Identity recognition method, device, equipment, medium and program product
By acquiring the current electrocardiogram (ECG) signal and extracting identity features, and combining time intervals and training models, the problem of poor stability across time periods in traditional ECG recognition methods is solved, thereby improving the accuracy and stability of identity recognition.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-29
- Publication Date
- 2026-04-03
AI Technical Summary
Traditional electrocardiogram (ECG) identification methods suffer from poor stability across time periods, which affects the accuracy of identity verification.
By acquiring the current electrocardiogram (ECG) signal, extracting the target's identity features, and determining the target's time features based on the time interval, combined with a trained identity recognition model, the target object's identity type is predicted. By using ECG signals from the target's acquisition time period as training samples, the impact of time drift is reduced.
It improves the accuracy of identity recognition in scenarios with cross-time period changes, reduces the impact of time drift, and achieves more stable identity recognition.
Smart Images
Figure CN121786749A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of identity recognition technology, and in particular to an identity recognition method, apparatus, device, medium, and program product. Background Technology
[0002] With the continuous development of identity recognition technology, identity recognition has gradually extended from external physiological features such as fingerprints and faces to recognition based on physiological signals such as electrocardiograms.
[0003] However, traditional ECG recognition methods often suffer from stability issues across different time periods. This is because various factors, such as the user's state, emotions, posture, sensor position, and electrode contact, can affect the waveform of even the same user's ECG signal at different times. This temporal drift phenomenon can easily affect the accuracy of identification. Summary of the Invention
[0004] Therefore, it is necessary to provide an identity recognition method, device, equipment, medium, and program product to address the aforementioned technical problems, thereby improving the accuracy of identity recognition.
[0005] Firstly, this application provides an identity verification method, including:
[0006] Obtain the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target's identity features from the current ECG signal;
[0007] The target time characteristic is determined based on the time interval between the current data collection period and the target data collection period; the target data collection period is earlier than the current data collection period.
[0008] Based on the target's identity characteristics and target's time characteristics, determine the predicted identity characteristics of the target object within the target data collection period;
[0009] Based on the trained identity recognition model, the target identity type of the target object is determined according to the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected from the identity authentication object during the target collection time period.
[0010] In one embodiment, the target time feature is determined based on the time interval between the current acquisition time period and the target acquisition time period, including: obtaining the target acquisition time period corresponding to the electrocardiogram signal acquired by the target object before the current acquisition time period; determining the time interval between the current acquisition time period and the target acquisition time period; and performing feature encoding processing on the time interval to obtain the target time feature.
[0011] In one embodiment, the target time feature includes time feature elements of at least one encoded dimension; correspondingly, feature encoding processing is performed on the time interval to obtain the target time feature, including: for each encoded dimension, determining an objective function with the encoded dimension and time interval as independent variables and the time feature elements as dependent variables according to the parity of the encoded dimension; and based on the objective function, determining the time feature elements under the encoded dimension according to the encoded dimension and time interval.
[0012] In one embodiment, determining the predicted identity features of a target object within a target acquisition time period based on the target identity features and the target time features includes: fusing the target identity features and the target time features to predict the initial predicted identity features of the target object within the target acquisition time period; reconstructing the initial predicted identity features to obtain a reconstructed signal; and extracting the predicted identity features from the reconstructed signal.
[0013] In one embodiment, extracting target identity features from the current electrocardiogram (ECG) signal includes: determining the target heartbeat cycle based on the current ECG signal; extracting at least one heartbeat segment from the current ECG signal according to the target heartbeat cycle; normalizing the at least one heartbeat segment to obtain a summary heartbeat signal; and extracting target identity features from the summary heartbeat signal.
[0014] In one embodiment, after determining the target identity type of the target object, the method further includes: retraining the identity recognition model based on the current electrocardiogram signal within the current collection time period; and updating the target collection time period based on the current collection time period.
[0015] Secondly, this application also provides an identity recognition device, comprising:
[0016] The extraction module is used to acquire the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target's identity features from the current ECG signal;
[0017] The first determining module is used to determine the target time characteristics based on the time interval between the current collection time period and the target collection time period; the target collection time period is earlier than the current collection time period.
[0018] The second determination module is used to determine the predicted identity features of the target object within the target collection time period based on the target identity features and the target time features.
[0019] The third determination module is used to determine the target identity type of the target object based on the trained identity recognition model and the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected from the identity authentication object during the target collection time period.
[0020] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0021] Obtain the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target's identity features from the current ECG signal;
[0022] The target time characteristic is determined based on the time interval between the current data collection period and the target data collection period; the target data collection period is earlier than the current data collection period.
[0023] Based on the target's identity characteristics and target's time characteristics, determine the predicted identity characteristics of the target object within the target data collection period;
[0024] Based on the trained identity recognition model, the target identity type of the target object is determined according to the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected from the identity authentication object during the target collection time period.
[0025] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:
[0026] Obtain the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target's identity features from the current ECG signal;
[0027] The target time characteristic is determined based on the time interval between the current data collection period and the target data collection period; the target data collection period is earlier than the current data collection period.
[0028] Based on the target's identity characteristics and target's time characteristics, determine the predicted identity characteristics of the target object within the target data collection period;
[0029] Based on the trained identity recognition model, the target identity type of the target object is determined according to the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected from the identity authentication object during the target collection time period.
[0030] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:
[0031] Obtain the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target's identity features from the current ECG signal;
[0032] The target time characteristic is determined based on the time interval between the current data collection period and the target data collection period; the target data collection period is earlier than the current data collection period.
[0033] Based on the target's identity characteristics and target's time characteristics, determine the predicted identity characteristics of the target object within the target data collection period;
[0034] Based on the trained identity recognition model, the target identity type of the target object is determined according to the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected from the identity authentication object during the target collection time period.
[0035] The aforementioned identity recognition methods, devices, equipment, media, and program products acquire the current electrocardiogram (ECG) signal of the target object within the current acquisition time period and extract target identity features from the current ECG signal, thus providing a data foundation for the subsequent identity recognition process. By determining the target time feature based on the time interval between the current acquisition time period and the target acquisition time period, the time interval between the two acquisition moments is quantified as the time variation difference between different acquisition moments, facilitating the subsequent perception of predicted identity features under corresponding time span changes by combining the target time feature. By determining the predicted identity feature of the target object within the target acquisition time period based on the target identity feature and the target time feature, the theoretical identity feature that the target object should present within the target acquisition time period can be estimated or predicted. Based on a trained identity recognition model, the target identity type of the target object is determined according to the predicted identity feature. Since the training samples of the identity recognition model include ECG signals collected from the identity authentication object within the target acquisition time period, the impact of time drift can be reduced and the accuracy of identity recognition improved in cross-time period change scenarios. Attached Figure Description
[0036] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0037] Figure 1A This is a flowchart illustrating an identity recognition method in one embodiment;
[0038] Figure 1B This is a schematic diagram of the processing chain of an identity recognition method in one embodiment;
[0039] Figure 2 This is a flowchart illustrating the steps for determining the target time feature in one embodiment;
[0040] Figure 3 This is a schematic diagram of a cascaded reconstruction structure in one embodiment;
[0041] Figure 4This is a flowchart illustrating the identity recognition method in another embodiment;
[0042] Figure 5 This is a structural block diagram of an identity recognition device in one embodiment;
[0043] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0044] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0045] In one embodiment, such as Figure 1A As shown, an identity recognition method is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. The terminal may include a personal computer, tablet computer, mobile terminal, or edge computing platform. This application does not limit the specific type of terminal. In this embodiment, the method includes the following steps:
[0046] S110. Obtain the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target identity features from the current ECG signal.
[0047] Here, the current ECG signal can be understood as the ECG signal of the target object collected within the current acquisition time period. Target identity features can be understood as the features extracted from the current ECG signal used to characterize the identity of the target object.
[0048] In an optional embodiment, the raw electrocardiogram (ECG) signal of the target object during the current acquisition time period can be extracted, and the raw ECG signal can be preprocessed to obtain the current ECG signal.
[0049] Optionally, preprocessing may include at least one of resampling, bandpass filtering, and baseline drift removal. Specifically, preprocessing the raw ECG signal aims to obtain a smooth and standardized current ECG signal. The raw ECG signal may be acquired using different acquisition devices, and the number of raw ECG signals may be at least one.
[0050] For example, the original electrocardiogram signal can be resampled based on a preset sampling rate. The preset sampling rate can be 500Hz.
[0051] For example, bandpass filtering can be performed according to a preset frequency range. The preset frequency range can be from 0.5Hz to 40Hz.
[0052] For example, baseline drift and high-frequency interference in the original ECG signal can be removed, and transient noise can be eliminated based on East China mean filtering, thereby obtaining a smooth and standardized current ECG signal.
[0053] It should be noted that the above-mentioned preset sampling rate and preset frequency range can be set by technicians according to their needs or experience, or determined through a large number of experiments. This application does not impose any restrictions on them.
[0054] In an optional embodiment, the current electrocardiogram (ECG) signal can be input into an identity feature extraction model to extract the target identity features from the current ECG signal.
[0055] In another optional embodiment, the target heartbeat cycle can be determined based on the current ECG signal; at least one heartbeat segment can be extracted from the current ECG signal according to the target heartbeat cycle; the at least one heartbeat segment can be normalized to obtain a summary heartbeat signal; and the target identity features can be extracted from the summary heartbeat signal.
[0056] Optionally, R-wave detection can be used to identify the target heartbeat cycle in the current ECG signal. The R-wave can be understood as the most prominent and highest-amplitude positive wave in the current ECG signal, and is a key wave used to identify features such as heart rate changes.
[0057] Optionally, at least one R wave can be acquired based on the target heartbeat cycle; for each R wave, a heartbeat segment is extracted centered on the R wave and within a sampling point window of a preset fixed length. The preset fixed length can be set by a technician based on needs or experience, or determined through extensive experimentation; this application does not impose any limitations on it. For example, the preset fixed length can be the sampling length corresponding to 400 sampling points.
[0058] Optionally, at least one heartbeat segment can be amplitude normalized and then averaged to obtain a summary heartbeat signal representing the identity characteristics of the target object. By normalizing the amplitude of at least one heartbeat segment and then averaging it, the influence of heart rate variations and noise can be reduced.
[0059] Optionally, an autoencoder can be used to encode the summary heartbeat signal. The autoencoder may include four convolutional units with 8, 32, 128 and 512 channels respectively. The convolutional units are used to progressively extract identity features to extract the target identity features from the summary heartbeat signal.
[0060] S120. Determine the target time characteristics based on the time interval between the current collection time period and the target collection time period; the target collection time period is earlier than the current collection time period.
[0061] The target acquisition time period can be understood as the acquisition time period of the electrocardiogram signal of the identity authentication object in the training sample of the identity recognition model.
[0062] For example, the target collection time period can be obtained by inputting the target object, or it can be a predefined collection time period. This application does not limit the specific method of obtaining the target collection time period.
[0063] In an optional embodiment, a location encoding vector can be constructed based on a time interval; the location encoding vector can then be used as a target temporal feature.
[0064] Optionally, the time interval can be mapped based on sine and cosine functions to obtain a location encoding vector; whereby the location encoding vector is used to characterize the time features across time periods, i.e., the target time features.
[0065] S130. Based on the target identity characteristics and target time characteristics, determine the predicted identity characteristics of the target object within the target collection time period.
[0066] Among them, the target identity feature and the target time feature can be represented by vectors, that is, the target identity feature can be a target identity feature vector, and the target time feature can be a target time feature vector.
[0067] In an optional embodiment, the target identity features and target time features can be input into a first fusion model to determine the predicted identity features of the target object within the target collection time period. The first fusion model can be a machine learning model or a neural network model, and this application does not impose any limitations on it.
[0068] In another optional embodiment, the target identity features and target time features can be fused to predict the initial predicted identity features of the target object within the target acquisition time period; the initial predicted identity features are reconstructed to obtain a reconstructed signal; and the predicted identity features are extracted from the reconstructed signal.
[0069] The initial predicted identity features may include the identity information of the target individual and information on changes over time, providing a data foundation for subsequent reconstruction and identification.
[0070] Optionally, the target identity features and target time features can be input into a second fusion model to determine the initial predicted identity features of the target object within the target collection time period. The second fusion model can be a machine learning model or a neural network model; this application does not impose any limitations on it. It is understood that during the fusion processing of the target identity features and target time features, the target time features serve as a time modulation factor for the target identity features, enabling the prediction of the identity features corresponding to the target collection time period.
[0071] Optionally, based on the first decoder, the initial predicted identity features can be reconstructed to obtain a reconstructed signal; the reconstructed signal can then be encoded using a re-encoder to extract latent identity features, i.e., predicted identity features. Understandably, by reconstructing the initial predicted identity features, the time-aware fused features can be written back to an approximate heartbeat; by encoding the reconstructed signal using a re-encoder, temporal perturbations can be removed to obtain the predicted identity features.
[0072] S140. Based on the trained identity recognition model, determine the target identity type of the target object according to the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected by the identity authentication object during the target collection time period.
[0073] The identity recognition model is used to classify and determine the target identity type. The identity recognition model can be a traditional machine learning model or a neural network model; this application makes no limitation on either.
[0074] The target identity type may include at least one of authenticated identity types and unauthenticated identity types. This application does not impose any specific limitations on the target identity type.
[0075] In this context, the identity authentication object can be understood as an individual whose identity has been clearly labeled or confirmed during the model training phase, and the electrocardiogram signal corresponding to the identity authentication object is used as a training sample to build the identity recognition model; the target object can be understood as the individual whose identity is to be recognized or verified.
[0076] In an optional embodiment, the identity recognition model can be retrained based on the current ECG signal within the current acquisition time period; the target acquisition time period can then be updated using the current acquisition time period. Specifically, during the retraining process, corresponding predicted identity features can be determined based on the current ECG signal within the current acquisition time period. These predicted identity features are obtained by fusing the target identity features and target time features corresponding to the current ECG signal. The identity recognition model is then retrained based on the predicted identity features and the target identity type label.
[0077] In an optional embodiment, the identity recognition model may include a multilayer perceptron (MLP) structure. The MLP may be a three-layer perceptron; the first and second layers of the three-layer perceptron may be fully connected layers, employing a LeakyReLU (Leaky Rectified Linear Unit) activation function; the third layer outputs an identity category probability distribution through a Softmax function (normalized exponential function). This identity category probability distribution characterizes the target identity type with the highest classification probability.
[0078] For example, during the training process of an identity recognition model, a joint loss function can be used to simultaneously optimize the reconstruction and classification tasks. The loss function is defined as follows:
[0079]
[0080] in, To compensate for the reconstruction error loss, the mean square error between the output signal (i.e., the reconstructed signal) and the original signal (i.e., the summary heartbeat signal) is calculated. Cross-entropy loss for identity classification; and is the weighting coefficient, used to balance feature stability and classification performance.
[0081] Understandably, the loss function described above includes a reconstruction loss term. and classification loss items By performing dynamic weight coefficient balancing reconstruction and classification tasks, feature discrimination and time stability can be optimized simultaneously. Thus, under this joint optimization strategy, an identity determination model that can maintain robustness under cross-time conditions is obtained, namely the ECG identity authentication model.
[0082] For example, during the training process of an identity recognition model, the predicted identity features can be reconstructed based on a second decoder to obtain the final reconstructed signal. Reconstruction error loss. This can be used to calculate the mean square error between the final reconstructed signal and the original signal. Understandably, under the aforementioned cascaded reconstruction mechanism—that is, the process of "decoding-re-encoding-re-decoding"—consistency constraints on the feature space of samples across time periods can be achieved, enabling the model to maintain a stable representation of identity features under different acquisition conditions. Simultaneously, through decoding and re-encoding, the final predicted identity features can be recognized by the identity recognition model, which helps improve the stability of identity recognition.
[0083] In another embodiment, the predicted identity features and the target identity features can be input into a trained identity recognition model to determine the target identity type of the target object.
[0084] refer to Figure 1B The diagram shows the processing flow of an identity recognition method. Among them, Figure 1B The document illustrates an identity recognition method that may include heartbeat segmentation and signal optimization, feature extraction and fusion, and an identity recognition process. The heartbeat segmentation and signal optimization process may include acquiring an ECG (electrocardiogram) signal; performing preprocessing, heartbeat segmentation, and signal optimization on the ECG signal sequentially to obtain a summary heartbeat signal. The feature extraction and fusion process may include extracting identity features from the summary heartbeat signal to obtain target identity features; acquiring the current acquisition period corresponding to the summary heartbeat signal; extracting time features based on the time interval between the current acquisition period and the target acquisition period to obtain target time features; fusing the target identity features and target time features to obtain predicted identity features. Finally, identity recognition is performed on the predicted identity features, and an authentication result is output, including the target identity type of the target object.
[0085] The aforementioned identity recognition method acquires the target object's current electrocardiogram (ECG) signal within the current acquisition time period and extracts the target identity features from the current ECG signal, thus providing a data foundation for the subsequent identity recognition process. By determining the target time feature based on the time interval between the current acquisition time period and the target acquisition time period, the time interval between the two acquisition moments is quantified as the time variation difference between different acquisition moments. This facilitates the subsequent perception of predicted identity features under corresponding time span changes by combining the target time feature. By determining the predicted identity features of the target object within the target acquisition time period based on the target identity feature and the target time feature, the theoretical identity features that the target object should exhibit during the target acquisition time period can be estimated or predicted. Based on a trained identity recognition model, the target identity type of the target object is determined according to the predicted identity features. Since the training samples of the identity recognition model include ECG signals collected from the identity verification object within the target acquisition time period, the impact of time drift can be reduced in cross-time period scenarios, improving the accuracy of identity recognition.
[0086] Among them, the above-mentioned identity recognition method introduces target time features to provide a data foundation for the subsequent elimination of time drift; the fusion of target time features and target identity features and the cascaded reconstruction process are mainly used to eliminate drift; by introducing dynamic joint loss, feature discrimination and time stability can be adaptively optimized in sync; the three work together in a closed loop to form a complete and robust recognition framework aimed at determining identity type based on electrocardiogram signals in cross-time period scenarios.
[0087] Based on the above embodiments, this application also provides an optional embodiment in which the steps for determining the target time features are refined.
[0088] refer to Figure 2 The steps for determining the target time features are shown below, including:
[0089] S210. Obtain the target acquisition time period corresponding to the electrocardiogram signals acquired by the target object before the current acquisition time period.
[0090] S220. Determine the time interval between the current data collection period and the target data collection period.
[0091] S230. Perform feature encoding on the time interval to obtain the target time feature.
[0092] In an optional embodiment, the target time feature includes time feature elements of at least one encoding dimension; correspondingly, for each encoding dimension, an objective function can be determined based on the parity of the encoding dimension, with the encoding dimension and time interval as independent variables and the time feature elements as dependent variables; based on the objective function, the time feature elements under the encoding dimension are determined according to the encoding dimension and time interval.
[0093] The encoding dimension is used to represent the encoding length, meaning that the target temporal feature can include the total number of temporal feature elements in the encoding dimension.
[0094] Optionally, when the encoding dimension is odd, the function type of the objective function can be determined to be a cosine function; when the encoding dimension is even, the function type of the objective function can be determined to be a sine function.
[0095] For example, the temporal feature elements in the encoding dimension can be determined based on the following formula:
[0096]
[0097]
[0098] Where T represents the time interval; d represents the total number of encoding dimensions; 2k represents the even-numbered dimension index of the encoding dimension, and 2k+1 represents the odd-numbered dimension index of the encoding dimension.
[0099] The time interval T can be understood as an implicit proxy variable for the change of physiological state over time. The time interval T is mapped to a time modulation vector (i.e., target time feature) through a multi-frequency sine and cosine basis, and the target time feature is used as the time modulation factor of the target identity feature, so as to determine the predicted identity feature of the target object within the target collection time period.
[0100] Understandably, the above formulas can be used to determine the time feature elements of different encoding dimensions, thus mapping time intervals into high-dimensional vectors. Introducing sine and cosine functions to construct a multi-frequency time base is beneficial for improving the ability to perceive changes across time periods.
[0101] Based on the above embodiments, feature fusion and signal reconstruction can be performed based on the cascaded reconstruction structure.
[0102] refer to Figure 3 The diagram shows a cascaded reconstruction structure. This cascaded reconstruction structure can include an upstream encoding fusion module and a cascaded decoding reconstruction module.
[0103] In the upstream encoding and fusion module, the summary heartbeat signal S2 can be acquired. Based on the encoder, identity features are extracted from the summary heartbeat signal S2 to obtain the target identity features. The current acquisition period T2 corresponding to the summary heartbeat signal S2 is acquired, and the time interval T2_T1 between the current acquisition period T2 and the target acquisition period T1 is determined. Positional features are extracted from the time interval T2_T1 to obtain the target time features. The target identity features and the target time features are fused to obtain the initial predicted identity features. Furthermore, after obtaining the target identity features, the target identity features can be reconstructed based on the decoder to obtain the reconstructed signal corresponding to the target identity features, i.e., the second reconstructed signal.
[0104] In the cascaded decoding and reconstruction module, the initial predicted identity features can be reconstructed based on the decoder to obtain the reconstructed signal of the initial predicted identity features, i.e., the initial reconstructed signal S1'; based on the encoder, identity features are extracted from the reconstructed signal to obtain the predicted identity features; based on the classifier (i.e., the identity recognition model), the target identity category is determined according to the predicted identity features. After obtaining the predicted identity features, they can also be reconstructed based on the decoder to obtain the corresponding reconstructed signal, i.e., the first reconstructed signal S1.
[0105] Specifically, the reconstruction loss term loss_m can be determined by combining the relationship between the summary heartbeat signal and the second reconstruction signal, as well as the relationship between the initial reconstruction signal and the first reconstruction signal; the classification loss term loss_c can be determined by combining the output of the classifier and the target identity features.
[0106] Based on the above embodiments, this application also provides an optional embodiment in which the identity recognition method is described in detail.
[0107] refer to Figure 4 The diagram shown is a flowchart of an identity recognition method in another embodiment, including the following steps:
[0108] S401. Obtain the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and determine the target heart rate cycle based on the current ECG signal.
[0109] S402. According to the target heartbeat cycle, extract at least one heartbeat segment from the current electrocardiogram signal.
[0110] S403. Normalize at least one heartbeat segment to obtain a summary heartbeat signal, and extract the target identity features from the summary heartbeat signal.
[0111] S404. Obtain the target acquisition time period corresponding to the electrocardiogram signals acquired by the target object before the current acquisition time period.
[0112] S405. Determine the time interval between the current data collection period and the target data collection period.
[0113] S406. For each coding dimension, determine the objective function with the coding dimension and time interval as independent variables and the time feature element as the dependent variable, based on the parity of the coding dimension.
[0114] The target time feature includes time feature elements of at least one encoded dimension.
[0115] S407. Based on the objective function, determine the time feature elements under the encoding dimension according to the encoding dimension and time interval.
[0116] S408. Perform feature encoding on the time interval to obtain the target time feature.
[0117] S409. The target identity features and target time features are fused to predict the initial predicted identity features of the target object within the target collection time period.
[0118] S410. Reconstruct the signal from the initial predicted identity features to obtain the reconstructed signal.
[0119] S411. Extract predicted identity features from the reconstructed signal.
[0120] S412. Based on the trained identity recognition model, determine the target identity type of the target object according to the predicted identity features.
[0121] The training samples for the identity recognition model include electrocardiogram (ECG) signals collected from the identity-authenticating object during the target collection time period.
[0122] In some embodiments, the above-mentioned identity recognition method was validated based on a public database. The recognition rate reached 96.53% under non-cross-time conditions; 85.29% under cross-time conditions of 30 to 120 days; and 76.91% under cross-time conditions exceeding 120 days. This represents an improvement of over 30% compared to traditional baseline methods. These results validate the effectiveness and robustness of the above-mentioned identity recognition method in long-term scenarios.
[0123] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0124] Based on the same inventive concept, this application also provides an identity recognition device for implementing the aforementioned identity recognition method. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more identity recognition device embodiments provided below can be found in the limitations of the identity recognition method described above, and will not be repeated here.
[0125] In one exemplary embodiment, such as Figure 5 As shown, an identity recognition device is provided, including: an extraction module 510, a first determination module 520, a second determination module 530, and a third determination module 540, wherein:
[0126] Extraction module 510 is used to acquire the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target identity features from the current ECG signal;
[0127] The first determining module 520 is used to determine the target time characteristics based on the time interval between the current collection time period and the target collection time period; the target collection time period is earlier than the current collection time period.
[0128] The second determining module 530 is used to determine the predicted identity features of the target object within the target collection time period based on the target identity features and the target time features.
[0129] The third determination module 540 is used to determine the target identity type of the target object based on the trained identity recognition model and the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected by the identity authentication object during the target collection time period.
[0130] In one embodiment, the first determining module 520 includes: a first acquiring unit, used to acquire the target acquisition time period corresponding to the electrocardiogram signal acquired by the target object before the current acquisition time period; a first determining unit, used to determine the time interval between the current acquisition time period and the target acquisition time period; and an encoding unit, used to perform feature encoding processing on the time interval to obtain the target time feature.
[0131] In one embodiment, the target time feature includes time feature elements of at least one encoding dimension; correspondingly, the encoding unit includes: a first determining subunit, configured to determine, for each encoding dimension, an objective function with the encoding dimension and time interval as independent variables and the time feature elements as dependent variables, based on the parity of the encoding dimension; and a second determining subunit, configured to determine the time feature elements under the encoding dimension based on the objective function, according to the encoding dimension and time interval.
[0132] In one embodiment, the second determining module 530 includes: a fusion unit for fusing target identity features and target time features to predict the initial predicted identity features of the target object within the target acquisition time period; a reconstruction unit for reconstructing the initial predicted identity features to obtain a reconstructed signal; and a first extraction unit for extracting the predicted identity features from the reconstructed signal.
[0133] In one embodiment, the extraction module 510 includes: a second determining unit, configured to determine the target heartbeat cycle based on the current electrocardiogram signal; a truncating unit, configured to truncate at least one heartbeat segment from the current electrocardiogram signal according to the target heartbeat cycle; a processing unit, configured to normalize the at least one heartbeat segment to obtain a summary heartbeat signal; and a second extraction unit, configured to extract target identity features from the summary heartbeat signal.
[0134] In one embodiment, the third determining module 540 includes: a training unit, configured to retrain the identity recognition model based on the current electrocardiogram signal within the current acquisition time period; and an updating unit, configured to update the target acquisition time period based on the current acquisition time period.
[0135] Each module in the aforementioned identity recognition device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0136] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements an identification method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.
[0137] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0138] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0139] Obtain the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target's identity features from the current ECG signal;
[0140] The target time characteristic is determined based on the time interval between the current data collection period and the target data collection period; the target data collection period is earlier than the current data collection period.
[0141] Based on the target's identity characteristics and target's time characteristics, determine the predicted identity characteristics of the target object within the target data collection period;
[0142] Based on the trained identity recognition model, the target identity type of the target object is determined according to the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected from the identity authentication object during the target collection time period.
[0143] In one embodiment, when the processor executes the computer program, it further performs the following steps: obtaining the target acquisition time period corresponding to the electrocardiogram signal acquired by the target object before the current acquisition time period; determining the time interval between the current acquisition time period and the target acquisition time period; and performing feature encoding processing on the time interval to obtain the target time feature.
[0144] In one embodiment, the target time feature includes time feature elements of at least one encoded dimension; correspondingly, when the processor executes the computer program, it further implements the following steps: for each encoded dimension, determining an objective function with the encoded dimension and time interval as independent variables and the time feature elements as dependent variables according to the parity of the encoded dimension; and determining the time feature elements under the encoded dimension based on the objective function, according to the encoded dimension and time interval.
[0145] In one embodiment, when the processor executes the computer program, it further performs the following steps: fusing the target identity features and the target time features to predict the initial predicted identity features of the target object within the target acquisition time period; reconstructing the initial predicted identity features to obtain a reconstructed signal; and extracting the predicted identity features from the reconstructed signal.
[0146] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining the target heartbeat cycle based on the current electrocardiogram (ECG) signal; extracting at least one heartbeat segment from the current ECG signal according to the target heartbeat cycle; normalizing the at least one heartbeat segment to obtain a summary heartbeat signal; and extracting the target identity features from the summary heartbeat signal.
[0147] In one embodiment, when the processor executes the computer program, it further performs the following steps: retraining the identity recognition model based on the current electrocardiogram signal within the current acquisition time period; and updating the target acquisition time period based on the current acquisition time period.
[0148] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:
[0149] Obtain the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target's identity features from the current ECG signal;
[0150] The target time characteristic is determined based on the time interval between the current data collection period and the target data collection period; the target data collection period is earlier than the current data collection period.
[0151] Based on the target's identity characteristics and target's time characteristics, determine the predicted identity characteristics of the target object within the target data collection period;
[0152] Based on the trained identity recognition model, the target identity type of the target object is determined according to the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected from the identity authentication object during the target collection time period.
[0153] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: obtaining the target acquisition time period corresponding to the electrocardiogram signal acquired by the target object before the current acquisition time period; determining the time interval between the current acquisition time period and the target acquisition time period; and performing feature encoding processing on the time interval to obtain the target time feature.
[0154] In one embodiment, the target time feature includes time feature elements of at least one encoded dimension; correspondingly, when the computer program is executed by the processor, it further implements the following steps: for each encoded dimension, determining an objective function with the encoded dimension and time interval as independent variables and the time feature elements as dependent variables according to the parity of the encoded dimension; and based on the objective function, determining the time feature elements under the encoded dimension according to the encoded dimension and time interval.
[0155] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: fusing the target identity features and the target time features to predict the initial predicted identity features of the target object within the target acquisition time period; reconstructing the initial predicted identity features to obtain a reconstructed signal; and extracting the predicted identity features from the reconstructed signal.
[0156] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the target heartbeat cycle based on the current electrocardiogram (ECG) signal; extracting at least one heartbeat segment from the current ECG signal according to the target heartbeat cycle; normalizing the at least one heartbeat segment to obtain a summary heartbeat signal; and extracting the target identity features from the summary heartbeat signal.
[0157] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: retraining the identity recognition model based on the current electrocardiogram signal within the current acquisition time period; and updating the target acquisition time period based on the current acquisition time period.
[0158] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:
[0159] Obtain the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target's identity features from the current ECG signal;
[0160] The target time characteristic is determined based on the time interval between the current data collection period and the target data collection period; the target data collection period is earlier than the current data collection period.
[0161] Based on the target's identity characteristics and target's time characteristics, determine the predicted identity characteristics of the target object within the target data collection period;
[0162] Based on the trained identity recognition model, the target identity type of the target object is determined according to the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected from the identity authentication object during the target collection time period.
[0163] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: obtaining the target acquisition time period corresponding to the electrocardiogram signal acquired by the target object before the current acquisition time period; determining the time interval between the current acquisition time period and the target acquisition time period; and performing feature encoding processing on the time interval to obtain the target time feature.
[0164] In one embodiment, the target time feature includes time feature elements of at least one encoded dimension; correspondingly, when the computer program is executed by the processor, it further implements the following steps: for each encoded dimension, determining an objective function with the encoded dimension and time interval as independent variables and the time feature elements as dependent variables according to the parity of the encoded dimension; and based on the objective function, determining the time feature elements under the encoded dimension according to the encoded dimension and time interval.
[0165] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: fusing the target identity features and the target time features to predict the initial predicted identity features of the target object within the target acquisition time period; reconstructing the initial predicted identity features to obtain a reconstructed signal; and extracting the predicted identity features from the reconstructed signal.
[0166] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the target heartbeat cycle based on the current electrocardiogram (ECG) signal; extracting at least one heartbeat segment from the current ECG signal according to the target heartbeat cycle; normalizing the at least one heartbeat segment to obtain a summary heartbeat signal; and extracting the target identity features from the summary heartbeat signal.
[0167] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: retraining the identity recognition model based on the current electrocardiogram signal within the current acquisition time period; and updating the target acquisition time period based on the current acquisition time period.
[0168] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.
[0169] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0170] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. An identity recognition method, characterized in that, The method includes: Acquire the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target identity features from the current ECG signal; The target time feature is determined based on the time interval between the current collection time period and the target collection time period; the target collection time period is earlier than the current collection time period. Based on the target identity features and the target time features, the predicted identity features of the target object within the target collection time period are determined; Based on the trained identity recognition model, the target identity type of the target object is determined according to the predicted identity features; the training samples of the identity recognition model include the electrocardiogram signals collected by the identity authentication object during the target collection time period.
2. The method according to claim 1, characterized in that, The step of determining the target time feature based on the time interval between the current collection time period and the target collection time period includes: Obtain the target acquisition time period corresponding to the electrocardiogram signal acquired by the target object before the current acquisition time period; Determine the time interval between the current data collection period and the target data collection period; The time interval is subjected to feature encoding processing to obtain the target time feature.
3. The method according to claim 2, characterized in that, The target time feature includes at least one time feature element with an encoded dimension; correspondingly, the feature encoding process performed on the time interval to obtain the target time feature includes: For each encoding dimension, an objective function is determined based on the parity of the encoding dimension, with the encoding dimension and the time interval as independent variables and the time feature element as the dependent variable. Based on the objective function, the temporal feature elements under the encoding dimension are determined according to the encoding dimension and the time interval.
4. The method according to any one of claims 1-3, characterized in that, The step of determining the predicted identity features of the target object within the target data collection time period based on the target identity features and the target time features includes: The target identity features and the target time features are fused together to predict the initial predicted identity features of the target object within the target collection time period; The initial predicted identity features are reconstructed to obtain a reconstructed signal; The predicted identity features are extracted from the reconstructed signal.
5. The method according to any one of claims 1-3, characterized in that, Extracting target identity features from the current electrocardiogram signal includes: Based on the current ECG signal, determine the target heart rate cycle; According to the target heart rate cycle, extract at least one heart rate segment from the current electrocardiogram signal; The at least one heartbeat segment is normalized to obtain a summary heartbeat signal; The target identity features are extracted from the summary heartbeat signal.
6. The method according to any one of claims 1-3, characterized in that, After determining the target identity type of the target object, the method further includes: The identity recognition model is retrained based on the current electrocardiogram signal within the current collection time period; The target collection time period is updated based on the current collection time period.
7. An identity recognition device, characterized in that, The device includes: The extraction module is used to acquire the current electrocardiogram (ECG) signal of the target object within the current acquisition time period, and extract the target identity features from the current ECG signal; The first determining module is used to determine a target time feature based on the time interval between the current collection time period and the target collection time period; the target collection time period is earlier than the current collection time period. The second determining module is used to determine the predicted identity features of the target object within the target collection time period based on the target identity features and the target time features; The third determination module is used to determine the target identity type of the target object based on the trained identity recognition model and the predicted identity features; the training samples of the identity recognition model include electrocardiogram signals collected from the identity authentication object during the target collection time period.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.