Artificial intelligence-based data evaluation method and device, computer device and medium

Abstract

This application belongs to the field of artificial intelligence technology and relates to a data evaluation method, apparatus, computer equipment, and storage medium based on artificial intelligence. The method includes: processing acquired specified physiological signals to obtain target signals; extracting features from the target signals based on feature types to obtain target features; performing prediction processing on the target features based on an anomaly prediction model to obtain an anomaly prediction result; if the anomaly prediction result meets the alarm triggering conditions, generating corresponding event data based on the target signal and target features; acquiring long-term health data, user basic information, medical history data, and environmental data corresponding to the user; evaluating and processing the event data, long-term health data, user basic information, medical history data, and environmental data based on a risk assessment model, and sending the obtained risk assessment result to the user. This application can be applied to health assessment scenarios in the medical technology field and can effectively improve the accuracy of risk assessment result generation.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology and can be applied to fields such as healthcare, particularly to data evaluation methods, devices, computer equipment, and storage media based on artificial intelligence. Background Technology

[0002] In the current field of cardiac health monitoring systems, fixed rules are commonly used to evaluate relevant data and generate risk assessment results. However, this approach has revealed numerous problems in terms of accuracy and latency, resulting in unsatisfactory accuracy in risk assessment. Specifically, fixed-rule assessment models are often based on a single preset standard or simple conditional judgment, lacking in-depth consideration of the complexity and dynamism of cardiac health data. This makes it difficult to accurately capture subtle changes in cardiac status and individual differences, leading to discrepancies between assessment results and actual conditions, resulting in low accuracy.

[0003] For example, in postoperative monitoring of cardiovascular diseases, traditional cardiac health monitoring systems typically rely on fixed heart rate and blood pressure thresholds to assess a patient's cardiac risk. If a patient's cardiac function is in a phase of gradual recovery but not yet fully stable, their heart rate, blood pressure, and other indicators may fluctuate within a certain range. Traditional systems, following fixed rules, trigger risk warnings whenever indicators exceed predetermined thresholds, potentially leading to frequent false alarms, causing unnecessary psychological stress for patients and increasing the workload of medical staff. Conversely, traditional systems may fail to detect potential cardiac abnormalities that have not yet reached fixed thresholds but are already showing signs of deterioration, thus delaying treatment. This inaccuracy in assessment methods seriously affects the postoperative recovery outcomes and quality of care for patients with heart disease.

[0004] Therefore, there is an urgent need to provide a new type of cardiac health monitoring system to improve the accuracy of risk assessment, thereby enhancing the accuracy and reliability of cardiac health monitoring. Summary of the Invention

[0005] The purpose of this application is to propose a data evaluation method, device, computer equipment, and storage medium based on artificial intelligence, so as to solve the technical problem that the data evaluation methods of existing heart health monitoring systems have low accuracy.

[0006] Firstly, an artificial intelligence-based data evaluation method is provided, including: The user's physiological signals are collected based on a preset multimodal sensor, and a specified physiological signal corresponding to a preset data type is obtained from the physiological signals. The specified physiological signal is processed based on a preset preprocessing strategy to obtain the corresponding target signal; Based on a preset feature type, feature extraction is performed on the target signal to obtain the corresponding target features; The target features are predicted based on a preset anomaly prediction model to obtain the corresponding anomaly prediction results. If the anomaly prediction result meets the preset alarm triggering conditions, then corresponding event data is generated based on the target signal and the target features; Acquire long-term health data, basic user information, medical history data, and environmental data corresponding to the user; The event data, long-term health data, user basic information, medical history data, and environmental data are evaluated and processed based on a preset risk assessment model to obtain the corresponding risk assessment results. The risk assessment results are sent to the user.

[0007] Secondly, an artificial intelligence-based data evaluation device is provided, comprising: The processing module is used to collect the user's physiological signals based on a preset multimodal sensor, and to obtain a specified physiological signal corresponding to a preset data type from the physiological signals; The preprocessing module is used to process the specified physiological signal based on a preset preprocessing strategy to obtain the corresponding target signal; The extraction module is used to extract features from the target signal based on a preset feature type to obtain the corresponding target features; The prediction module is used to perform prediction processing on the target features based on a preset anomaly prediction model to obtain the corresponding anomaly prediction results. The first generation module is used to generate corresponding event data based on the target signal and the target features if the anomaly prediction result meets the preset alarm triggering conditions. The acquisition module is used to acquire long-term health data, basic user information, medical history data, and environmental data corresponding to the user. The assessment module is used to assess and process the event data, the long-term health data, the user's basic information, the medical history data, and the environmental data based on a preset risk assessment model, and obtain the corresponding risk assessment results. The first sending module is used to send the risk assessment results to the user.

[0008] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described artificial intelligence-based data evaluation method.

[0009] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned artificial intelligence-based data evaluation method.

[0010] In the aforementioned scheme implemented by the data evaluation method, device, computer equipment, and storage medium based on artificial intelligence, the user's physiological signals are first collected based on a preset multimodal sensor, and a specified physiological signal corresponding to a preset data type is obtained from the physiological signals; then, the specified physiological signal is processed based on a preset preprocessing strategy to obtain a corresponding target signal; and features are extracted from the target signal based on a preset feature type to obtain corresponding target features; then, the target features are predicted based on a preset anomaly prediction model to obtain a corresponding anomaly prediction result; if the anomaly prediction result meets a preset alarm triggering condition, corresponding event data is generated based on the target signal and the target features; subsequently, long-term health data, user basic information, medical history data, and environmental data corresponding to the user are acquired; further, the event data, long-term health data, user basic information, medical history data, and environmental data are evaluated based on a preset risk assessment model to obtain a corresponding risk assessment result; finally, the risk assessment result is sent to the user. Based on the above automated processing flow, this application collects users' physiological signals using multimodal sensors, preprocesses specified physiological signals corresponding to preset data types obtained from the physiological signals to obtain corresponding target signals, then extracts features from the target signals based on feature types to obtain target features, and then uses an anomaly prediction model to predict the target features to obtain anomaly prediction results. When the anomaly prediction results are detected to meet preset alarm triggering conditions, event data is generated based on the target signals and target features. Subsequently, a risk assessment model is used to evaluate the event data, acquired long-term health data, user basic information, medical history data, and environmental data, and the generated risk assessment results are sent to the user. Thus, unlike existing methods that use fixed rules to evaluate relevant data and generate risk assessment results, this application provides a novel assessment method based on anomaly prediction and risk assessment models, which can automatically and accurately complete the data evaluation processing corresponding to the user's physiological signals, effectively improving the accuracy of the generated risk assessment results. Attached Figure Description

[0011] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is an exemplary system architecture diagram to which this application can be applied; Figure 2 This is a flowchart of an embodiment of the AI-based data evaluation method according to this application; Figure 3 yes Figure 2 A schematic diagram of a specific implementation method for step S207; Figure 4 yes Figure 3 A schematic diagram of a specific implementation of step S2072; Figure 5 This is a schematic diagram of a structure of an embodiment of the artificial intelligence-based data evaluation device according to this application; Figure 6 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0013] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0014] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0015] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0016] like Figure 1As shown, system architecture 100 may include terminal device 101, network 102, and server 103. Terminal device 101 may be a laptop 1011, tablet 1012, or mobile phone 1013. Network 102 is used as a medium to provide a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0017] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0018] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptops 1011, tablets 1012, or mobile phones 1013, terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer, and a desktop computer, etc.

[0019] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0020] It should be noted that the AI-based data evaluation method provided in this application is generally executed by a server / terminal device, and correspondingly, the AI-based data evaluation device is generally located in the server / terminal device.

[0021] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0022] Continue to refer to Figure 2This document illustrates a flowchart of an embodiment of the AI-based data evaluation method according to this application. The order of steps in the flowchart can be changed, and some steps can be omitted, depending on different needs. The AI-based data evaluation method provided in this application can be applied to any scenario requiring data evaluation, and therefore can be applied to products in these scenarios, such as data evaluation products in the medical technology field. The AI-based data evaluation method includes the following steps: Step S201: Collect the user's physiological signals based on a preset multimodal sensor, and obtain a specified physiological signal corresponding to a preset data type from the physiological signals.

[0023] In this embodiment, the data evaluation method based on artificial intelligence runs on an electronic device (e.g., Figure 1 The server / terminal device shown can acquire the user's physiological signals via wired or wireless connection. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra-wideband) connections, and other currently known or future-developed wireless connection methods. The implementing entity of this application is specifically a data evaluation system, which can be simply referred to as the system.

[0024] The above-mentioned process for acquiring user physiological signals based on preset multimodal sensors includes: ECG signal acquisition: using a three-lead dry electrode system (RA, LA, LL analog leads), a closed loop is formed through the wristband and finger contact points. The sampling rate is fixed at 250Hz to capture the details of the QRS complex. The user can initiate a 30-second short recording by actively triggering a button to verify abnormal events (such as palpitations).

[0025] PPG signal acquisition: Dual-wavelength LEDs (530nm green and 660nm red) are used as the light source, and reflected light signals are received via dual photodiodes. The sampling frequency is dynamically adjusted according to the user's status: 100Hz at rest and 200Hz during movement, to balance accuracy and power consumption. The adaptive gain control (AGC) circuit dynamically adjusts the LED emission intensity by monitoring skin reflectivity in real time, ensuring that the signal amplitude remains stable within the optimal range.

[0026] Acceleration signal acquisition: A triaxial MEMS sensor (range ±8g) samples at 50Hz and identifies user posture (sitting, walking, running, sleeping) and motion artifact intensity by calculating the acceleration modulus (ax2+ay2+az2).

[0027] Auxiliary signal acquisition: A temperature sensor (accuracy ±0.1℃) is used to correct PPG baseline drift (temperature changes cause photodiode response shift); a GSR sensor monitors sympathetic nerve activity through changes in skin conductance; an optional SpO2 module uses dual-wavelength PPG signals to assess blood oxygen saturation.

[0028] All sensors are synchronized via a hardware interrupt mechanism, and sampling points are marked with a unified timestamp (accuracy ±1ms). For example, the detection of the R peak in PPG and ECG needs to be combined with acceleration data at the same moment to determine the motion state, avoiding motion artifact interference.

[0029] In addition, the above data type can specifically be an ECG signal type, and the corresponding specified physiological signal can specifically be an ECG signal.

[0030] The multimodal sensing terminal acquires physiological signals (PPG, ECG) and environmental signals (acceleration, temperature, etc.) through multiple sensors in a collaborative manner, and optimizes signal quality using dynamic sampling rate and adaptive gain control. A time synchronization mechanism ensures that multimodal data is strictly aligned on the time axis, providing reliable input for subsequent edge processing.

[0031] Step S202: The specified physiological signal is processed based on a preset preprocessing strategy to obtain the corresponding target signal.

[0032] In this embodiment, the preprocessing of the specified physiological signal (ECG signal) includes: 1) 50 / 60Hz notch filtering. A double-T notch filter is designed to eliminate power frequency interference. Power frequency interference refers to electromagnetic interference generated by the power system at 50Hz (in China) or 60Hz (in the United States). This interference is superimposed on the ECG signal, affecting the signal quality. The double-T notch filter is a filter with a specific frequency response. It can form a deep notch at the center frequency (50Hz or 60Hz), significantly attenuating the interference signal at that frequency while having less impact on signals at other frequencies, thereby effectively removing power frequency interference and preserving the effective physiological information in the ECG signal.

[0033] 2) High-pass filtering. A 0.5Hz high-pass filter is used to remove baseline drift. Baseline drift in ECG signals mainly manifests as a slow change in the overall vertical direction of the signal. This change may be caused by low-frequency fluctuations due to physiological activities such as breathing. A 0.5Hz high-pass filter allows frequency components above 0.5Hz to pass through while blocking components below 0.5Hz, thereby removing baseline drift and stabilizing the ECG signal baseline at a level, facilitating subsequent analysis and processing.

[0034] 3) R-peak localization and RR interval generation. First, QRS complexes are initially located using threshold detection. A suitable threshold is set; when the amplitude of the ECG signal exceeds this threshold, a QRS complex is preliminarily considered to have been detected. However, due to potential noise or other interference in the signal, the initial localization result may contain false positives. Therefore, further confirmation using morphological verification is necessary. Morphological verification mainly examines the width, slope, and other characteristics of the QRS complex. For example, a normal QRS complex has a certain width range and slope variation pattern. If the detected complex does not conform to these characteristics, it is considered a false positive and is discarded. After accurately locating the QRS complex, the time interval between adjacent QRS complexes, i.e., the RR interval, is calculated. To ensure the accuracy of the RR interval data, outliers also need to be removed. An outlier removal method based on the interquartile range (IQR) is used. First, the quartiles Q1 and Q3 of the RR interval are calculated, and then the interquartile range IQR = Q3 - Q1 is calculated. If a certain RR interval exceeds the range of Q1 - 1.5IQR or Q3 + 1.5IQR, then the RR interval is considered an outlier and is removed to ensure the reliability of the RR interval data used in subsequent analyses.

[0035] Step S203: Extract features from the target signal based on the preset feature type to obtain the corresponding target features.

[0036] In this embodiment, the above-mentioned feature types may include time-domain feature types, frequency-domain feature types, nonlinear feature types, and morphological feature types.

[0037] The specific implementation process of extracting features from the target signal based on the preset feature type to obtain the corresponding target features will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0038] Step S204: Based on the preset anomaly prediction model, the target features are predicted to obtain the corresponding anomaly prediction results.

[0039] In this embodiment, the specific implementation process of predicting the target features based on the preset anomaly prediction model to obtain the corresponding anomaly prediction result will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0040] Step S205: If the anomaly prediction result meets the preset alarm triggering conditions, then corresponding event data is generated based on the target signal and the target features.

[0041] In this embodiment, if an abnormal prediction result is detected from the output of the abnormal prediction model, such as a certain type of abnormal probability being greater than a preset abnormal threshold (e.g., 85%), it is considered that a relatively serious cardiac abnormal event may have occurred. At this time, the corresponding event package, i.e. the event data, will be automatically generated based on the target signal and target features.

[0042] The specific implementation process of generating corresponding event data based on the target signal and the target features will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0043] In addition, the system also has a motion state correction function. It can combine acceleration data to determine the user's motion state. The acceleration modulus reflects the intensity of human motion in three-dimensional space. When the acceleration modulus is greater than 0.5g, the user is considered to be in a state of vigorous exercise. During vigorous exercise, the body's physiological state will change significantly, such as increased heart rate and rapid breathing. These changes may interfere with the acquisition and analysis of cardiac signals, leading to model misjudgment. Therefore, when vigorous exercise is detected, the initial alarm is corrected. If it is a delayed alarm, that is, the alarm is not triggered temporarily, and the judgment is re-evaluated after the user's motion state stabilizes; if it is downgraded to an "observation event", for example, an event that was originally judged as atrial fibrillation with a probability of 90% is downgraded to an observation event, reducing the probability of atrial fibrillation to 70%, in order to reduce the false alarm rate and avoid unnecessary trouble for the user caused by false alarms due to vigorous exercise.

[0044] Step S206: Obtain long-term health data, basic user information, medical history data, and environmental data corresponding to the user.

[0045] In this embodiment, the aforementioned long-term health data includes the user's heart rate variability (HRV) trend, which reflects the regulatory function of the cardiac autonomic nervous system; sleep quality score, which reflects the impact of the user's sleep status on cardiac health; and resting heart rate trend, which reflects the heart's function at rest. The aforementioned basic user information can be filled in during user registration or obtained synchronously through health records. The aforementioned medical history data is synchronously obtained from the Hospital Information System (HIS) or Electronic Health Record (EHR) with user authorization. The aforementioned environmental data can be obtained in real time from third-party environmental data platforms (such as meteorological bureaus and air quality monitoring stations) via API.

[0046] Since the multi-source data comes from different collection devices and time points, it is mapped to a unified timeline using timestamps to enable comprehensive analysis. For example, ECG events can be correlated with sleep stages on the same day to analyze whether abnormal cardiac events occurred during specific sleep stages, revealing the potential relationship between sleep and heart health. This data alignment process allows for a more comprehensive and accurate understanding of a user's health status at different times and in different states, providing a more reliable data foundation for subsequent risk assessments.

[0047] Step S207: Based on a preset risk assessment model, the event data, the long-term health data, the user's basic information, the medical history data, and the environmental data are evaluated and processed to obtain the corresponding risk assessment results.

[0048] In this embodiment, the specific implementation process of evaluating and processing the event data, long-term health data, user basic information, medical history data, and environmental data based on the preset risk assessment model to obtain the corresponding risk assessment results will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0049] Step S208: Send the risk assessment results to the user.

[0050] In this embodiment, the generated risk assessment results can be sent to the user via methods such as APP push notifications, SMS notifications, and telephone notifications.

[0051] This application first collects the user's physiological signals based on a preset multimodal sensor, and then extracts a specified physiological signal corresponding to a preset data type from the physiological signals. Next, it processes the specified physiological signal based on a preset preprocessing strategy to obtain a corresponding target signal. Then, it extracts features from the target signal based on a preset feature type to obtain corresponding target features. Afterward, it performs prediction processing on the target features based on a preset anomaly prediction model to obtain a corresponding anomaly prediction result. If the anomaly prediction result meets preset alarm triggering conditions, it generates corresponding event data based on the target signal and the target features. Subsequently, it acquires long-term health data, user basic information, medical history data, and environmental data corresponding to the user. Further, it evaluates the event data, long-term health data, user basic information, medical history data, and environmental data based on a preset risk assessment model to obtain a corresponding risk assessment result. Finally, it sends the risk assessment result to the user. Based on the above automated processing flow, this application collects users' physiological signals using multimodal sensors, preprocesses specified physiological signals corresponding to preset data types obtained from the physiological signals to obtain corresponding target signals, then extracts features from the target signals based on feature types to obtain target features, and then uses an anomaly prediction model to predict the target features to obtain anomaly prediction results. When the anomaly prediction results are detected to meet preset alarm triggering conditions, event data is generated based on the target signals and target features. Subsequently, a risk assessment model is used to evaluate the event data, acquired long-term health data, user basic information, medical history data, and environmental data, and the generated risk assessment results are sent to the user. Thus, unlike existing methods that use fixed rules to evaluate relevant data and generate risk assessment results, this application provides a novel assessment method based on anomaly prediction and risk assessment models, which can automatically and accurately complete the data evaluation processing corresponding to the user's physiological signals, effectively improving the accuracy of the generated risk assessment results.

[0052] In some alternative implementations, step S203 includes the following steps: The target signal is subjected to time-domain feature extraction to obtain the corresponding time-domain features.

[0053] In this embodiment, the process of extracting the time-domain features includes: calculating the mean RR interval (μ_RR), which reflects the average cycle of the heartbeat and is an important indicator for measuring heart rhythm; calculating the standard deviation (SDNN), which reflects the overall variability of the RR interval and can reflect the long-term fluctuations of the heartbeat; calculating the root mean square (RMSSD) of the difference between adjacent RR intervals, which mainly reflects the short-term variability of the heartbeat and is closely related to the autonomic nervous system regulation function of the heart; and calculating the proportion of adjacent RR interval differences > 50ms (pNN50), which is also an indicator reflecting the short-term variability of the heart. The larger the value, the more obvious the short-term changes in the heartbeat.

[0054] Frequency domain features are extracted from the target signal to obtain the corresponding frequency domain features.

[0055] In this embodiment, the frequency domain feature extraction process includes: calculating the power spectral density using the Welch method. The Welch method is a commonly used power spectrum estimation method. It divides the signal into multiple overlapping segments, performs a Fourier transform on each segment, calculates the power spectrum of each segment, and finally averages the power spectra of all segments to obtain the power spectral density of the entire signal. Based on the power spectral density, low-frequency (LF, 0.04 - 0.15 Hz) and high-frequency (HF, 0.15 - 0.4 Hz) power are extracted. LF power mainly reflects the activity of the sympathetic nervous system, while HF power mainly reflects the activity of the vagus nerve. Calculating the LF / HF ratio can reflect the balance of the autonomic nervous system, and changes in this ratio can provide important information about the human physiological state and stress response.

[0056] Nonlinear features are extracted from the target signal to obtain the corresponding nonlinear features.

[0057] In this embodiment, the implementation process of the above-mentioned nonlinear feature extraction includes: drawing a Poincaré diagram, which is a graph plotted on a plane using adjacent RR intervals as coordinate points, with the scatter points distributed as (RR_i, RR_i + 1), where RR_i represents the i-th RR interval and RR_i + 1 represents the (i + 1)-th RR interval. By analyzing the distribution of the scatter points in the Poincaré diagram, SD1 and SD2 can be calculated. SD1 reflects the short-term variability of the heartbeat, which is related to the distribution of the scatter points in the direction perpendicular to the line RR_i = RR_i + 1; SD2 reflects the long-term variability of the heartbeat, which is related to the distribution of the scatter points in the direction parallel to the line RR_i = RR_i + 1. The sample entropy (SampEn) is calculated. Sample entropy is an indicator that quantifies the complexity of heart rhythm by comparing vector similarity. It reflects the regularity and complexity of the signal; a larger sample entropy value indicates a more complex heart rhythm and more active autonomic nervous system regulation.

[0058] The target signal is subjected to morphological feature extraction to obtain the corresponding morphological features.

[0059] In this embodiment, the above-mentioned morphological feature extraction process includes: extracting the QRS wave width from the ECG signal, that is, the duration from the start of the Q wave to the end of the S wave. The QRS wave width can reflect the time of ventricular depolarization, and its changes may be related to heart disease; detecting the visibility of the P wave. The P wave reflects the depolarization process of the atrium. The presence of the P wave is detected by setting an appropriate threshold. Abnormalities of the P wave may be related to atrial disease; calculating the ST segment offset trend. The ST segment is a line segment from the end of the QRS complex to the beginning of the T wave. The slope change 60ms after the J point is calculated. The ST segment offset and slope change can reflect myocardial ischemia, damage, etc., and are of great significance for the diagnosis of heart disease.

[0060] The time-domain features, frequency-domain features, nonlinear features, and morphological features are integrated to obtain the corresponding integrated features.

[0061] In this embodiment, the extracted time-domain features, frequency-domain features, nonlinear features, and morphological features are integrated and processed, and the resulting integrated features are used as the corresponding target features.

[0062] The integrated feature is used as the target feature.

[0063] This application simultaneously extracts time-domain features from the target signal to obtain corresponding time-domain features; extracts frequency-domain features from the target signal to obtain corresponding frequency-domain features; extracts nonlinear features from the target signal to obtain corresponding nonlinear features; and extracts morphological features from the target signal to obtain corresponding morphological features. Then, the time-domain features, frequency-domain features, nonlinear features, and morphological features are integrated to obtain corresponding integrated features. These integrated features are then used as the target features. Based on the above processing flow, this application extracts multiple features from the target signal, including time-domain features, frequency-domain features, nonlinear features, and morphological features. These features describe the physiological state of the heart and the autonomic nervous system's regulatory function from different perspectives, which is beneficial for providing multi-dimensional and accurate data input for subsequent model inference.

[0064] In some optional implementations of this embodiment, step S204 includes the following steps: A preset anomaly prediction model is invoked; wherein the anomaly prediction model includes a feature extraction layer, a temporal modeling layer, and a classification head.

[0065] In this embodiment, the above-mentioned anomaly prediction model can specifically adopt a lightweight model (TinyHeartNet) consisting of a feature extraction layer, a temporal modeling layer, and a classification head.

[0066] Based on the feature extraction layer, information is extracted from the target features to obtain the corresponding key feature information.

[0067] In this embodiment, the feature extraction layer employs 1D convolution and depthwise separable convolution to extract time-frequency features. 1D convolution operates along a one-dimensional direction of the signal, capturing local features. By setting different kernel sizes and numbers, features at different scales can be extracted. Depthwise separable convolution is a special type of convolution that decomposes the standard convolution operation into two steps: depthwise convolution and pointwise convolution. This significantly reduces the number of model parameters while maintaining feature extraction capabilities to some extent. During feature extraction, a Squeeze-and-Excitation (SE) module is introduced. The SE module first compresses the feature map of each channel into a scalar using global average pooling. This scalar can be considered a global representation of the channel's features. Then, channel weights are generated through fully connected layers and other operations. These weights reflect the importance of features in different channels. By multiplying the channel weights with the original feature map, QRS wave-related features can be highlighted. For example, when the kernel size is 3, it can better capture the high-frequency components in the signal. Since QRS waves are high-frequency components in ECG signals, the SE module can enhance the weights of feature channels related to QRS waves, making the model pay more attention to these important features.

[0068] The feature extraction layer (1D convolution + depthwise separable convolution + SE module) extracts time-frequency features from the original input (ECG segment or PPG + accelerometer fusion signal) and outputs a feature vector or feature map. Specifically, for ECG signals, the feature extraction layer may output a feature map of shape (T, C), where T is the time step (e.g., the length of the sequence after convolution) and C is the number of channels (e.g., the number of convolution kernels).

[0069] Based on the time-series modeling layer, the key feature information is processed using time-series modeling to obtain the corresponding output data.

[0070] In this embodiment, the aforementioned temporal modeling layer employs a bidirectional LSTM (Long Short-Term Memory) network to capture long-term dependencies in the RR interval sequence. The RR interval sequence reflects the rhythmic changes of the heartbeat; some heart diseases, such as atrial fibrillation, can cause irregularities in the RR interval. The bidirectional LSTM consists of a forward LSTM and a backward LSTM. The forward LSTM processes data starting from the beginning of the sequence, capturing the influence of earlier information on the current moment; the backward LSTM processes data starting from the end of the sequence, capturing the influence of later information on the current moment. By fusing the forward and backward information, the bidirectional LSTM can more comprehensively understand the contextual information of the RR interval sequence, thereby better capturing long-term dependencies, such as accurately identifying irregular patterns in the RR intervals of patients with atrial fibrillation.

[0071] The temporal modeling layer (bidirectional LSTM) takes as input the output of the feature extraction layer. Its function is to perform temporal modeling on these features, capturing long-term dependencies in the sequence (such as the variation of RR intervals). The bidirectional LSTM considers both forward and backward information of the sequence, outputting a hidden state sequence or the hidden state of the last time step (depending on the implementation), typically with a shape of (T, H) or (H,), where H is the dimension of the LSTM's hidden layer. Furthermore, the output of the temporal modeling layer (such as the hidden state of the last time step) contains temporal information of the sequence but does not directly correspond to a specific anomaly category.

[0072] The output data is classified based on the classification header to obtain the corresponding classification result.

[0073] In this embodiment, the classification head uses the Softmax function as the output layer. The Softmax function can convert the model's output into the probability distribution of each abnormal category (i.e., the abnormal classification result). For example, for abnormal categories such as atrial fibrillation and premature ventricular contractions, the Softmax function will calculate the probability value corresponding to each category based on the input feature vector, and the sum of these probability values ​​is 1.

[0074] The role of the classification head (Softmax function) is to map the output of the time series modeling layer to a class probability distribution: if the output of the time series modeling layer is a vector of (H,), the classification head will map it to the dimension of (n_classes,) through a fully connected layer (optional) (n_classes is the number of abnormal classes, such as atrial fibrillation, premature ventricular contractions, etc.), and then convert it into probabilities through the Softmax function.

[0075] The classification result is used as the anomaly prediction result.

[0076] This application utilizes a pre-defined anomaly prediction model, which includes a feature extraction layer, a temporal modeling layer, and a classification head. The feature extraction layer extracts information from the target features to obtain corresponding key feature information. The temporal modeling layer then performs temporal modeling processing on the key feature information to obtain corresponding output data. Subsequently, the classification head classifies the output data to obtain a classification result. Finally, the classification result is used as the anomaly prediction result. Based on this processing flow, the anomaly prediction model used in this application employs a lightweight design, utilizing 1D convolution, depthwise separable convolution, and SE modules, which reduces the number of model parameters. This allows the model to perform local real-time inference on edge devices, enabling rapid and accurate analysis and judgment of input cardiac signals, thus ensuring the accuracy of the generated anomaly prediction results.

[0077] In some alternative implementations, step S205 includes the following steps: Obtain the timestamp information corresponding to the anomaly prediction result.

[0078] In this embodiment, the timestamp information mentioned above is the specific time when the event occurred, such as the generation time of the above-mentioned anomaly prediction result, which facilitates subsequent time series analysis and tracing of the event.

[0079] Perform a hash operation on the target signal to obtain the corresponding hash data.

[0080] In this embodiment, a unique identifier, i.e., a hash value, or hash data, can be obtained by performing hash calculations on the target signal using a common hash algorithm (such as SHA-256, SHA-512, etc.). The hash value is unique and irreversible, and can be used for data integrity verification to ensure that the signal fragment has not been tampered with during transmission and processing.

[0081] The target signal, the target feature, the timestamp information, and the hash data are integrated to obtain the corresponding integrated data.

[0082] In this embodiment, the target signal, target features, and generated timestamp information and hash data can be integrated and processed, and the resulting integrated data (integration package) can be used as the corresponding event data.

[0083] The integrated data is used as the event data.

[0084] In this embodiment, the generated event data packets can be encrypted using an encryption algorithm (such as AES-256) before being uploaded to the cloud. AES-256 is an advanced encryption standard with high security and encryption strength, effectively protecting the data in the event packets from theft and tampering. The original event data is stored locally for a period of 30 days. This is done to ensure data traceability while preventing user privacy leaks. Because the original event data contains detailed cardiac physiological information about the user, uploading it arbitrarily to the cloud could pose privacy risks. Local storage, on the other hand, can better protect user privacy while meeting data retrieval needs within a certain timeframe.

[0085] This application obtains the timestamp information corresponding to the anomaly prediction result; then performs a hash operation on the target signal to obtain corresponding hash data; subsequently, it integrates the target signal, target features, timestamp information, and hash data to obtain integrated data; and finally, it uses the integrated data as the event data. Based on the above processing flow, this application obtains the timestamp information corresponding to the anomaly prediction result, performs a hash operation on the target signal to obtain hash data, and then integrates the target signal, target features, and the generated timestamp information and hash data, using the resulting integrated data as the corresponding event data. This enables rapid generation of event data, improving the efficiency of event data generation for further analysis and processing.

[0086] Please refer to Figure 3 In some alternative implementations, step S207 includes the following steps: Step S2071: Extract features from the multi-source data to obtain corresponding feature data; wherein, the multi-source data includes the event data, the long-term health data, the user's basic information, the medical history data, and the environmental data.

[0087] In this embodiment, real-time features can be obtained by extracting features from the aforementioned event data. These real-time features reflect the user's current cardiovascular status. The real-time features include: RR interval variability (SDNN), which originates from the ECG signal. R-peaks or pulse peaks can be extracted using a peak detection algorithm, and the time difference between adjacent peaks can be calculated to obtain the RR interval sequence. The standard deviation (SDNN) of the RR interval sequence is then calculated to reflect the overall level of heart rate variability (HRV); ST trend (slope change), which originates from the ST segment of the ECG signal (the interval from the end of the QRS wave to the beginning of the T wave). Sliding window analysis of the ST segment slope change can detect myocardial ischemia or damage (such as ST segment elevation / depression); and heart rate mutation rate, which originates from the RR interval sequence. The proportion of adjacent RR interval differences >20% reflects abnormal heart rate fluctuations (such as premature beats or arrhythmias caused by atrial fibrillation).

[0088] Historical features can be obtained by extracting features from the aforementioned long-term health data. These historical features are generated through aggregated analysis of the user's long-term health data, reflecting the user's cardiovascular health trends or chronic risks. Historical features include: 7-day average HRV, sourced from daily HRV values ​​(such as SDNN, RMSSD, etc.), which can be obtained by taking the average HRV over the most recent 7 days to reflect the long-term state of autonomic nervous system function; minimum nocturnal heart rate, sourced from resting heart rate data during sleep (usually collected by smart bracelets / watches), which can be obtained by extracting the minimum heart rate value during each night's sleep and taking the average or minimum value over the most recent 7 days; and atrial fibrillation event frequency, sourced from atrial fibrillation event records detected by edge ECG monitoring devices, which can be obtained by statistically analyzing the number of atrial fibrillation attacks in the most recent 30 days or the hourly atrial fibrillation load (percentage).

[0089] By extracting features from the aforementioned basic user information, corresponding demographic characteristics can be obtained. These demographic characteristics are either filled in by the user during registration or obtained synchronously from their health record, reflecting an individual's underlying risk. Demographic characteristics include: age, sourced from user registration or synchronously from an electronic health record (EHR); gender, sourced from user registration or synchronously from an EHR; and BMI, calculated from the user's height and weight (BMI = weight (kg) / height² (m²)) or synchronously from a smart body fat scale.

[0090] By extracting features from the aforementioned medical history data, corresponding medical history characteristics can be obtained. These characteristics, after user authorization, are synchronized from the Hospital Information System (HIS) or Electronic Health Record (EHR), reflecting the impact of chronic diseases or treatments on the cardiovascular system. Medical history characteristics include: whether beta-blockers are taken (e.g., metoprolol, bisoprolol, etc.), which reduce HRV; pacemaker status (e.g., whether a pacemaker is implanted and its operating mode); and history of hypertension / diabetes (e.g., extracted from diagnostic records via ICD coding or keyword matching).

[0091] By extracting features from the aforementioned environmental data, corresponding environmental characteristics can be obtained. These characteristics are acquired in real-time from third-party environmental data platforms (such as meteorological bureaus and air quality monitoring stations) via API, reflecting the short-term impact of the environment on the cardiovascular system. Environmental characteristics include: the maximum daily temperature difference, obtained from the difference between the highest and lowest temperatures of the day via meteorological API (a temperature difference >10℃ may increase cardiovascular burden); PM2.5 index, obtained from the current PM2.5 concentration via air quality API (>75μg / m³ may trigger inflammatory responses and affect vascular endothelial function); and air pressure, obtained from the current air pressure value via meteorological API (low air pressure may be associated with headaches or blood pressure fluctuations).

[0092] Among these features, the corresponding feature data can be obtained by integrating the aforementioned real-time features, historical features, demographic features, medical history features, and environmental features.

[0093] Step S2072: Obtain the feature weights corresponding to the feature data.

[0094] In this embodiment, the specific implementation process of obtaining the feature weights corresponding to the feature data will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0095] Step S2073: Call the preset risk score calculation formula.

[0096] In this embodiment, the formula for calculating the risk score is as follows: Where e is the natural constant. It is the intercept term. The weights for each feature, These are the values ​​of each feature data.

[0097] Step S2074: Based on the risk assessment model, the feature data and feature weights are calculated using the risk score calculation formula to obtain the corresponding risk score.

[0098] In this embodiment, the risk assessment model uses a logistic regression model to calculate the risk level (0-100 points). Specifically, the feature data and corresponding feature weights are substituted into the risk calculation formula for calculation, so as to integrate the information of multiple features and calculate a risk score between 0 and 100.

[0099] Step S2075: Perform level mapping processing on the risk score to obtain the corresponding target risk level.

[0100] In this embodiment, different risk levels can be divided according to the risk score: 0-30 points are low risk, which means that the user's current cardiovascular health is relatively good; 31-60 points are medium risk, which indicates that the user has certain cardiovascular risks and needs attention; 61-85 points are high risk, which indicates that the user's cardiovascular risk is high and further measures need to be taken; 86-100 points are emergency risk, which means that the user may be in a serious cardiovascular crisis and needs immediate intervention.

[0101] Step S2076: The target risk level is used as the risk assessment result.

[0102] In this embodiment, to help doctors understand the model's decision-making basis, SHAP scores are used to sort and display the main risk factors. SHAP scores are a method for interpreting the prediction results of machine learning models; they quantify the contribution of each feature to the model's predictions. For example, if the report shows "HRV decrease contribution +28 points," it means that the feature of HRV decrease increased the user's final risk score by 28 points, making it one of the important factors leading to increased risk. Through this interpretable report, doctors can more clearly understand how the model arrives at risk assessment results based on input data, thereby better developing treatment plans for patients.

[0103] This application extracts features from multi-source data to obtain corresponding feature data; then, it obtains the feature weights corresponding to the feature data; and calls a preset risk score calculation formula; subsequently, based on the risk assessment model, it uses the risk score calculation formula to calculate and process the feature data and the feature weights to obtain the corresponding risk score; next, it performs level mapping processing on the risk score to obtain the corresponding target risk level; finally, it uses the target risk level as the risk assessment result. Based on the above processing flow, this application, by using a risk assessment model, can comprehensively consider the impact of various factors on cardiovascular health, automatically and accurately use the risk score calculation formula to calculate and process the feature data and corresponding feature weights to generate corresponding risk assessment results, realizing an upgrade from single-event detection to long-term risk assessment, and improving the accuracy and intelligence of the generated risk assessment results.

[0104] Please refer to Figure 4 In some optional implementations of this embodiment, step S2072 includes the following steps: Step S20721: Invoke the preset weight generation algorithm.

[0105] In this embodiment, the weight generation algorithm described above can specifically be the XGBoost algorithm.

[0106] Step S20722: Perform weight generation processing on the specified feature data based on the weight generation algorithm to obtain the corresponding initial weight; wherein, the specified feature data is any one of the data contained in the feature data.

[0107] In this embodiment, the XGBoost algorithm is a powerful machine learning algorithm that can automatically learn the degree of influence of each feature on the risk score, thereby determining appropriate weights and generating the initial weights of each feature contained in the above feature data.

[0108] Step S20723: Adjust the initial weights based on the preset weight constraint strategy to obtain the corresponding specified weights.

[0109] In this embodiment, the aforementioned weight constraint strategy refers to domain-knowledge-based weight constraints. The goal is to leverage medical domain knowledge to impose logical constraints on feature weights, preventing model output from contradicting clinical common sense. The corresponding implementation methods include: Weight range limitation: Based on clinical research, the direction of some features' impact on cardiovascular risk is clear (e.g., a history of hypertension should increase risk, while regular exercise should decrease risk). The weights output by XGBoost are truncated or scaled to ensure their sign and range conform to expectations. For example, if the weight of a history of hypertension is negative (contradicting expectations), it is forcibly set to a positive minimum threshold (e.g., 0.1). Weight ratio adjustment: The weight ratio is adjusted according to the clinical importance of the feature. For example, if the weight of age is much lower than that of PM2.5 (an environmental factor), but clinically age is considered a stronger risk factor, the weight of age can be proportionally increased.

[0110] The initial weights can be adjusted based on the implementation of the weight constraint strategy, and the resulting specified weights can be used as the feature weights of the specified feature data.

[0111] Step S20724: Use the specified weight as the feature weight of the specified feature data.

[0112] This application calls a preset weight generation algorithm; then, based on the weight generation algorithm, it performs weight generation processing on specified feature data to obtain corresponding initial weights; wherein, the specified feature data is any data contained in the feature data; subsequently, it adjusts the initial weights based on a preset weight constraint strategy to obtain corresponding specified weights; and finally, it uses the specified weights as the feature weights of the specified feature data. Based on the above processing flow, this application uses a weight generation algorithm to perform weight generation processing on specified feature data to obtain initial weights, and then intelligently adjusts the initial weights based on the use of a weight constraint strategy, and uses the adjusted specified weights as the feature weights of the specified feature data. This intelligently and significantly improves the adaptability and accuracy of feature weights based on the use of the weight generation algorithm, thereby contributing to the optimization of the accuracy and reliability of the subsequently generated risk scores.

[0113] In some optional implementations of this embodiment, after step S207, the electronic device may further perform the following steps: Early warning information corresponding to the risk assessment results is generated based on a preset information generation strategy.

[0114] In this embodiment, a warning information template can be pre-constructed according to actual warning needs. Then, by filling the corresponding positions in the warning information template with the generated risk assessment results, the corresponding warning information is generated.

[0115] Determine the target notification method corresponding to the risk assessment results.

[0116] In this embodiment, based on the calculated risk level (i.e., risk assessment result), the system will adopt a tiered notification approach. For low-risk users, a push notification will be sent via the app to inform them of their current cardiovascular health status and remind them to maintain healthy lifestyle habits, such as a balanced diet and moderate exercise. For medium-risk users, a text message will suggest a follow-up examination, reminding them to go to the hospital for further examination to determine if there are any potential cardiovascular diseases. For high-risk users, family members and doctors will be notified to keep family members informed of the user's health status, while doctors can contact the user promptly to develop an appropriate treatment plan. For users at emergency risk, the platform will trigger the SOS function, automatically dialing emergency numbers and sending location information to ensure that the user receives professional medical assistance in the shortest possible time.

[0117] Based on the target notification method, the warning information is sent to the user.

[0118] In this embodiment, the generated warning information can be sent to the aforementioned user according to the selected target notification method.

[0119] This application generates early warning information corresponding to the risk assessment result based on a preset information generation strategy; then determines the target notification method corresponding to the risk assessment result; and subsequently sends the early warning information to the user based on the target notification method. Based on the above processing flow, this application automatically generates early warning information corresponding to the risk assessment result by using an information generation strategy, thus improving the efficiency of early warning information generation. Furthermore, it intelligently sends the early warning information to the user based on the target notification method corresponding to the risk assessment result, thereby improving the intelligence of early warning information delivery and enhancing the user experience.

[0120] In some alternative implementations, the user information obtained is subject to user consent and complies with relevant laws and policies.

[0121] Furthermore, any software tools or components not belonging to our company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.

[0122] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0123] It should be emphasized that, to further ensure the privacy and security of the above risk assessment results, the risk assessment results can also be stored in a blockchain node.

[0124] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0125] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0126] 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 instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0127] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by 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 accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0128] Further reference Figure 5 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of an artificial intelligence-based data evaluation device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0129] like Figure 5 As shown, the artificial intelligence-based data evaluation device 300 described in this embodiment includes: a processing module 301, a preprocessing module 302, an extraction module 303, a prediction module 304, a first generation module 305, an acquisition module 306, an evaluation module 307, and a first sending module 308. Wherein: The processing module 301 is used to collect the user's physiological signals based on a preset multimodal sensor, and to obtain a specified physiological signal corresponding to a preset data type from the physiological signals. The preprocessing module 302 is used to process the specified physiological signal based on a preset preprocessing strategy to obtain the corresponding target signal; Extraction module 303 is used to extract features from the target signal based on a preset feature type to obtain the corresponding target features; The prediction module 304 is used to perform prediction processing on the target features based on a preset anomaly prediction model to obtain the corresponding anomaly prediction result. The first generation module 305 is used to generate corresponding event data based on the target signal and the target features if the anomaly prediction result meets the preset alarm triggering conditions. The acquisition module 306 is used to acquire long-term health data, basic user information, medical history data and environmental data corresponding to the user; The assessment module 307 is used to assess the event data, the long-term health data, the user's basic information, the medical history data and the environmental data based on a preset risk assessment model, and obtain the corresponding risk assessment results. The first sending module 308 sends the risk assessment result to the user.

[0130] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data evaluation method in the aforementioned implementation method, and will not be repeated here.

[0131] In some optional implementations of this embodiment, the extraction module 303 includes: The first extraction submodule is used to extract time-domain features from the target signal to obtain the corresponding time-domain features; The second extraction submodule is used to extract frequency domain features from the target signal to obtain the corresponding frequency domain features; The third extraction submodule is used to extract nonlinear features from the target signal to obtain the corresponding nonlinear features; The fourth extraction submodule is used to extract morphological features from the target signal to obtain the corresponding morphological features; The first integration submodule is used to integrate the time-domain features, the frequency-domain features, the nonlinear features, and the morphological features to obtain the corresponding integrated features; The first determining submodule is used to use the integrated feature as the target feature.

[0132] In some optional implementations of this embodiment, the prediction module 304 includes: The first calling submodule is used to call a preset anomaly prediction model; wherein, the anomaly prediction model includes a feature extraction layer, a temporal modeling layer, and a classification head; The fifth extraction submodule is used to extract information from the target features based on the feature extraction layer to obtain the corresponding key feature information; The modeling submodule is used to perform time-series modeling processing on the key feature information based on the time-series modeling layer to obtain the corresponding output data. The classification submodule is used to classify the output data based on the classification header to obtain the corresponding classification result. The second determination submodule is used to use the classification result as the anomaly prediction result.

[0133] In some optional implementations of this embodiment, the first generation module 305 includes: The first acquisition submodule is used to acquire timestamp information corresponding to the anomaly prediction result; The computation submodule is used to perform hash operations on the target signal to obtain the corresponding hash data; The second integration submodule is used to integrate the target signal, the target feature, the timestamp information and the hash data to obtain the corresponding integrated data. The third determining submodule is used to use the integrated data as the event data.

[0134] In some optional implementations of this embodiment, the evaluation module 307 includes: The sixth extraction submodule is used to extract features from multi-source data to obtain corresponding feature data; wherein, the multi-source data includes the event data, the long-term health data, the user's basic information, the medical history data, and the environmental data; The second acquisition submodule is used to acquire the feature weights corresponding to the feature data; The second submodule is used to invoke the preset risk score calculation formula; The calculation submodule is used to calculate and process the feature data and feature weights based on the risk assessment model and the risk score calculation formula to obtain the corresponding risk score. The mapping submodule is used to perform level mapping processing on the risk score to obtain the corresponding target risk level; The third determining submodule is used to take the target risk level as the risk assessment result.

[0135] In some optional implementations of this embodiment, the second acquisition submodule includes: The calling unit is used to invoke the preset weight generation algorithm; The processing unit is configured to perform weight generation processing on the specified feature data based on the weight generation algorithm to obtain the corresponding initial weights; wherein, the specified feature data is any one of the data contained in the feature data; An adjustment unit is used to adjust the initial weights based on a preset weight constraint strategy to obtain the corresponding specified weights. A determining unit is used to use the specified weight as the feature weight of the specified feature data.

[0136] In some optional implementations of this embodiment, the artificial intelligence-based data evaluation device further includes: The second generation module is used to generate early warning information corresponding to the risk assessment results based on a preset information generation strategy. The determination module is used to determine the target notification method corresponding to the risk assessment result; The second sending module is used to send the warning information to the user based on the target notification method.

[0137] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 6 , Figure 6 This is a basic structural block diagram of the computer device in this embodiment.

[0138] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0139] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0140] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 4. Of course, the memory 41 may also include both the internal storage unit and its external storage device of the computer device 4. In this embodiment, the memory 41 is typically used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for data evaluation methods based on artificial intelligence. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.

[0141] In some embodiments, the processor 42 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 42 is typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or to process data, for example, to execute computer-readable instructions for the artificial intelligence-based data evaluation method.

[0142] The network interface 43 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 4 and other electronic devices.

[0143] Compared with the prior art, the embodiments of this application have the following beneficial effects: In this embodiment, the application collects the user's physiological signals using a multimodal sensor, preprocesses the selected physiological signals corresponding to preset data types to obtain target signals, then extracts features from the target signals based on feature types to obtain target features, and then uses an anomaly prediction model to predict the target features to obtain anomaly prediction results. When the anomaly prediction results meet preset alarm triggering conditions, event data is generated based on the target signals and target features. Subsequently, a risk assessment model is used to evaluate the event data, acquired long-term health data, user basic information, medical history data, and environmental data, and the generated risk assessment results are sent to the user. Thus, unlike existing methods that use fixed rules to evaluate relevant data and generate risk assessment results, this application provides a novel evaluation method based on anomaly prediction and risk assessment models, which can automatically and accurately complete the data evaluation processing corresponding to the user's physiological signals, effectively improving the accuracy of the generated risk assessment results.

[0144] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the artificial intelligence-based data evaluation method described above.

[0145] Compared with the prior art, the embodiments of this application have the following main advantages: In this embodiment, the application collects the user's physiological signals using a multimodal sensor, preprocesses the selected physiological signals corresponding to preset data types to obtain target signals, then extracts features from the target signals based on feature types to obtain target features, and then uses an anomaly prediction model to predict the target features to obtain anomaly prediction results. When the anomaly prediction results meet preset alarm triggering conditions, event data is generated based on the target signals and target features. Subsequently, a risk assessment model is used to evaluate the event data, acquired long-term health data, user basic information, medical history data, and environmental data, and the generated risk assessment results are sent to the user. Thus, unlike existing methods that use fixed rules to evaluate relevant data and generate risk assessment results, this application provides a novel evaluation method based on anomaly prediction and risk assessment models, which can automatically and accurately complete the data evaluation processing corresponding to the user's physiological signals, effectively improving the accuracy of the generated risk assessment results.

[0146] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0147] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

[0148] It should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application has been authorized (with the knowledge and consent) by the relevant parties or has been fully authorized by all parties, and the executing entity may obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals.

Claims

1. A data evaluation method based on artificial intelligence, characterized in that, Includes the following steps: The user's physiological signals are collected based on a preset multimodal sensor, and a specified physiological signal corresponding to a preset data type is obtained from the physiological signals. The specified physiological signal is processed based on a preset preprocessing strategy to obtain the corresponding target signal; Based on a preset feature type, feature extraction is performed on the target signal to obtain the corresponding target features; The target features are predicted based on a preset anomaly prediction model to obtain the corresponding anomaly prediction results. If the anomaly prediction result meets the preset alarm triggering conditions, then corresponding event data is generated based on the target signal and the target features; Acquire long-term health data, basic user information, medical history data, and environmental data corresponding to the user; The event data, long-term health data, user basic information, medical history data, and environmental data are evaluated and processed based on a preset risk assessment model to obtain the corresponding risk assessment results. The risk assessment results are sent to the user.

2. The data evaluation method based on artificial intelligence according to claim 1, characterized in that, The step of extracting features from the target signal based on a preset feature type to obtain the corresponding target features specifically includes: The target signal is subjected to time-domain feature extraction to obtain the corresponding time-domain features; Frequency domain features are extracted from the target signal to obtain the corresponding frequency domain features; Nonlinear feature extraction is performed on the target signal to obtain the corresponding nonlinear features; The target signal is subjected to morphological feature extraction to obtain the corresponding morphological features; The time-domain features, frequency-domain features, nonlinear features, and morphological features are integrated to obtain the corresponding integrated features; The integrated feature is used as the target feature.

3. The data evaluation method based on artificial intelligence according to claim 1, characterized in that, The step of predicting the target features based on a preset anomaly prediction model to obtain the corresponding anomaly prediction result specifically includes: Invoke a preset anomaly prediction model; wherein the anomaly prediction model includes a feature extraction layer, a temporal modeling layer, and a classification head; Based on the feature extraction layer, information is extracted from the target features to obtain the corresponding key feature information; Based on the aforementioned temporal modeling layer, the key feature information is processed using temporal modeling to obtain the corresponding output data; The output data is classified based on the classification header to obtain the corresponding classification result. The classification result is used as the anomaly prediction result.

4. The data evaluation method based on artificial intelligence according to claim 1, characterized in that, The step of generating corresponding event data based on the target signal and the target features specifically includes: Obtain the timestamp information corresponding to the anomaly prediction result; Perform a hash operation on the target signal to obtain the corresponding hash data; The target signal, the target feature, the timestamp information, and the hash data are integrated to obtain the corresponding integrated data. The integrated data is used as the event data.

5. The data evaluation method based on artificial intelligence according to claim 1, characterized in that, The step of evaluating and processing the event data, long-term health data, user basic information, medical history data, and environmental data based on a preset risk assessment model to obtain the corresponding risk assessment results specifically includes: Feature extraction is performed on multi-source data to obtain corresponding feature data; wherein, the multi-source data includes the event data, the long-term health data, the user's basic information, the medical history data, and the environmental data; Obtain the feature weights corresponding to the feature data; Call the preset risk score calculation formula; Based on the risk assessment model, the feature data and feature weights are calculated and processed using the risk score calculation formula to obtain the corresponding risk score; The risk scores are then subjected to a level mapping process to obtain the corresponding target risk level; The target risk level is used as the risk assessment result.

6. The data evaluation method based on artificial intelligence according to claim 5, characterized in that, The step of obtaining the feature weights corresponding to the feature data specifically includes: Invoke the preset weight generation algorithm; The weight generation algorithm is used to generate weights for the specified feature data to obtain the corresponding initial weights; wherein the specified feature data is any one of the data contained in the feature data. The initial weights are adjusted based on a preset weight constraint strategy to obtain the corresponding specified weights; The specified weights are used as the feature weights of the specified feature data.

7. The data evaluation method based on artificial intelligence according to claim 1, characterized in that, After the step of evaluating and processing the event data, long-term health data, user basic information, medical history data, and environmental data based on a preset risk assessment model to obtain the corresponding risk assessment results, the method further includes: Based on a preset information generation strategy, early warning information corresponding to the risk assessment results is generated; Determine the target notification method corresponding to the risk assessment results; Based on the target notification method, the warning information is sent to the user.

8. A data evaluation device based on artificial intelligence, characterized in that, include: The processing module is used to collect the user's physiological signals based on a preset multimodal sensor, and to obtain a specified physiological signal corresponding to a preset data type from the physiological signals; The preprocessing module is used to process the specified physiological signal based on a preset preprocessing strategy to obtain the corresponding target signal; The extraction module is used to extract features from the target signal based on a preset feature type to obtain the corresponding target features; The prediction module is used to perform prediction processing on the target features based on a preset anomaly prediction model to obtain the corresponding anomaly prediction results. The first generation module is used to generate corresponding event data based on the target signal and the target features if the anomaly prediction result meets the preset alarm triggering conditions. The acquisition module is used to acquire long-term health data, basic user information, medical history data, and environmental data corresponding to the user. The assessment module is used to assess and process the event data, the long-term health data, the user's basic information, the medical history data, and the environmental data based on a preset risk assessment model, and obtain the corresponding risk assessment results. The first sending module is used to send the risk assessment results to the user.

9. A computer device, characterized in that, It includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the data evaluation method based on artificial intelligence as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the data evaluation method based on artificial intelligence as described in any one of claims 1 to 7.

Images