0199] In some embodiments, the dynamic baseline of the ECG signal may be extracted by complex nonlinear calculation methods such as wavelet transform and empirical mode decomposition (EMD).
[0200] However, the above method has large computational complexity, poor real-time performance, and may cause waveform distortion.
[0201] Therefore, in some other embodiments, the electronic device may perform median filtering on the ECG signal to obtain a dynamic baseline of the filtered ECG signal.
[0202] For example, the electronic device may input the ECG signal into a median filter, which may be a sliding window median filter. The window length of the sliding window median filter may be determined according to the sampling frequency, etc., and the window length may include at least three sampling points of the ECG signal.
[0203] This embodiment uses median filtering to reduce the amount of calculation for dynamic baseline extraction, improve the real-time performance of the calculation, and avoid waveform distortion.
[0204] After the dynamic baseline of the ECG signal is extracted, step 903 may be executed.
[0205] Step 903: extracting effective ECG signals from the ECG signals based on the dynamic baseline.
[0206] In some embodiments, step 903 may include: acquiring a baseline range of a valid ECG signal; screening a target signal segment in the ECG signal, wherein a dynamic baseline of the target signal segment is within the baseline range; and determining a valid ECG signal based on the target signal segment.
[0207] Further, in some embodiments, obtaining a baseline range of a valid ECG signal includes: obtaining the peak value of the R wave in the ECG signal; calculating the average height of the peak value of the R wave in the ECG signal; and obtaining the baseline range of the valid ECG signal based on the average value and a preset range coefficient.
[0208] Further, in some embodiments, determining a valid ECG signal based on a target signal segment may include: if a signal duration of the target signal segment is not less than a preset valid signal duration, treating the target signal segment as a valid ECG signal.
[0209] The above step 903 is substantially the same as step 104 and will not be described in detail in this embodiment.
[0210] In some embodiments, after step 104, the valid ECG signal may be input into the ECG signal classification model. This step is substantially the same as step 105 and will not be described in detail in this embodiment.
[0211] This embodiment can eliminate dirty data with large baseline drift in the ECG signal by screening the target signal segments. Furthermore, by screening the valid ECG signal in the target signal segments, an effective, stable and low-noise ECG signal can be obtained, thereby improving the quality of the ECG signal input into the ECG signal classification model, thereby facilitating the analysis and processing of the ECG signal classification model.
[0212] The present application also provides a signal processing method, referring to Fig.10 As shown, the signal processing method includes:
[0213] Step 1001, obtaining an electrocardiogram signal.
[0214] In some embodiments, step 1001 may include: acquiring an original ECG signal; and preprocessing the original ECG signal to obtain the ECG signal acquired in step 1001 .
[0215] Among them, the preprocessing step may include: filtering the original ECG signal to obtain a filtered ECG signal; extracting a dynamic baseline of the filtered ECG signal; and extracting a valid ECG signal from the filtered ECG signal based on the dynamic baseline.
[0216] The specific details of step 1001 can refer to the description of steps 101 to 104, and will not be described in detail in this embodiment.
[0217] It is understandable that if the quality of the received original ECG signal is good, the original ECG signal may only be filtered, or only valid ECG signal extraction may be performed, or the received original ECG signal may be directly used as the ECG signal after filtering.
[0218] That is to say, the preprocessing step of the original ECG signal can be set according to the quality of the original ECG signal, and this embodiment does not limit the specific preprocessing step of the ECG signal.
[0219] Step 1002: Segment the ECG signal into a plurality of sub-ECG signals.
[0220] The segmentation method of the ECG signal can be set according to actual application requirements. For example, the ECG signal can be segmented into multiple sub-ECG signals according to a preset signal segmentation time.
[0221] The signal segmentation time length can be set according to the length of the input signal of multiple parallel artificial neural networks in the ECG classification model, and the length of the input signal of the artificial neural network can be set according to actual application requirements. For example, if the length of the input signal of the artificial neural network is 4 to 10 seconds, the signal segmentation time length can also be 4 to 10 seconds.
[0222] In some embodiments, there may or may not be signal segments that overlap in time between the multiple ECG sub-signals.
[0223] Among them, step 1002 and step 1051 of this embodiment are substantially the same and will not be described in detail here.
[0224] Step 1003: input each sub-ECG signal into a different artificial neural network in the ECG classification model to obtain a classification result of the ECG signal.
[0225] Among them, the ECG classification model includes multiple parallel artificial neural networks.
[0226] In some embodiments, again referring to Figure 5 As shown, the ECG signal classification model also includes a classification layer, each parallel artificial neural network includes a fully connected layer group, that is, each parallel artificial neural network includes at least one fully connected layer, and the last fully connected layer of each parallel artificial neural network is combined into a new fully connected layer, and the new fully connected layer is connected to the classification layer; wherein step 1003 may include:
[0227] Inputting each of the sub-ECG signals into different artificial neural networks of the ECG signal classification model to obtain the output result of the new fully connected layer;
[0228] The output result is input into the classification layer to obtain the classification result of the electrocardiogram signal.
[0229] Among them, step 1003 is substantially the same as step 1052 and will not be described again here.
[0230] The embodiment of the present application divides the effective ECG signal into segments, and then simultaneously inputs the segments into multiple parallel convolutional neural networks to obtain classification results, so that the ECG signal classification model can process ECG signals of different lengths, and multiple convolutional neural networks are arranged in parallel, so that multiple sub-ECG signals can be processed synchronously, and the calculation time will not be prolonged due to the lengthening of the ECG signal. That is, by processing the ECG signal through the ECG signal classification model, the classification result of the complete ECG signal can be quickly obtained.
[0231] Fig.11 This is a schematic diagram of an embodiment of an electronic device of the present application.
[0232] The electronic device 100 includes a memory 20, a processor 30, and a computer program 40 stored in the memory 20 and executable on the processor 30. When the processor 30 executes the computer program 40, the steps in the above-mentioned signal processing method embodiment are implemented, for example Figure 1 Steps 101 to 105 shown in FIG. 101 , or as shown in FIG. Fig. 9 Steps 901 to 903 shown in FIG. 1 , or as shown in FIG. 1 Fig.10 Steps 1001 to 1003 are shown.
[0233] Exemplarily, the computer program 40 may also be divided into one or more modules / units, which are stored in the memory 20 and executed by the processor 30. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 40 in the electronic device 100.
[0234] Those skilled in the art will understand that the schematic diagram is merely an example of the electronic device 100 and does not constitute a limitation of the electronic device 100. The electronic device 100 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device 100 may also include input and output devices, network access devices, buses, etc.
[0235] The processor 30 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, a single-chip microcomputer, or the processor 30 may also be any conventional processor, etc.
[0236] The memory 20 can be used to store the computer program 40 and / or the module / unit. The processor 30 realizes various functions of the electronic device 100 by running or executing the computer program and / or the module / unit stored in the memory 20 and calling the data stored in the memory 20. The memory 20 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 20 can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.
[0237] If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.
[0238] In the several embodiments provided in this application, it should be understood that the disclosed electronic device and method can be implemented in other ways. For example, the electronic device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation.
[0239] In addition, each functional unit in each embodiment of the present application may be integrated into the same processing unit, each unit may exist physically separately, or two or more units may be integrated into the same unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.
[0240] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or basic features of the present application. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or electronic devices stated in the electronic device claim can also be implemented by the same unit or electronic device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.
[0241] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the above embodiments, a person of ordinary skill in the art should understand that the technical solution of the present application may be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present application.
Claims
1. A signal processing method, It is characterized in that include: Get the original ECG signal; Performing filtering on the original electrocardiogram signal to obtain a filtered electrocardiogram signal; Extracting a dynamic baseline of the filtered ECG signal; Based on the dynamic baseline, extracting a valid ECG signal from the filtered ECG signal; The effective electrocardiogram signal is input into an electrocardiogram signal classification model to obtain a classification result of the original electrocardiogram signal.
2. The signal processing method according to claim 1, It is characterized in that The extracting a valid ECG signal from the ECG signal after the filtering based on the dynamic baseline comprises: Acquiring a baseline range of the effective electrocardiogram signal; In the filtered ECG signal, screening a target signal segment, wherein the dynamic baseline of the target signal segment is within the baseline range; A valid ECG signal is determined based on the target signal segment.
3. The signal processing method according to claim 2, It is characterized in that The determining of a valid electrocardiogram signal based on the target signal segment comprises: If the signal duration of the target signal segment is not less than the preset valid signal duration, the target signal segment is used as a valid ECG signal.
4. The signal processing method according to claim 2, It is characterized in that The obtaining of the baseline range of the effective electrocardiogram signal includes: Obtaining the peak value of each R wave in the filtered electrocardiogram signal; Calculating the average height of the peaks of the R waves; Based on the average height and a preset range coefficient, a baseline range of the effective electrocardiogram signal is obtained.
5. The signal processing method according to any one of claims 1 to 4, It is characterized in that The ECG signal classification model includes a plurality of parallel artificial neural networks. The inputting of the effective ECG signal into the ECG signal classification model to obtain the classification result of the original ECG signal includes: Splitting the effective ECG signal into a plurality of sub-ECG signals; Each sub-ECG signal is input into a different artificial neural network in the ECG signal classification model to obtain a classification result of the original ECG signal.
6. The signal processing method according to claim 5, It is characterized in that The ECG signal classification model further includes a classification layer, each parallel artificial neural network includes a fully connected layer group, and the last fully connected layer of each parallel artificial neural network is combined into a new fully connected layer, and the new fully connected layer is connected to the classification layer; The step of inputting each sub-ECG signal into a different artificial neural network in the ECG signal classification model to obtain a classification result of the original ECG signal includes: Inputting each sub-ECG signal into the different artificial neural networks respectively to obtain the output result of the new fully connected layer; The output result is input into the classification layer to obtain the classification result of the electrocardiogram signal.
7. The signal processing method according to claim 5, It is characterized in that There are signal segments that overlap in time between the multiple sub-ECG signals.
8. A signal processing method, It is characterized in that include: Obtain ECG signals; Extracting a dynamic baseline of the electrocardiogram signal; Based on the dynamic baseline, a valid electrocardiogram signal is extracted from the electrocardiogram signal.
9. The signal processing method according to claim 8, It is characterized in that The extracting a valid electrocardiogram signal from the electrocardiogram signal based on the dynamic baseline includes: Acquiring a baseline range of the effective electrocardiogram signal; In the electrocardiogram signal, screening a target signal segment, wherein a dynamic baseline of the target signal segment is within the baseline range; A valid ECG signal is determined based on the target signal segment.
10. The signal processing method according to claim 9, It is characterized in that The determining of a valid electrocardiogram signal based on the target signal segment comprises: If the signal duration of the target signal segment is not less than the preset valid signal duration, the target signal segment is used as a valid ECG signal.
11. The signal processing method according to claim 9, It is characterized in that The obtaining of the baseline range of the effective electrocardiogram signal includes: Obtaining the peak value of each R wave in the electrocardiogram signal; Calculating the average height of the peaks of the R waves; Based on the average height and a preset range coefficient, a baseline range of the effective electrocardiogram signal is obtained.
12. A signal processing method, It is characterized in that The signal processing method comprises: Obtain ECG signals; Splitting the electrocardiogram signal into a plurality of sub-electrocardiogram signals; Inputting each sub-ECG signal into a different artificial neural network in an ECG classification model to obtain a classification result of the ECG signal; Wherein, the ECG classification model includes multiple parallel artificial neural networks.
13. The signal processing method according to claim 12, It is characterized in that The method for segmenting the multiple sub-ECG signals includes: segmenting the ECG signal into multiple sub-ECG signals according to a preset signal segmentation duration.
14. The signal processing method according to claim 12, It is characterized in that The ECG signal classification model further includes a classification layer, each parallel artificial neural network includes a fully connected layer group, and the last fully connected layer of each parallel artificial neural network is combined into a new fully connected layer, and the new fully connected layer is connected to the classification layer; The step of inputting each sub-ECG signal into different artificial neural networks in the ECG classification model to obtain the classification result of the ECG signal includes: Inputting each sub-ECG signal into the different artificial neural networks to obtain the output result of the new fully connected layer; The output result is input into the classification layer to obtain the classification result of the electrocardiogram signal.
15. The signal processing method according to any one of claims 12 to 14, It is characterized in that There are signal segments that overlap in time between the multiple sub-ECG signals.
16. An electronic device, comprising a processor and a memory, It is characterized in that The memory is used to store instructions, and the processor is used to call the instructions in the memory, so that the electronic device executes the signal processing method as described in any one of claims 1 to 7, or executes the signal processing method as described in any one of claims 8 to 11, or executes the signal processing method as described in any one of claims 12 to 15.
17. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer instructions, and when the computer instructions are executed on an electronic device, the electronic device executes the signal processing method as described in any one of claims 1 to 7, or executes the signal processing method as described in any one of claims 8 to 11, or executes the signal processing method as described in any one of claims 12 to 15.