Emotional state recognition method and device based on electrocardiosignal and deep learning
By using an emotion state recognition method based on electrocardiogram signals and deep learning, the problem of feature extraction relying on manual intervention and insufficient model mapping capabilities in traditional methods is solved, achieving more efficient identification of depressive states, especially accurate identification of early depressive states.
Patent Information
- Application Number
- CN202511258017.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-04
- Publication Date
- 2025-12-23
AI Technical Summary
Existing ECG-based depressive state recognition technologies rely on extensive experience and feature extraction from numerous experiments. Furthermore, the limited spatial mapping capabilities of traditional machine learning models hinder the improvement of recognition performance, particularly in the timely and accurate identification of early depressive states.
An emotion state recognition method based on electrocardiogram (ECG) signals and deep learning is adopted, including collecting standard lead I ECG signals, preprocessing, constructing a deep learning model, splitting the training set and validation set for training, and testing with a test set to finally recognize the emotion state.
By automatically extracting features through deep learning, the loss of implicit features in the data is avoided, which improves the accuracy and efficiency of identifying depressive states, especially in the identification of early depressive states.
Smart Images
Figure CN121176908A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the interdisciplinary field of computer technology and physiological electrical signal processing, and in particular to a method and apparatus for emotion state recognition based on electrocardiogram signals and deep learning. Background Technology
[0002] Contemporary psychiatric research (WHO, 2024) shows that persistent psychological stress is the primary environmental factor triggering depressive states. Currently, a large number of individuals are chronically exposed to stressors such as work intensity, family conflicts, or social competition. The accelerated pace of modern society is reshaping the psychological resilience of individuals, leading to a vicious cycle of "stress-neural damage-emotional regulation failure." Depression is no longer simply a medical problem, but a concrete projection of multiple social pressures at the neurobiological level. Because early symptoms of depression are often subtle, early-stage depression is usually difficult to identify promptly and accurately.
[0003] Traditional methods for identifying depressive states based on electrocardiogram signals primarily rely on heart rate variability parameters.
[0004] Compared to subjects in a normal emotional state, subjects in a depressive state showed significantly lower RR interval standard deviation, root mean square of adjacent RR interval interpolation, and high-frequency component proportion, while significantly higher low-frequency component proportion and low-frequency to high-frequency component ratio. With the development of artificial intelligence technology, depressive state recognition technology based on electrocardiogram (ECG) signals and machine learning methods has made some progress. Existing techniques extract time-domain, frequency-domain, and nonlinear features from subjects' ECG sequence signals and input these features into a Bayesian network for training, achieving an accuracy of 86.4% on a private dataset. Other existing techniques collect ECG signals from depressed and healthy subjects during psychological tasks and extract their time-domain, frequency-domain, and nonlinear features to obtain a 100-dimensional heart rate variability vector, which is then trained using a support vector machine, resulting in a model with an accuracy of 74.4%.
[0005] However, traditional machine learning methods rely heavily on rich experience and extensive experiments for feature extraction and selection, and traditional Bayesian, support vector machine, and random forest models have limited spatial mapping capabilities. Therefore, the performance of depressive state recognition based on machine learning and electrocardiogram signals needs to be improved. Summary of the Invention
[0006] The embodiments of this application provide a method and apparatus for emotion state recognition based on electrocardiogram signals and deep learning, so as to at least have the technical effect of combining electrocardiogram signals with deep learning to perform accurate emotion recognition.
[0007] To address the aforementioned technical problems, according to one aspect of this application, a method for emotion state recognition based on electrocardiogram signals and deep learning is provided, comprising:
[0008] Collect standard lead I electrocardiogram signals from the target personnel;
[0009] The standard lead I ECG signal is preprocessed, including low-pass filtering and baseline drift removal to improve the signal-to-noise ratio of the ECG signal;
[0010] An initial electrocardiogram (ECG) signal model based on deep learning is constructed to perform feature encoding, fusion, and decoding of the standard lead I ECG signal;
[0011] A dataset is constructed, which is divided into a training set and a validation set to train the initial ECG signal model. The initial ECG signal model is then tested using a test set to obtain the optimized target ECG signal model.
[0012] The target electrocardiogram (ECG) signal model is used to identify the emotional state of the collected ECG signals.
[0013] According to another aspect of this application, a device for emotion state recognition based on electrocardiogram signals and deep learning is also claimed, comprising:
[0014] The acquisition unit is configured to acquire standard lead I electrocardiogram signals from the target personnel.
[0015] The first processing unit is configured to preprocess the standard lead I ECG signal, including low-pass filtering and baseline drift removal to improve the signal-to-noise ratio of the ECG signal;
[0016] The second processing unit is configured to construct an initial electrocardiogram (ECG) signal model based on deep learning, for feature encoding, fusion, and decoding of the standard lead I ECG signal;
[0017] The third processing unit is configured to construct a dataset, which is divided into a training set and a validation set to train the initial ECG signal model, and a test set to test the initial ECG signal model to obtain an optimized target ECG signal model.
[0018] The determination unit is configured to use the target electrocardiogram signal model to perform emotional state recognition on the acquired electrocardiogram signals.
[0019] According to another aspect of this application, a computer-readable storage medium is also claimed, the computer-readable storage medium including a stored program wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the emotion state recognition method based on electrocardiogram signals and deep learning described above.
[0020] According to another aspect of this application, an electronic device is also claimed, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including methods for performing the emotion state recognition method based on electrocardiogram signals and deep learning as described in any of the preceding claims.
[0021] This application has the following beneficial effects:
[0022] (1) Compared with the traditional depression state recognition model that uses one-dimensional digital electrocardiogram signal as input to deep learning network, the format conversion of the present invention rearranges the "temporal-spatial" features of the one-dimensional digital electrocardiogram signal, which is conducive to the network extracting and fusing features related to depression state recognition.
[0023] (2) Compared with traditional depression state recognition models that use machine learning, this invention automatically extracts features associated with depression state recognition through deep learning networks, thus avoiding the loss of implicit features in the data. Attached Figure Description
[0024] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:
[0025] Figure 1 A flowchart illustrating an emotion state recognition method based on electrocardiogram signals and deep learning according to an embodiment of this application is shown.
[0026] Figure 2 A schematic diagram of a standard lead configuration for digital electrocardiogram (ECG) signal acquisition according to an embodiment of this application is shown.
[0027] Figure 3 A schematic diagram of a digital electrocardiogram signal format conversion process according to an embodiment of this application is shown;
[0028] Figure 4 A schematic diagram of the structure of a depressive state recognition model provided according to an embodiment of this application is shown;
[0029] Figure 5 A schematic diagram of the Fusion Block module structure provided according to an embodiment of this application is shown;
[0030] Figure 6 The model training parameters provided according to embodiments of this application are shown;
[0031] Figure 7A structural block diagram of an apparatus for an emotion state recognition method based on electrocardiogram signals and deep learning, according to an embodiment of this application, is shown. Detailed Implementation
[0032] As introduced in the background section, contemporary psychiatric research (WHO, 2024) shows that persistent psychological stress is the primary environmental factor triggering depressive states. Currently, a large number of individuals are chronically exposed to stressors such as work intensity, family conflicts, or social competition. The accelerated pace of modern society is reshaping the psychological resilience of individuals, leading to a vicious cycle of "stress-neural damage-emotional regulation failure." Depression is no longer simply a medical problem, but a concrete projection of multiple social pressures at the neurobiological level. Because early symptoms of depression are often subtle, early-stage depression is usually difficult to identify promptly and accurately.
[0033] Research has found that when a person is in a depressive state, there are abnormalities in brain activity and neural connections, which in turn affect the autonomic nervous system. The human autonomic nervous system includes the sympathetic and parasympathetic nervous systems, which dynamically coordinate to control the activity of organs such as the heart. Its peripheral efferent nervous system regulates the activity of the myocardium, cardiovascular smooth muscle, and glands, thus forming a physiological signal change chain of "depressive state – changes in EEG characteristics – changes in ECG characteristics," laying the physiological foundation for identifying depressive states based on ECG signals.
[0034] Deep learning-based methods are increasingly attracting researchers' attention, using electrocardiogram (ECG) signals as direct input to deep learning models to automatically extract effective deep-level features. Because deep learning methods eliminate the need for manual feature extraction and selection, some studies have already used ECG signals as input and deep learning models for the diagnosis of arrhythmias and other diseases. Hannun et al. developed a deep neural network (DNN) using single-lead ECG signals as input to classify 12 types of arrhythmias, achieving good classification performance on different datasets. Attia et al. developed a convolutional neural network (CNN) using 12-lead ECG signals as input to diagnose left ventricular dysfunction, achieving an accuracy of 85.7%.
[0035] Although methods based on electrocardiogram (ECG) signals and deep learning have achieved some research results in the diagnosis of heart diseases, demonstrating the feasibility of extracting features from ECG signals using deep learning methods for depressive state identification, specific methods in this area are still rarely reported. An analysis of research on depressive state identification using both deep learning and traditional ECG-based methods reveals several challenges in applying ECG signals and deep learning to identify depressive states.
[0036] First, at the feature extraction level, since depressive states can be identified based on heart rate variability parameters, the features of depressive states reflected in ECG signals have a clear temporal domain nature. Second, since the aforementioned high-frequency and low-frequency features reflect the frequency domain characteristics of ECG signals over a period of time, this indicates that the features of depressive states reflected in ECG signals have a clear transform domain nature. Through inverse transform theory, these features can also be considered as the spatial domain nature of depressive states reflected in ECG signals. In other words, when ECG signals are used as input to a deep learning model, the model must simultaneously possess the ability to extract features in both the spatial and temporal domains. Third, since the spatial and temporal features of ECG signals serve together as identification features for depressive states, the deep learning model must possess a good ability to fuse spatiotemporal features.
[0037] Therefore, to address the existing difficulties in detecting and identifying depressive moods, embodiments of this application provide a method, apparatus, computer, readable storage medium, and electronic device for identifying emotional states based on electrocardiogram signals and deep learning.
[0038] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0039] This embodiment provides a method for emotion state recognition based on electrocardiogram signals and deep learning. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than that shown here.
[0040] Figure 1 This is a schematic diagram illustrating a process for emotion state recognition based on electrocardiogram signals and deep learning, according to an embodiment of this application. Figure 1 As shown, the method includes the following steps:
[0041] Step S101: Collect standard lead I electrocardiogram signals from the target personnel;
[0042] Step S102 involves preprocessing the standard lead I ECG signal, including low-pass filtering and baseline drift removal to improve the signal-to-noise ratio of the ECG signal.
[0043] Step S103: Construct an initial ECG signal model based on deep learning to perform feature encoding, fusion, and decoding of the standard I-lead ECG signal;
[0044] Step S104: Construct a dataset, which is divided into a training set and a validation set to train the initial ECG signal model, and a test set to test the initial ECG signal model to obtain the optimized target ECG signal model.
[0045] Step S105: Use the target ECG signal model to identify the emotional state of the collected ECG signals.
[0046] In one embodiment, S101 includes: an ECG signal acquisition device with standard leads, ECG electrodes connected to the left hand, right hand and right foot of the target person, and right leg drive technology to improve the quality of the ECG signal, wherein the sampling frequency of the ECG signal acquisition device is not less than 250Hz.
[0047] In one embodiment, S104 includes the step of constructing the dataset, which includes: the dataset is used to assess the depressive state of the target personnel by using scales, electroencephalograms, and subject self-reports, and the collected electrocardiogram signals are labeled.
[0048] In one embodiment, S102 includes:
[0049] S1 performs digital low-pass filtering, high-pass filtering, 50Hz notch filtering, and baseline drift removal;
[0050] S2, normalize the standard lead I ECG signal after the above processing;
[0051] S3 performs windowing processing on the normalized signal, dividing it into multiple time segments;
[0052] S4, based on wavelet packet decomposition, performs format conversion processing on the digital electrocardiogram signal in each time segment;
[0053] S5, perform N-level wavelet packet decomposition on the digital ECG signal within a time segment; label the decomposed sub-signals as X1X2...XN, where Xi=L / H, representing the wavelet packet decomposition label when analyzing from the perspective of the filter bank;
[0054] S6. Following the L-priority principle, the decomposed sub-signals are arranged row by row to obtain a 2^N row two-dimensional data.
[0055] It also includes S7, which takes the obtained two-dimensional data as input, uses the judgment of professionals in the dataset as the label of the input data, and trains the initial ECG signal model to obtain the target ECG signal model.
[0056] refer to Figures 2-7 This application also provides an emotion state recognition device based on electrocardiogram signals and deep learning for recognizing depressive emotions, including:
[0057] The acquisition unit is configured to acquire standard lead I electrocardiogram signals from the target personnel.
[0058] The first processing unit is configured to preprocess the standard lead I ECG signal, including low-pass filtering and baseline drift removal to improve the signal-to-noise ratio of the ECG signal;
[0059] The second processing unit is configured to construct an initial electrocardiogram (ECG) signal model based on deep learning, for feature encoding, fusion, and decoding of the standard lead I ECG signal;
[0060] The third processing unit is configured to construct a dataset, which is divided into a training set and a validation set to train the initial ECG signal model, and a test set to test the initial ECG signal model to obtain an optimized target ECG signal model.
[0061] The determination unit is configured to use the target electrocardiogram signal model to perform emotional state recognition on the acquired electrocardiogram signals.
[0062] It should be noted that the emotion state recognition device based on electrocardiogram signals and deep learning in this application embodiment can be used to execute the emotion state recognition method based on electrocardiogram signals and deep learning provided in this application embodiment. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.
[0063] The following describes the emotion state recognition device based on electrocardiogram signals and deep learning provided in the embodiments of this application.
[0064] Step 1: Digital ECG signal acquisition.
[0065] This embodiment uses traditional ECG acquisition equipment. During the acquisition process, the subject can be in a normal sitting or lying position, and other movements should be reduced or stopped. Disposable "Ag-AgCl" electrodes are used on the subject's left hand, right hand, and right leg to establish a stable connection between the ECG acquisition equipment and the subject, as shown in the attached diagram. Figure 1As shown. The sampling rate of the ECG acquisition device was set to 250Hz, the digital ECG signal storage bit depth was set to 12 bits, and the acquisition time was controlled to be no less than 3 minutes. To determine whether the subjects were currently in a depressive state, a portable EEG acquisition device was used simultaneously during the data preparation process to acquire the subjects' digital EEG signals, and the subjects were asked to describe their current state and complete the SCL90 scale. Based on the above information, professionals labeled whether the subjects were currently in a depressive state.
[0066] Step 2: Digital ECG signal preprocessing.
[0067] From the acquired digital electrocardiogram (ECG) signal data, a 3-minute duration with relatively weak baseline drift was selected as the ECG signal for subsequent processing. A linear-phase finite-length digital low-pass filter was used to process the digital ECG signal, with a 3dB cutoff frequency set to 5Hz and a stopband attenuation set to 40dB. A linear-phase finite-length digital high-pass filter was used to process the digital ECG signal, with a 3dB cutoff frequency set to 0.01Hz and a stopband attenuation set to 40dB. A linear-phase finite-length 50Hz digital notch filter was used to process the digital ECG signal, with an attenuation set to 40dB. A quadratic median filter was used to remove baseline drift from the digital ECG signal. After the above processing, the processed digital ECG signal was normalized to normalize its amplitude to the range [1,0].
[0068] Step 3: Windowing processing of digital electrocardiogram signals.
[0069] The digital ECG signal, after preprocessing in step two, was windowed sequentially into multiple time segments. The window width was set to 6 seconds, and the step size to 3 seconds. After windowing, each time segment contained 1500 points of digital ECG signal. To facilitate wavelet packet decomposition, zero-padding was performed at the end of the digital ECG signal in each time segment. After zero-padding, the signal length within each time segment was 2048 points.
[0070] Step 4: Format conversion of digital electrocardiogram signals.
[0071] The digital electrocardiogram (ECG) signal within each time segment is used as the processing object, and wavelet packet decomposition is employed for processing. The complete format conversion process is attached. Figure 2 As shown. In this embodiment, the Daubechies mother wavelet function is used to perform a three-level wavelet packet decomposition on the digital electrocardiogram signal, that is, the signals in both the low-scale and high-scale signal spaces after a certain decomposition are subjected to the next round of wavelet decomposition. Referring to the wavelet analysis method based on filter banks, the sub-signals after the three-level wavelet packet decomposition are labeled, and the labels of the eight sub-signals obtained in sequence are as follows:
[0072] The sub-signals are named "LLL,LLH,LHL,LHH,HLL,HLH,HHL,HHH", each with a data length of 256 points. Following the "L" priority principle, the decomposed sub-signals are arranged row by row to obtain a new data format similar to a "two-dimensional image". For example, the LLL sub-signal after wavelet packet decomposition is the first row of the "two-dimensional image", the LHL sub-signal is the third row, and the HHH sub-signal is the last row. After format conversion, a new data format resembling a "two-dimensional image" with a resolution of "8×256" is obtained.
[0073] Step 5: Construct a model for recognizing depressive states.
[0074] In this embodiment, ResNet18 is used as the backbone network, and the overall network structure is shown in the attached figure. Figure 3 As shown in the figure. To better achieve the extraction and fusion of spatiotemporal features of ECG signals, this embodiment first replaces some Basic Blocks in ResNet18 with redesigned Fusion Blocks, the structure of which is shown in the attached figure. Figure 4 As shown in the diagram. In Fusion Block, this embodiment uses the CBAM mechanism to replace traditional residual connections to optimize cross-layer information transfer. Traditional residual connections can effectively alleviate the gradient vanishing problem in deep network training, while promoting cross-layer information flow, enabling the network to learn features better. However, residual connections simply pass the input from the previous layer to the subsequent layer, avoiding the loss of important features, but they cannot filter and enhance key features. In contrast, CBAM introduces channel attention and spatial attention mechanisms. Both attention mechanisms involve pooling and generate a probability tensor of the same size as the output, representing the importance of features. This probability tensor is dot-producted with the input matrix, allowing the model to autonomously focus on important features. Channel attention filters out primary and secondary channels by compressing and restoring the channel dimension and reallocates weights to highlight key features; spatial attention explicitly establishes connections between pixels by applying a larger convolution kernel to the entire feature map, highlighting key regions in the feature map and thus improving feature representation. Compared to residual connections, CBAM retains the ability to transfer information across layers while filtering features, enabling the network to learn and utilize information more efficiently and improving performance.
[0075] For the dataset used to train the depression recognition model in this embodiment, after data cleaning, it consists of 1794 data points from 50 subjects processed according to steps 1-4 of the present invention. Of these, 15 subjects were identified as depressed, and 35 were identified as not depressed. A total of 580 data points were labeled as depressed, and the remaining 1214 data points were labeled as not depressed. The subjects were divided into training, validation, and test sets in an 8:1:1 ratio. In the test set, 58 data points were labeled as depressed, and 121 data points were labeled as not depressed. During training, weighted cross-entropy loss was used to calculate the loss value, and the AdamW optimizer was used to update the model parameters. The weight of each category in the loss function is the reciprocal of the normalized count of each category; that is, the larger the sample size of a certain category, the smaller its loss weight. The parameters involved in model training are shown in the appendix. Figure 5 As shown.
[0076] After training, the depression recognition model in this example correctly identified 53 out of 58 data points labeled as depressed, and 111 out of 121 data points labeled as non-depressed. The metrics used to evaluate the depression recognition model are as follows:
[0077] accuracy
[0078]
[0079] Accuracy
[0080]
[0081] Recall rate
[0082]
[0083] F1 score
[0084]
[0085] The aforementioned emotion state recognition method and apparatus based on electrocardiogram signals and deep learning includes a processor and a memory. The acquisition unit, first processing unit, second processing unit, third processing unit, and determination unit are all stored as program units in the memory. The processor executes these program units stored in the memory to achieve the corresponding functions. All of the above modules are located in the same processor; alternatively, the modules may be located in different processors in any combination.
[0086] The processor contains a kernel, which retrieves the corresponding program unit from memory. One or more kernels can be configured, and adjusting kernel parameters can address issues such as measurement inaccuracies or discrepancies among target personnel.
[0087] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.
[0088] This invention provides a computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device containing the computer-readable storage medium to perform the emotion state recognition method based on electrocardiogram signals and deep learning.
[0089] This invention provides a processor for running a program, wherein the program executes the above-mentioned emotion state recognition method based on electrocardiogram signals and deep learning.
[0090] This application also provides an electronic device, which includes: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors.
[0091] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.
[0092] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0093] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0094] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0095] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0096] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0097] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0098] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0099] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.
[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A method for emotion state recognition based on electrocardiogram signals and deep learning, characterized in that, include: Collect standard lead I electrocardiogram signals from the target personnel; The standard lead I ECG signal is preprocessed, including low-pass filtering and baseline drift removal to improve the signal-to-noise ratio of the ECG signal; An initial electrocardiogram (ECG) signal model based on deep learning is constructed to perform feature encoding, fusion, and decoding of the standard lead I ECG signal; A dataset is constructed, which is divided into a training set and a validation set to train the initial ECG signal model. The initial ECG signal model is then tested using a test set to obtain the optimized target ECG signal model. The target electrocardiogram (ECG) signal model is used to identify the emotional state of the collected ECG signals.
2. The emotion state recognition method based on electrocardiogram signals and deep learning according to claim 1, characterized in that, The specific steps for collecting standard lead I electrocardiogram signals from the target personnel include: The ECG signal acquisition device uses standard leads and connects ECG electrodes to the target person's left hand, right hand, and right foot. Right leg drive technology is used to improve the quality of the ECG signal. The sampling frequency of the ECG signal acquisition device is not less than 250Hz.
3. The emotion state recognition method based on electrocardiogram signals and deep learning according to claim 1, characterized in that, The steps for constructing the dataset include: using scales, electroencephalograms (EEGs), and self-reported information from the subjects to assess the subjects' depressive state, and labeling the collected electrocardiogram signals.
4. The emotion state recognition method based on electrocardiogram signals and deep learning according to claim 1, characterized in that, The steps for preprocessing the standard lead I ECG signal include: Perform digital low-pass filtering, high-pass filtering, 50Hz notch filtering, and baseline drift removal; The standard lead I ECG signal after the above processing is normalized. The normalized signal is then windowed and divided into multiple time segments. Based on wavelet packet decomposition, the digital electrocardiogram signal in each time segment is processed for format conversion. Perform N-level wavelet packet decomposition on a digital electrocardiogram signal within a time segment; label the decomposed sub-signals as X1X2...XN, where Xi=L / H represents the wavelet packet decomposition label when analyzing from the perspective of the filter bank; Following the L-priority principle, the decomposed sub-signals are arranged row by row to obtain a two-dimensional data with 2^N rows.
5. The emotion state recognition method based on electrocardiogram signals and deep learning according to claim 4, characterized in that, It also includes using the obtained two-dimensional data as input, and using the judgments of professionals in the dataset as labels for the input data, to train the initial electrocardiogram signal model to obtain the target electrocardiogram signal model.
6. An emotion state recognition device based on electrocardiogram signals and deep learning, characterized in that, include: The acquisition unit is configured to acquire standard lead I electrocardiogram signals from the target personnel. The first processing unit is configured to preprocess the standard lead I ECG signal, including low-pass filtering and baseline drift removal to improve the signal-to-noise ratio of the ECG signal; The second processing unit is configured to construct an initial electrocardiogram (ECG) signal model based on deep learning, for feature encoding, fusion, and decoding of the standard lead I ECG signal; The third processing unit is configured to construct a dataset, which is divided into a training set and a validation set to train the initial ECG signal model, and a test set to test the initial ECG signal model to obtain an optimized target ECG signal model. The determination unit is configured to use the target electrocardiogram signal model to perform emotional state recognition on the acquired electrocardiogram signals.
7. The emotion state recognition device based on electrocardiogram signals and deep learning according to claim 6, characterized in that, The ECG signal is acquired using a standard lead ECG acquisition device or a portable ECG acquisition device, with the portable ECG acquisition device outputting a standard lead I ECG signal.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein when the program is executed, it controls the device containing the computer-readable storage medium to perform the emotion state recognition method based on electrocardiogram signals and deep learning as described in any one of claims 1 to 5.
9. An electronic device, characterized in that, The method includes: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include methods for performing the emotion state recognition method based on electrocardiogram signals and deep learning as described in any one of claims 1 to 5.