Wearable undergarment heart rate detection method and system

Abstract

The present application relates to the technical field of heart rate detection, and particularly relates to a wearable underwear heart rate detection method and system, the method comprising the following steps: acquiring motion information of a user, identifying a specific body posture change based on the motion information of the user; generating a physiological artifact signal related to the specific body posture change based on the specific body posture change; removing the physiological artifact signal from an original physiological signal to obtain a purified physiological signal; and performing heart rate anomaly judgment on the purified physiological signal. Through the above scheme, the physiological artifact signal can be effectively removed, and the accuracy of heart rate detection is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of heart rate detection, and in particular to a wearable underwear heart rate detection method and system. BACKGROUND

[0002] Currently, in wearable health monitoring devices, real-time analysis of heart rate data to identify potential abnormal conditions is one of its core functions. However, when users are engaged in daily activities, the small and irregular activities of the device can cause changes in the sensor-skin contact state, thereby introducing non-periodic motion artifacts. SUMMARY

[0003] The present application aims to address the above-mentioned deficiencies by providing a wearable underwear heart rate detection method and system.

[0004] The present application adopts the following technical solutions:

[0005] A wearable underwear heart rate detection method, comprising the following steps:

[0006] Obtaining user motion information, and identifying specific body posture changes based on the user motion information;

[0007] Based on the specific body posture changes, generating a physiological artifact signal related to the specific body posture changes;

[0008] Removing the physiological artifact signal from the original physiological signal to obtain a purified physiological signal;

[0009] Performing heart rate anomaly judgment on the purified physiological signal.

[0010] Through the above-mentioned scheme, the physiological artifact signal can be effectively removed, the accuracy of heart rate detection can be improved, and false negatives caused by data discarding can be avoided.

[0011] Optionally, the present application further provides that the step of generating a physiological artifact signal related to the specific body posture changes based on the specific body posture changes comprises:

[0012] Extracting dynamic parameters related to the specific body posture changes from the user's motion information;

[0013] Based on the specific body posture changes and the corresponding dynamic parameters, generating a physiological artifact signal through an artifact prediction model, wherein at least one of the shape, amplitude and duration of the physiological artifact signal is dynamically adjusted according to the dynamic parameters.

[0014] Through the above-mentioned scheme, the shape, amplitude and duration of the physiological artifact signal can be dynamically adjusted according to the dynamic parameters, making the artifact removal more accurate.

[0015] Optionally, the application further proposes that, after the step of removing the physiological artifact signal from the original physiological signal to obtain the purified physiological signal, further comprising:

[0016] identifying a sustained fluctuation signal component in the purified physiological signal, the sustained fluctuation signal component being a non-pulsatile signal, and the frequency of the sustained fluctuation signal component partially overlapping with the heart rate frequency range;

[0017] performing suppression processing on the sustained fluctuation signal component.

[0018] Through the above scheme, the sustained fluctuation signal component partially overlapping with the heart rate frequency range can be further identified and suppressed, and the signal purity is further improved.

[0019] Optionally, the application further proposes that the step of performing heart rate abnormality judgment on the purified physiological signal comprises:

[0020] obtaining and updating heart rate ranges and heart rate variability parameters of the user in different activity states to form a parameter set;

[0021] identifying the current activity state of the user;

[0022] According to the current activity state of the user, selecting the heart rate range and the heart rate variability parameter corresponding to the current activity state of the user from the parameter set as the heart rate abnormality judgment standard;

[0023] identifying the heart rate pulsation from the purified physiological signal;

[0024] comparing the heart rate pulsation with the heart rate abnormality judgment standard;

[0025] determining whether there is a heart rate abnormality according to the comparison result.

[0026] Through the above scheme, the heart rate abnormality judgment standard can be dynamically adjusted according to the current activity state of the user, and the accuracy and adaptability of the abnormality judgment are improved.

[0027] Optionally, the application further proposes that the step of obtaining and updating the heart rate range and the heart rate variability parameter of the user in different activity states to form the parameter set comprises:

[0028] obtaining the original mechanical impact signal of the user, and pre-processing the original mechanical impact signal to obtain the pre-processed mechanical impact signal;

[0029] identifying the heart rate pulsation from the purified physiological signal;

[0030] taking the occurrence time point of the heart rate pulsation as a starting point, and identifying a characteristic event synchronized with the heart rate pulsation from the pre-processed mechanical impact signal within a preset time window after the starting point;

[0031] calculate a time difference between a time point of occurrence of the heart rate beat and a time point of occurrence of the characteristic event;

[0032] determine whether the time difference falls within a time threshold range to confirm authenticity of the heart rate beat;

[0033] use the heart rate beat confirmed to be authentic to obtain and update a heart rate range and a heart rate variability parameter of the user in different activity states, to form a parameter set;

[0034] The parameter set is limited to data corresponding to a case where the purified physiological signal is not determined to be abnormal.

[0035] Through the above scheme, the authenticity of the heart rate beat can be confirmed through the characteristic event synchronous with the mechanical impact signal and the heart rate beat, and the reliability of the parameter set used for heart rate abnormality determination can be ensured.

[0036] Optionally, the step of performing suppression processing on the sustained fluctuation signal component includes:

[0037] performing frequency analysis on the purified physiological signal to extract dominant frequency characteristics and energy distribution intervals of the sustained fluctuation signal component;

[0038] configuring parameters of the filter based on the dominant frequency characteristics and energy distribution intervals of the sustained fluctuation signal component;

[0039] applying the configured filter to the purified physiological signal to suppress the sustained fluctuation signal component.

[0040] Through the above scheme, the filter parameters can be configured based on the dominant frequency characteristics and energy distribution intervals of the sustained fluctuation signal component, and accurate suppression of the sustained fluctuation signal component can be achieved.

[0041] Optionally, the application further proposes that, in the process of suppressing the sustained fluctuation signal component, the integrity of the beat waveform in the purified physiological signal is monitored, and when beat waveform distortion is detected, the suppression strength of the filter is reduced.

[0042] Through the above scheme, the integrity of the beat waveform can be monitored in the suppression process, and the distortion of the real heart rate signal caused by excessive filtering can be avoided.

[0043] Optionally, the application further proposes that the step of monitoring the integrity of the beat waveform in the purified physiological signal includes:

[0044] identifying a heart rate beat from the purified physiological signal;

[0045] extracting beat feature parameters from the heart rate beat, the beat feature parameters including a peak amplitude, a waveform width, and a rising edge slope;

[0046] acquire current wearing state information of the user;

[0047] adjust a reference range of the pulsation feature parameter according to the current wearing state information of the user;

[0048] compare the pulsation feature parameter with the adjusted reference range of the pulsation feature parameter;

[0049] determine the integrity of the pulsation waveform in the purified physiological signal according to the comparison result.

[0050] Through the above scheme, the integrity of the pulsation waveform can be accurately determined by extracting the pulsation feature parameter and dynamically adjusting the reference range according to the wearing state.

[0051] Optionally, the application further provides that the step of adjusting the reference range of the pulsation feature parameter according to the current wearing state information of the user comprises:

[0052] acquire inertial measurement data and strain data of multiple positions in the underwear;

[0053] determine the current wearing state of the user based on a preset wearing state recognition model;

[0054] output a corresponding wearing state label according to the current wearing state of the user, and the wearing state label is used for reference range adjustment of the pulsation feature parameter.

[0055] Through the above scheme, the wearing state of the user can be accurately determined by acquiring the inertial measurement data and the strain data and based on the wearing state recognition model, thereby providing a basis for dynamic adjustment of the pulsation feature parameter.

[0056] Optionally, the application further provides a wearable underwear heart rate detection system applied to the above wearable underwear heart rate detection method, and the system comprises:

[0057] an information processing module configured to acquire motion information of the user and recognize a specific body posture change based on the motion information of the user;

[0058] a physiological artifact signal generation module configured to generate a physiological artifact signal related to the specific body posture change based on the specific body posture change;

[0059] a physiological signal purification module configured to remove the physiological artifact signal from an original physiological signal to obtain a purified physiological signal;

[0060] a heart rate anomaly judgment module configured to perform heart rate anomaly judgment on the purified physiological signal.

[0061] Through the above scheme, a system for implementing the above heart rate detection method is provided, which is convenient for actual application and deployment.

[0062] From the above, the wearable underwear heart rate detection method and system provided by the application can effectively remove physiological artifact signals, improve the accuracy of heart rate detection, and avoid false negatives caused by data loss.

[0063] To enable further understanding of the features and technical contents of the present application, please refer to the following detailed description and drawings of the present application. However, the drawings provided are only for reference and illustration, and are not intended to limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 The method flow chart of the wearable underwear heart rate detection method of the present application is shown in the following figure.

[0065] Figure 2 The structure schematic diagram of the wearable underwear heart rate detection system of the present application is shown in the following figure. DETAILED DESCRIPTION

[0066] The following is to illustrate the embodiments of the present application through specific specific embodiments, and the advantages and effects of the present application can be understood by the person skilled in the art from the disclosed content of the present application. The present application can be implemented or applied through other different specific embodiments, and each detail in the present application can be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. In addition, the drawings of the present application are only simple schematic illustrations, not actual size drawings, and the prior declaration is made. The following embodiments will further illustrate the related technical content of the present application in detail, but the disclosed content is not intended to limit the protection scope of the present application.

[0067] The present embodiment provides a wearable underwear heart rate detection method and system, which combines Figure 1 and Figure 2 as shown.

[0068] Referring to Figure 1 , a wearable underwear heart rate detection method, the method comprising the following steps:

[0069] obtaining the motion information of the user, and identifying the specific body posture change based on the motion information of the user;

[0070] generating a physiological artifact signal related to the specific body posture change based on the specific body posture change;

[0071] removing the physiological artifact signal from the original physiological signal to obtain a purified physiological signal;

[0072] judging the heart rate anomaly of the purified physiological signal.

[0073] wherein, the motion information refers to data reflecting the user's physical activity state, which can be realized by inertial measurement unit data, accelerometer data, gyroscope data or pressure sensor data, such as collected by sensors built-in the underwear, mainly to obtain the user's body behavior, providing data for recognizing posture changes. The specific body posture change refers to a body posture or action transition associated with the generation of physiological artifacts, which can be realized by threshold-based accelerometer signal analysis, pattern recognition algorithm or machine learning model, such as the transition from sitting to standing, arm lifting or torso bending, etc., mainly to identify the behavior patterns that may cause physiological artifacts. The physiological artifact signal refers to the interference component superimposed on the original physiological signal caused by non-cardiac physiological activity or body movement, which can be realized by artifact prediction model, signal synthesis technology or artifact template library based on empirical data, such as signal fluctuations caused by muscle contraction, blood vessel pressure change or poor sensor-skin contact, mainly to simulate and characterize the interference caused by specific body posture changes for subsequent removal. The original physiological signal refers to the unprocessed bioelectric or optical signal containing heart rate information and various disturbances collected directly from the user's body, which can be obtained by photoplethysmography (PPG) sensor, electrocardiogram (ECG) sensor or bioelectric impedance sensor, such as the optical signal collected by the PPG sensor integrated in the underwear, mainly to provide a data stream for heart rate detection. The purified physiological signal refers to the physiological data that has been processed, removed or suppressed physiological artifact signal, which can be realized by adaptive filtering, signal decomposition and reconstruction or artifact subtraction, such as the signal obtained by subtracting the predicted artifact signal from the original PPG signal, mainly to provide a signal closer to the heart rate beat, improving the accuracy of subsequent heart rate analysis. The heart rate abnormality judgment refers to the process of evaluating whether the heart rate beat identified in the purified physiological signal deviates from the normal physiological range or pattern according to the set standard or model, which can be realized by methods such as heart rate range comparison, heart rate variability analysis or machine learning classifier, such as comparing the current heart rate with the user's historical heart rate data or medical standards, mainly to find possible cardiovascular health problems of the user.

[0074] The core innovation of the present application is to combine the acquisition of user motion information and the recognition of specific body posture changes with the generation of physiological artifact signals based on specific body posture changes to predict and remove, thereby solving the problem that existing methods are difficult to distinguish heart rate beats from physiological artifacts when the user's body posture changes, achieving the effect of improving heart rate detection accuracy in motion scenarios and avoiding false negatives of heart rate abnormalities caused by data discarding.

[0075] In particular, the scheme of the present application is to acquire the motion information of the user, such as data from the inertial measurement unit, and then identify the specific body posture change that the user is currently in. This identification process has an effect because it enables the system to cope with the physiological disturbance caused by different posture changes. Based on the identified specific body posture change, the system can generate a physiological artifact signal related to the posture change. This generation process is not simply the application of a general model, but dynamically simulates the morphology, amplitude and duration of the artifact according to the situation of the posture change, so as to reflect the actual disturbance. Subsequently, the generated physiological artifact signal is removed from the original physiological signal, such as by signal subtraction or adaptive filtering technique, to obtain a purified physiological signal. This step eliminates the disturbance introduced by the body posture change, making the signal closer to the physiological pulsation. Finally, the purified physiological signal is judged for heart rate abnormalities. Since the signal has been processed by artifact removal, the judgment process can avoid misjudging the artifact as a heart rate abnormality, thereby improving the correct judgment of heart rate abnormality detection. It is precisely because of this strategy of posture change-based artifact prediction and removal that the system can still obtain and analyze heart rate data stably when the user is performing daily activities, especially when the body posture changes, avoiding the problem of abnormal omission caused by data disturbance in existing methods.

[0076] In some embodiments, the present application is implemented as follows: first, the inertial measurement unit integrated in the underwear continuously acquires acceleration data and angular velocity data of the user, which constitute the motion information of the user. A built-in posture recognition algorithm, such as a model based on support vector machine or decision tree, analyzes these motion information in real time to identify specific body posture changes, such as the transition from sitting to standing, the lifting action of the arm or the bending of the torso. Once a specific body posture change is identified, such as the user switching from sitting to standing, the system generates a physiological artifact signal related to this posture change according to an artifact prediction model. The model can be a signal generator that synthesizes a fluctuating signal, such as a wave signal, with a specific frequency and amplitude according to the amplitude, duration, etc. of the posture change. Subsequently, from the original physiological signal acquired by the photoplethysmography (PPG) sensor in the underwear, the previously generated physiological artifact signal is removed by filtering algorithm or signal subtraction, to obtain a purified physiological signal. For example, a least mean square (LMS) filter can use the generated artifact signal as input to separate the artifact component from the original PPG signal. Finally, the purified physiological signal is judged for heart rate abnormalities. This can be achieved by identifying the waveform in the purified signal, calculating the heart rate, and comparing it with the user's heart rate range. For example, if the heart rate value for several seconds in a row exceeds the user's heart rate range in the current activity state, the system will determine that there is a heart rate abnormality.

[0077] The application further proposes that the step of generating a physiological artifact signal related to a specific body posture change based on the specific body posture change comprises:

[0078] extracting a dynamic parameter related to the specific body posture change from the motion information of the user;

[0079] generating a physiological artifact signal through an artifact prediction model based on the specific body posture change and the corresponding dynamic parameter, wherein at least one of the morphology, amplitude, and duration of the physiological artifact signal is dynamically adjusted according to the dynamic parameter.

[0080] Wherein, the motion information refers to data that can reflect the user's body movement state, which can be achieved by using data collected by inertial measurement unit (IMU) sensors such as accelerometers, gyroscopes, magnetometers, or other means such as visual sensors, pressure sensors, etc., the purpose of which is to provide basic data for identifying body posture changes and extracting dynamic parameters; the specific body posture change refers to the process of the user's body changing from one posture to another, such as from a sitting posture to a standing posture, from stillness to walking, or the lifting of an arm, etc., which can be identified by using a pre-set posture recognition algorithm combined with feature patterns in the motion information, the purpose of which is to determine the specific scenario in which the physiological artifact signal needs to be generated; the dynamic parameter refers to a numerical value that can quantify the kinematic characteristics such as intensity, speed, acceleration, or duration of the specific body posture change, which can be characterized by using indicators such as amplitude, rate of change, frequency components, or energy distribution of the motion information, the purpose of which is to provide refined input for the artifact prediction model to achieve dynamic adjustment of the artifact signal; the artifact prediction model refers to a calculation model that can predict the characteristics of the physiological artifact signal according to the input data, which can be implemented by using a regression model based on machine learning, a neural network model (such as a recurrent neural network or a long short-term memory network), or a lookup table or mathematical function established based on empirical data, the purpose of which is to generate the corresponding physiological artifact signal according to the body posture change and its dynamic parameters; the morphology, amplitude, and duration of the physiological artifact signal refer to the waveform characteristics, signal strength, and length of time that the physiological artifact signal exists in the time domain, which can be described by using the waveform shape, peak value, root mean square value, and time span from the beginning to the end of the signal, the purpose of which is to ensure that the generated artifact signal can more accurately simulate the characteristics of the real physiological artifact, thereby improving the accuracy of artifact removal.

[0081] The scheme of the present application improves the accuracy of heart rate detection by fine control of the generation process of physiological artifact signals. Specifically, after identifying a specific body posture change, the system further extracts dynamic parameters related to the specific body posture change from the user's motion information. These dynamic parameters, such as the intensity, speed or duration of the motion, can quantify the specific characteristics of the body posture change. Subsequently, the system generates physiological artifact signals based on the identified specific body posture change and these extracted dynamic parameters through an artifact prediction model. The artifact prediction model does not simply generate a fixed pattern of artifact signals, but can dynamically adjust the morphology, amplitude or duration of the generated physiological artifact signals according to the input dynamic parameters. For example, when the dynamic parameters indicate that the body posture change is more intense, the artifact prediction model can generate physiological artifact signals with larger amplitude and longer duration; conversely, when the dynamic parameters indicate that the body posture change is relatively gentle, the artifact prediction model generates artifact signals with smaller amplitude and shorter duration. This dynamic adjustment mechanism makes the generated physiological artifact signals more accurately simulate the physiological artifacts caused by different degrees of body posture change in real scenarios. This accurate physiological artifact signal generation method, combined with the step of removing physiological artifact signals from the original physiological signals in the present application, forms a more complete artifact removal process. By providing an artifact signal that closely matches the actual physiological artifact, the artifact component can be more effectively removed or suppressed from the original physiological signal, resulting in a more pure physiological signal. This pure physiological signal provides a high-quality data basis for subsequent heart rate abnormality judgment, significantly reducing the risk of false positives or false negatives due to artifact residues, thereby improving the overall reliability and accuracy of heart rate detection. It is precisely because of the improvement in the accuracy of artifact signal generation that the entire heart rate detection method can still maintain high performance in complex motion scenarios, effectively solving the problem of decreased heart rate detection accuracy caused by incomplete artifact removal in traditional methods.

[0082] In some preferred embodiments, the present application is implemented as follows: when the inertial measurement unit (IMU) sensors, such as tri-axial accelerometers and tri-axial gyroscopes, in the wearable undergarment worn by the user collect the motion information of the user, the system can first preprocess these motion information to identify specific body posture changes, such as the action of the user switching from a sitting posture to a standing posture. Once such specific body posture changes are identified, the system further extracts dynamic parameters related to this posture change from the accelerometer data, for example, the peak value of acceleration, the root mean square value of acceleration, or the duration of the posture change can be calculated. Subsequently, the system can use a pre-trained artifact prediction model to generate a physiological artifact signal. The artifact prediction model can be a deep learning-based recurrent neural network (RNN), such as a long short-term memory network (LSTM), whose input includes the type of specific body posture change identified and the dynamic parameters extracted. The model is trained to learn the complex mapping relationship between different body posture changes and their dynamic parameters and the corresponding physiological artifact signals. In generating the physiological artifact signal, the model can dynamically adjust the morphology, amplitude, or duration of the generated artifact signal according to the input dynamic parameters. For example, if the extracted acceleration peak value is high, indicating that the posture change is more intense, the artifact prediction model can generate a physiological artifact signal with a larger amplitude and a steeper waveform; if the duration of the posture change is longer, the model can correspondingly extend the duration of the generated artifact signal. In this way, the generated physiological artifact signal can more accurately reflect the physiological artifact caused by the actual motion intensity and duration of the user, thereby providing a highly matched reference signal for the subsequent artifact removal step.

[0083] The present application further proposes that after the step of removing the physiological artifact signal from the original physiological signal to obtain the purified physiological signal, further comprising:

[0084] Identifying the sustained fluctuation signal component in the purified physiological signal, the sustained fluctuation signal component being a non-pulsatile signal, and the frequency of the sustained fluctuation signal component partially overlapping with the heart rate frequency range;

[0085] Performing suppression processing on the sustained fluctuation signal component.

[0086] The identifying the persistent fluctuation signal component in the purified physiological signal refers to distinguishing the non-pulsatile interference signal with a frequency partially overlapping with the heart rate frequency range in the purified physiological signal by analyzing the purified physiological signal. Specifically, it can be through time domain or frequency domain analysis method, for example, wavelet transform, Fourier transform or adaptive filtering technology can be used to detect whether there is persistent non-pulsatile fluctuation in a specific frequency range in the signal, and the purpose is to accurately locate and distinguish the specific noise type that causes interference to heart rate detection. The persistent fluctuation signal component is a non-pulsatile signal, and the frequency of the persistent fluctuation signal component partially overlaps with the heart rate frequency range, specifically, such signal is not generated by the periodic beating of the heart, but is caused by other physiological activities or external interference, for example, muscle tremor, respiratory movement or environmental noise, and its frequency characteristic overlaps with the frequency range of normal heart rate, so that the traditional heart rate filtering method is difficult to completely separate it, and the purpose is to clearly need to process the characteristics of the interference signal in order to take targeted suppression strategy. The suppression processing on the persistent fluctuation signal component refers to taking corresponding signal processing technology to reduce or eliminate the influence of the identified persistent fluctuation signal component on the purified physiological signal. Specifically, it can be through digital filtering, adaptive noise cancellation or signal reconstruction method, and the purpose is to maximize the reduction of the interference of the persistent fluctuation signal component without damaging the real heart rate signal, so as to improve the purity of the heart rate signal.

[0087] The scheme of the present application ensures the accuracy of heart rate detection by further processing the residual interference in the signal on the basis of the preliminary purification of the physiological signal. Specifically, after removing the physiological artifact signals caused by specific body posture changes from the original physiological signal and obtaining a preliminary purified physiological signal, the scheme does not stop here. On the contrary, it further analyzes the purified physiological signal in depth to identify the possible existence of sustained fluctuation signal components in it. These components are clearly defined as non-pulsatile signals, and their frequency partially overlaps with the heart rate frequency range, which means that they are not derived from heart beats, but are easily confused with real heart rate signals and significantly interfere with subsequent heart rate detection. It is precisely due to the particularity of this signal component, i.e. its non-pulsatile and overlapping nature with heart rate frequency, that it is difficult to completely eliminate by conventional artifact removal methods. Therefore, after identifying these specific sustained fluctuation signal components, the scheme will perform targeted suppression processing on them. This suppression processing aims to minimize the amplitude of these interference signals while trying to preserve the true heart rate beat information. Through this two-stage signal purification strategy, i.e. first removing the physiological artifacts related to body posture, and then identifying and suppressing other sustained fluctuation signal components, the present scheme can obtain a more pure physiological signal. This enables the subsequent step of heart rate anomaly judgment on the purified physiological signal to be based on higher quality data, significantly reducing the risk of false positives or false negatives, thereby improving the reliability and accuracy of heart rate detection. This hierarchical and refined signal processing procedure effectively solves the problem that traditional methods are difficult to completely remove all interference in a complex noise environment, resulting in limited accuracy of heart rate detection.

[0088] In some preferred embodiments, after removing the physiological artifact signal from the original physiological signal and obtaining the purified physiological signal, the following steps can be implemented. First, in order to identify the persistent fluctuation signal components in the purified physiological signal, frequency analysis can be performed on the purified physiological signal. For example, fast Fourier transform or short-time Fourier transform can be employed to obtain the frequency spectrum of the signal, so as to observe the energy distribution of the signal at different frequencies. By analyzing the frequency spectrum, persistent fluctuations with significant energy peaks but not caused by beats can be identified within the heart rate frequency range (for example, the resting heart rate of an adult is usually 60 to 100 times per minute, corresponding to 1 Hz to 1.67 Hz). These fluctuations can be manifested as specific narrowband noise or broadband noise, and their dominant frequency characteristics and energy distribution intervals can be extracted. Further, in order to perform suppression processing on these identified persistent fluctuation signal components, the parameters of the filter can be configured based on their dominant frequency characteristics and energy distribution intervals. For example, if a persistent tremor noise at a specific frequency is identified, an adaptive filter can be configured, with the center frequency set to the dominant frequency of the noise and the bandwidth adjusted according to its energy distribution interval. Alternatively, if the noise is manifested as a wider frequency range, a band-stop filter or Kalman filter can be used. The configured filter is applied to the purified physiological signal, thereby effectively attenuating or eliminating the persistent fluctuation signal components while maintaining the integrity of the heart rate beat signal as much as possible. For example, in practical applications, a digital signal processor or microcontroller can be used to perform these frequency analysis and filtering operations to achieve real-time or quasi-real-time signal processing.

[0089] The present application further proposes steps for heart rate abnormality judgment on the purified physiological signal, including:

[0090] Obtaining and updating the heart rate range and heart rate variability parameters of the user in different activity states to form a parameter set;

[0091] Identifying the current activity state of the user;

[0092] According to the current activity state of the user, selecting the heart rate range and heart rate variability parameters corresponding to the current activity state of the user from the parameter set as the heart rate abnormality judgment standard;

[0093] Identifying the heart rate beat from the purified physiological signal;

[0094] Comparing the heart rate beat with the heart rate abnormality judgment standard;

[0095] According to the comparison result, determining whether there is a heart rate abnormality.

[0096] The heart rate variability parameter refers to a quantitative index of the slight time interval change between heartbeat cycles, which can be measured by a time domain index such as the standard deviation of adjacent heartbeat intervals, the root mean square of adjacent heartbeat interval differences, or a frequency domain index such as high-frequency power, low-frequency power, and the like, and aims to reflect the regulation ability of the autonomic nervous system on heart activity; the activity state refers to the physiological or physical situation in which the user is currently located, which can be specifically identified by analyzing the user's motion sensor data (such as accelerometer and gyroscope data), geographic location information, or in combination with user input, and aims to provide necessary context information for heart rate judgment; and the parameter set refers to a database or model established for the user under different activity states, containing corresponding heart rate ranges and heart rate variability parameters, which can be implemented by a pre-trained model, a user personalized model updated through continuous learning, or a rule set based on expert experience, and aims to provide personalized and dynamically adjusted judgment basis for heart rate abnormality judgment.

[0097] The scheme of the present application constructs a parameter set for individual differences of the user by acquiring and updating the heart rate range and heart rate variability parameter of the user under different activity states. This parameter set can change over time and can continuously learn and adjust according to the actual physiological data of the user, thereby ensuring that the judgment standard can timely reflect the current situation and has a high degree of accuracy. Subsequently, the system identifies the current activity state of the user, such as rest, walking, running, or sleep, etc. Based on the identified activity state, the system selects the heart rate range and heart rate variability parameter matching the current activity state from the pre-established parameter set, and uses these parameters as the basis for current heart rate abnormality judgment. After obtaining the purified physiological signal, the system identifies the heart rate beats from it, which is the basis for heart rate analysis. Then, the identified heart rate beats are compared with the judgment standard selected according to the current activity state. This comparison not only focuses on whether the heart rate exceeds the normal range, but also considers whether the heart rate variability conforms to the physiological characteristics under the current activity state. Finally, according to the comparison result, the system determines whether there is a heart rate abnormality. This series of steps forms a judgment process that can self-adjust and adapt to the user's situation.

[0098] It is worth noting that the purified physiological signal used in the present scheme is obtained after the original physiological signal is subjected to artifact removal and continuous fluctuation signal suppression processing in the previous steps. This preprocessing ensures the purity of the input signal and effectively reduces the adverse effects of motion artifacts, environmental noise, and other non-cardiac interference on heart rate judgment. Therefore, heart rate abnormality judgment is performed on the basis of a pure physiological signal, which enables the present scheme to accurately identify real heart rate beats and make detailed personalized judgments based on this, thereby effectively avoiding misjudgment and missed judgment caused by signal pollution and improving the stability of heart rate abnormality judgment.

[0099] In some preferred embodiments, the specific process of heart rate abnormality judgment on purified physiological signals can be implemented as follows. First, in order to obtain and update the heart rate range and heart rate variability parameters of the user in different activity states, a parameter set is formed. The system can continuously collect purified physiological signals of the user in different scenarios in daily life, and combine with other sensor data to automatically identify the activity state of the user, such as rest, light activity, moderate activity, strenuous exercise or sleep. For each activity state, the system can establish a baseline model using historical data, such as by calculating the average value and standard deviation of heart rate in this state, and the distribution range of heart rate variability parameters (such as RMSSD). These baseline data can be stored in a local database and updated periodically according to new, confirmed normal physiological data, such as using a moving average or exponential weighted average method to smooth the update, ensuring the dynamic adaptability of the parameter set.

[0100] Then, in order to identify the current activity state of the user, the system can analyze the data from the motion sensors (such as three-axis accelerometer and gyroscope) built into the underwear in real time. By feature extraction on these sensor data, such as calculating activity intensity index, posture change rate, etc., and inputting them into a pre-trained activity classification model (such as support vector machine or neural network), the current activity state label of the user is output.

[0101] Subsequently, according to the current activity state of the user, the system can query and select the heart rate range (for example, the heart rate range in the resting state is 50-90 times / minute) and the normal range of heart rate variability parameters (for example, the normal range of RMSSD in the resting state) corresponding to the activity state label from the previously constructed parameter set. These selected parameters are used as the current heart rate abnormality judgment standard.

[0102] After that, the heart rate beats can be identified from the purified physiological signals using mature signal processing algorithms, such as the method based on wave peak detection. This method can first perform band-pass filtering on the purified physiological signals to remove high-frequency noise and low-frequency baseline drift, and then apply adaptive threshold or morphological operation to identify the R wave peak in the signal, thereby determining the occurrence time point of each heartbeat.

[0103] Then, in order to compare the heart rate beats and the heart rate abnormality judgment standard, the system can calculate the average heart rate and heart rate variability parameters in the current time window. For example, the number of identified heart rate beats in the last 30 seconds can be calculated to obtain the average heart rate, and the RMSSD value of these beat intervals can be calculated. Then, these calculated heart rate and RMSSD values are compared with the previously selected heart rate abnormality judgment standard. For example, it is judged whether the current average heart rate is beyond the selected range, or the RMSSD value is below its normal lower limit.

[0104] Finally, according to the comparison result, it is determined whether there is a heart rate abnormality. If any one or more of the current heart rate or heart rate variability parameters exceeds the corresponding judgment standard, the system can determine that there is a heart rate abnormality, and can trigger a corresponding warning mechanism, such as through vibration, indicator light flickering, or sending a notification to a connected smart device.

[0105] The application further proposes to obtain and update the heart rate range and heart rate variability parameters of the user in different activity states, and the steps of constructing the parameter set include: obtaining the original mechanical impact signal of the user, preprocessing the original mechanical impact signal to obtain the preprocessed mechanical impact signal; identifying the heart rate beat in the purified physiological signal; taking the occurrence time point of the heart rate beat as the starting point, identifying the characteristic event synchronous with the heart rate beat from the preprocessed mechanical impact signal within the preset time window after the starting point; calculating the time difference between the occurrence time point of the heart rate beat and the occurrence time point of the characteristic event; determining whether the time difference falls within the time threshold range to confirm the authenticity of the heart rate beat; using the confirmed authentic heart rate beat to obtain and update the heart rate range and heart rate variability parameters of the user in different activity states to construct the parameter set; wherein the parameter set is limited to the data corresponding to the case where the purified physiological signal is not determined to be heart rate abnormality.

[0106] The original mechanical impact signal refers to the original data stream generated by user body movement or external physical contact and collected by a mechanical sensor, which can be implemented by a three-axis accelerometer, a MEMS gyroscope or a piezoelectric force sensor. The preprocessing refers to a series of signal processing operations on the original mechanical impact signal to remove noise, artifacts or irrelevant components, improve signal quality and feature recognizability, which can be implemented by low-pass filtering, high-pass filtering, band-pass filtering, notch filtering, wavelet denoising or moving average. The preset time window refers to a limited time interval set in advance after the time point of heart rate beat occurrence, which is used to limit the range of searching for synchronous feature events in the mechanical impact signal, which can be implemented by a fixed time period, a time period dynamically adjusted according to user physiological characteristics or a time period adjusted according to activity state. The feature event refers to the recognizable signal pattern or peak value in the preprocessed mechanical impact signal, which has synchronization or specific correlation with the heart rate beat in time, which can be implemented by the peak value, valley value, specific waveform feature or signal energy mutation point of the signal. The time difference refers to the time interval between the occurrence time point of the heart rate beat recognized in the purified physiological signal and the occurrence time point of the synchronous feature event recognized in the preprocessed mechanical impact signal, which can be implemented by the absolute time difference, relative time difference or phase difference of the two event occurrence time. The time threshold refers to an acceptable time range for judging whether the time difference meets the preset synchronization relationship, only when the time difference falls within this range, the heart rate beat is considered to be real, which can be implemented by a fixed numerical range, a range adjusted according to individual physiological differences or a range dynamically adjusted according to activity intensity.

[0107] The scheme of the present application significantly improves the reliability of heart rate beat data by introducing a multi-modal signal verification mechanism, thereby optimizing the accuracy of heart rate abnormality judgment. Specifically, the scheme first acquires the original mechanical impact signal of the user and pre-processes it to obtain high-quality pre-processed mechanical impact signal. This step lays the foundation for subsequent accurate identification of characteristic events from the mechanical impact signal, ensuring the purity of the signal. At the same time, the system identifies the heart rate beat in the purified physiological signal, and takes the occurrence time point of these beats as the time reference. Subsequently, taking the occurrence time point of each heart rate beat as the starting point, within the preset time window, the system searches and identifies the characteristic events synchronized with the heart rate beat from the pre-processed mechanical impact signal. By limiting the time window, the search range can be effectively focused, improving the efficiency and accuracy of synchronous event identification. On this basis, the scheme further calculates the time difference between the occurrence time point of the heart rate beat and the occurrence time point of the identified characteristic event. This time difference is a key indicator of the degree of synchronization of the two signals. Then, the system judges whether the calculated time difference falls within the preset time threshold range. Only when the time difference meets this threshold condition, the heart rate beat is confirmed as a real and valid beat. This multi-modal signal-based time synchronization verification mechanism can effectively eliminate false heart rate beats caused by noise or artifacts, thereby ensuring that the heart rate beat data used for subsequent analysis has a high degree of authenticity. Finally, only these heart rate beats that have been confirmed as real are used to obtain and update the user's heart rate range and heart rate variability parameters under different activity states, thereby forming a parameter set. In addition, in order to further ensure the purity and reliability of the parameter set, the construction and update of the parameter set are limited to the data corresponding to the case where the purified physiological signal is not judged as heart rate abnormality. In this way, the present scheme avoids including disturbed or abnormal heart rate data into the parameter set, thereby ensuring the accuracy and representativeness of the parameter set. Through the above mechanism, the present scheme provides more accurate and reliable judgment criteria for subsequent heart rate abnormality judgment. Compared with the method of relying solely on a single physiological signal to identify heart rate beats and construct a parameter set, the present scheme introduces mechanical impact signal for cross-validation, effectively overcoming the problem of inaccurate heart rate beat identification due to the interference of physiological signals. This multi-signal collaborative verification method makes the obtained heart rate range and heart rate variability parameters more close to the user's real physiological state, thereby significantly improving the reliability of heart rate abnormality judgment and reducing the risk of false positives and false negatives. This strict screening of heart rate beat authenticity directly improves the quality of heart rate abnormality judgment criteria, making the performance of the entire heart rate monitoring system substantially enhanced.

[0108] In some preferred embodiments, the present application is implemented as follows. Firstly, the original mechanical impact signal of the user can be acquired by using the three-axis accelerometer integrated in the wearable device. The accelerometer can capture the tiny vibration and impact of the user's body in real time. Subsequently, the acquired original mechanical impact signal is preprocessed, for example, a low-pass filter with a cutoff frequency of 20 Hz can be applied to remove high-frequency noise, and a moving average filter can be combined to smooth the signal, thereby obtaining the preprocessed mechanical impact signal. At the same time, the system identifies the heart rate pulsation from the purified physiological signal after artifact removal and suppression processing. This can be achieved by applying a mature peak detection algorithm, such as the Pan-Tompkins algorithm, which can accurately identify the R-wave peak in the PPG signal and record its occurrence time point. Then, taking the occurrence time point of each identified R-wave peak of the heart rate pulsation as the starting point, the system searches for a characteristic event in the preprocessed mechanical impact signal within a preset time window, for example, in the range of 50 milliseconds to 150 milliseconds, after the starting point, which is synchronized with the heart rate pulsation. This characteristic event can be specifically a certain waveform peak in the mechanical impact signal related to ventricular contraction, for example, a small and identifiable vibration peak in the mechanical impact signal can be generated after the occurrence of the heart rate pulsation due to the impact of blood ejection on the blood vessel wall. Subsequently, the system calculates the time difference between the occurrence time point of the R-wave peak of the heart rate pulsation and the occurrence time point of the peak of the identified mechanical impact signal characteristic event. For example, if the R-wave peak occurs at T1 and the mechanical impact characteristic event peak occurs at T2, the time difference is T2-T1. Further, the system determines whether the calculated time difference falls within a preset time threshold range, for example, the time threshold can be set to -20 milliseconds to +20 milliseconds. If the time difference is within this range, it is confirmed that the heart rate pulsation is real and valid; otherwise, the heart rate pulsation may be artifact or noise and will be discarded. Finally, only the heart rate pulsations that have passed the reality confirmation will be used to acquire and update the user's heart rate range and heart rate variability parameters in different activity states to form a parameter set. For example, when the user is in a resting state, the system will collect a large amount of verified resting heart rate pulsation data and calculate the average heart rate and heart rate variability (such as SDNN or RMSSD) as the parameters in the resting state. When the user enters a motion state, the system will switch to the parameter update mode in the motion state. It should be noted that only when the purified physiological signal is not determined to be abnormal in heart rate, the corresponding heart rate pulsation data will be included in the update of the parameter set, thereby avoiding the pollution of abnormal data to the normal parameter set.

[0109] The step of performing suppression processing on the sustained fluctuation signal component includes:

[0110] performing frequency analysis on the purified physiological signal to extract the dominant frequency feature and energy distribution interval of the sustained fluctuation signal component;

[0111] configure parameters of the filter based on the dominant frequency characteristic and the energy distribution interval of the continuous fluctuation signal component;

[0112] apply the configured filter to the purified physiological signal to suppress the continuous fluctuation signal component.

[0113] wherein the frequency analysis refers to a process of converting a time-domain signal into a frequency-domain signal through mathematical transformation, which can be implemented by using methods such as fast Fourier transform, short-time Fourier transform or wavelet transform, and the purpose is to reveal the intensity and distribution of different frequency components in the signal; the dominant frequency characteristic of the continuous fluctuation signal component refers to the frequency point or frequency range with the most concentrated energy of the continuous fluctuation signal component in the frequency analysis result, which can be expressed as the peak frequency in the frequency spectrum or the average frequency in a specific frequency band, and the purpose is to determine the central frequency of the noise that needs to be suppressed; the energy distribution interval refers to the range occupied by the continuous fluctuation signal component in the frequency domain and the distribution of its energy at different frequencies, which can be expressed as the bandwidth of the frequency spectrum, the energy decay curve or the energy proportion of a specific frequency range, and the purpose is to comprehensively understand the diffusion degree of the noise in the frequency domain, providing a basis for the accurate configuration of the filter parameters in the subsequent; the configuration of the parameters of the filter refers to the dynamic adjustment of each setting of the filter according to the frequency characteristic of the signal, which can specifically include the adjustment of the cutoff frequency, the center frequency, the bandwidth, the order or the type of the filter, and the purpose is to enable the filter to targetly suppress the target noise while retaining the effective signal to the maximum extent.

[0114] The scheme of the present application can accurately identify the frequency characteristics of the continuous fluctuation signal component in the purified physiological signal by performing frequency analysis on the purified physiological signal. Specifically, by performing frequency analysis on the purified physiological signal, the dominant frequency characteristics and energy distribution interval of the continuous fluctuation signal component can be extracted. These frequency characteristics and energy distribution information are the "fingerprint" of the continuous fluctuation signal component in the frequency domain, which reveals the frequency center and spread range of these non-pulsatile noises. Based on these real-time acquired frequency characteristics and energy distribution interval, the system can intelligently configure the parameters of the filter. This configuration is not based on fixed preset values, but is adaptively adjusted according to the actual frequency characteristics of the continuous fluctuation signal component in the current signal, for example, the cutoff frequency, bandwidth or stopband attenuation of the filter can be adjusted, so as to ensure that the filter can accurately cover and attenuate the target noise frequency. Subsequently, the filter thus finely configured is applied to the purified physiological signal to effectively suppress the continuous fluctuation signal component. The synergistic effect of this series of steps is that it overcomes the limitations of traditional fixed filters in dealing with continuous fluctuation signal components partially overlapping with the heart rate frequency range. On the basis of removing the physiological artifact signal from the original physiological signal and obtaining the purified physiological signal, the present scheme further provides a dynamic and adaptive suppression mechanism for the continuous fluctuation signal component in the purified physiological signal which is difficult to remove by simple artifact removal. By performing real-time frequency analysis on the noise and adjusting the filter accordingly, it can avoid excessive filtering that distorts the effective heart rate pulsation signal, while effectively removing the noise that overlaps with the heart rate signal. This method makes it possible to obtain a more pure heart rate signal in a complex noise environment, thereby providing a high-quality data basis for subsequent heart rate abnormality judgment, significantly improving the accuracy and reliability of heart rate detection.

[0115] In some preferred embodiments, when performing frequency analysis on the purified physiological signal, a fast Fourier transform algorithm can be employed. For example, a 256-point data window can be taken every second from the purified physiological signal for FFT processing, so as to obtain the frequency spectrum information of the signal in this time window. From the frequency spectrum information, the frequency peak of the sustained fluctuation signal component can be identified, for example, if a significant energy peak is detected between 5 Hz and 10 Hz, this frequency can be determined as the dominant frequency characteristic of the sustained fluctuation signal component. At the same time, the energy distribution around this peak can be analyzed, for example, the frequency range at which the energy on both sides of the peak drops to half of the peak value can be calculated, so as to determine its energy distribution interval. Based on these extracted dominant frequency characteristics and energy distribution intervals, an adaptive band-stop filter can be configured. For example, if the dominant frequency characteristic is 8 Hz and the energy distribution interval is 7 Hz to 9 Hz, the center frequency of the band-stop filter can be set to 8 Hz and the bandwidth can be set to 2 Hz. The type of filter can be a Butterworth filter or a Chebyshev filter, and the order of the filter can be dynamically adjusted according to the suppression requirement, for example, a 3rdor 4thorder filter can be selected according to the strength and bandwidth of the noise to achieve the required attenuation effect. Finally, the configured adaptive band-stop filter is applied to the purified physiological signal. This means that each data point of the purified physiological signal will be processed through this filter adjusted according to the real-time noise characteristics. Through this processing, the energy of the sustained fluctuation signal component in the range of 7 Hz to 9 Hz will be significantly attenuated, while the heart rate pulsation signal can be preserved, so as to obtain a more purified heart rate signal in which the sustained fluctuation signal component is effectively suppressed.

[0116] The present application further proposes that in the process of suppressing the sustained fluctuation signal component, the integrity of the pulsation waveform in the purified physiological signal is monitored, and when the pulsation waveform distortion is detected, the suppression strength of the filter is reduced.

[0117] The monitoring of the integrity of the pulsatile waveform in the purified physiological signal refers to real-time or periodic evaluation of whether the pulsatile waveform in the purified physiological signal maintains its expected morphology and characteristics, which can be achieved by waveform feature extraction and threshold comparison, template matching or machine learning model, etc. The purpose is to find the negative impact of filtering on the effective physiological signal in time. The pulsatile waveform distortion refers to the change of the key characteristics of the pulsatile waveform in the purified physiological signal after filtering, such as the wave peak amplitude, waveform width and rising slope, which deviates significantly from the normal or expected state. It can be manifested as waveform flattening, narrowing, amplitude reduction or abnormal sharp peak, etc. The purpose is to identify the damage to the signal quality caused by excessive filtering. Reducing the suppression strength of the filter refers to reducing the degree of attenuation of the target frequency component by the filter, so that the attenuation effect of the filter on the signal is weakened. It can be achieved by reducing the filter coefficient, adjusting the cutoff frequency or bandwidth of the filter, or switching to a filter type with lower suppression strength, etc. The purpose is to reduce the negative impact on the pulsatile waveform and restore its integrity.

[0118] The scheme of the present application introduces a real-time monitoring mechanism for the integrity of the pulsatile waveform in the purified physiological signal during the process of suppressing the continuous fluctuating signal component. Specifically, when the filter is configured according to the dominant frequency characteristics and energy distribution range of the continuous fluctuating signal component and applied to the purified physiological signal to suppress interference, the system will simultaneously and continuously evaluate the pulsatile waveform in the filtered purified physiological signal. Once the distortion of the pulsatile waveform is detected, such as abnormal attenuation of waveform amplitude, distortion of waveform shape or key characteristic parameters beyond the preset range, it indicates that the current filter suppression strength may be too high, causing unnecessary damage to the pulsatile waveform. At this time, the system will immediately respond and dynamically reduce the suppression strength of the filter, such as by adjusting the gain, bandwidth or attenuation coefficient of the filter, to reduce its attenuation effect on the signal. This dynamic adjustment mechanism enables the filter to effectively suppress the continuous fluctuating signal component while avoiding excessive damage to the heart rate pulsatile waveform, thereby balancing noise suppression and effective signal preservation. It is precisely due to this real-time feedback and adaptive adjustment that the scheme can effectively solve the pulsatile waveform distortion problem that may occur in traditional fixed parameter filtering or filtering methods based only on frequency analysis, thereby ensuring the accuracy and reliability of heart rate detection. This method, combined with the previous scheme of configuring filter parameters based on frequency analysis, forms a more robust and intelligent signal processing flow, which can better adapt to complex and variable physiological signal environments and avoid the loss or misjudgment of real heart rate information caused by excessive filtering.

[0119] In some preferred embodiments, when filtering the purified physiological signal to suppress the sustained fluctuation signal component, the following approach can be employed to monitor the integrity of the pulsation waveform in the purified physiological signal and make adjustments. First, each heart rate pulsation waveform can be identified from the purified physiological signal in real time, and its key pulsation feature parameters, such as the peak amplitude, the waveform width (e.g. the half-height width), and the rising edge slope, can be extracted. These parameters can reflect the energy, duration, and steepness of the pulsation waveform. Next, a set of reference ranges for the pulsation feature parameters can be pre-set, which can be obtained from statistical analysis of a large amount of normal physiological signal data, or personalized set according to the user's historical health data. During the filtering process, the currently extracted pulsation feature parameters are compared with the pre-set reference ranges. If any one or more of the pulsation feature parameters is outside its corresponding reference range, for example, the peak amplitude is significantly reduced, the waveform width is abnormally narrowed, or the rising edge slope is significantly reduced, it can be determined that the pulsation waveform has been distorted. Once the pulsation waveform distortion is detected, the system can immediately trigger a feedback mechanism to reduce the suppression strength of the current filter. Specifically, if a digital filter is used, the gain coefficient of the filter can be reduced, or its cutoff frequency can be adjusted to reduce the degree of attenuation of the sustained fluctuation signal component. For example, for a band-stop filter, its passband can be appropriately widened, or its stopband attenuation depth can be reduced. This dynamic adjustment can ensure that the original morphology of the heart rate pulsation waveform is maximally preserved while effectively removing the interference, avoiding loss or misjudgment of heart rate information due to excessive filtering.

[0120] The present application further proposes a step of monitoring the integrity of the pulsation waveform in the purified physiological signal, comprising:

[0121] identifying a heart rate pulsation from the purified physiological signal;

[0122] extracting pulsation feature parameters from the heart rate pulsation, the pulsation feature parameters including a peak amplitude, a waveform width, and a rising edge slope;

[0123] obtaining user current wearing state information;

[0124] dynamically adjusting the reference range of the pulsation feature parameters according to the user current wearing state information;

[0125] comparing the pulsation feature parameters with the adjusted reference range of the pulsation feature parameters;

[0126] determining the integrity of the pulsation waveform in the purified physiological signal according to the comparison result.

[0127] Among them, identifying heart rate beat refers to locating and distinguishing each independent heart rate beat cycle from the continuous purified physiological signal stream, which can be achieved by using technologies such as peak detection, template matching or machine learning classification, and the purpose is to provide analysis objects for subsequent waveform feature extraction; beat feature parameters refer to key values that can quantitatively describe the waveform form of heart rate beat, specifically the peak amplitude, waveform width and rising slope, which can be calculated by using signal processing algorithms on the identified heart rate beat waveform, and the purpose is to provide multi-dimensional data to evaluate the integrity of the waveform; the current wearing state information of the user refers to the data reflecting the contact condition and relative position of the wearable device and the user's body, which can be obtained by using the data collected by the built-in sensor (such as accelerometer, gyroscope, pressure sensor or strain sensor), the information manually input by the user or through image recognition, and the purpose is to provide a basis for subsequent dynamic adjustment of the reference range of beat feature parameters; dynamically adjusting the reference range of beat feature parameters refers to correcting the normal threshold interval of each feature parameter used to judge the integrity of the beat waveform according to the change of the user's wearing state, which can be achieved by using a pre-set lookup table, a statistical model based on historical data or an adaptive algorithm, and the purpose is to improve the accuracy and adaptability of the waveform integrity judgment; determining the integrity of the beat waveform in the purified physiological signal refers to comprehensively evaluating whether the current heart rate beat waveform maintains its inherent and undistorted form according to the comparison result of the beat feature parameters and the adjusted reference range, which can be achieved by using rule-based logical judgment, fuzzy reasoning system or machine learning classifier, and the purpose is to provide a decision basis for whether to adjust the filter suppression strength.

[0128] The scheme of the present application realizes fine monitoring of the integrity of the pulsation waveform in the purified physiological signal through a series of logically rigorous steps, thereby effectively avoiding excessive influence on the real heart rate pulsation waveform while suppressing the continuous fluctuation signal component. First, each independent heart rate pulsation is identified from the purified physiological signal, which is the basis for all subsequent waveform analysis, ensuring the focus on effective signals. On this basis, multi-dimensional pulsation feature parameters are extracted from the identified heart rate pulsations, including peak amplitude, waveform width and rising slope, which can fully quantify the morphological characteristics of the pulsation waveform and provide a quantitative basis for judging whether it is distorted. Considering the inherent influence of the wearing state of the wearable device on the morphological characteristics of the physiological signal waveform, the present scheme further acquires the current wearing state information of the user and dynamically adjusts the reference range of the pulsation feature parameters based on this information. This dynamic adjustment mechanism enables the waveform integrity judgment to adapt to different wearing conditions, avoiding false judgments caused by changes in wearing state and significantly improving the accuracy and adaptability of the judgment. Subsequently, the extracted pulsation feature parameters are compared with the dynamically adjusted reference range, and through this comparison, it can be accurately evaluated whether the current pulsation waveform deviates from its normal form. Finally, according to the comparison result, the system can judge the integrity of the pulsation waveform in the purified physiological signal. When the waveform is distorted, the system can reduce the suppression strength of the filter in time in combination with the foregoing scheme, thereby avoiding irreversible damage to the real heart rate pulsation waveform caused by excessive filtering. It is precisely due to this fine and adaptive waveform integrity monitoring mechanism that the continuous fluctuation signal component can be effectively suppressed while the original form of the heart rate pulsation waveform is maximally preserved, ensuring the accuracy and reliability of heart rate detection.

[0129] In some preferred embodiments, the present application is implemented as follows: when monitoring the integrity of the pulsation waveform in the purified physiological signal, a peak detection algorithm based on adaptive threshold can be first used to identify the heart rate pulsation, for example, by calculating the local mean and standard deviation of the signal to dynamically set the threshold for peak detection to adapt to the changes in signal amplitude. Once the heart rate pulsation is identified, the pulsation feature parameters can be extracted from each pulsation waveform. Specifically, the peak amplitude can be defined as the vertical distance between the peak of the pulsation waveform and the baseline; the waveform width can be defined as the duration at half the peak of the waveform (half-height width); the rising slope can be calculated as the average slope in the time period from the baseline to 80% of the peak of the waveform. To obtain the current wearing state information of the user, a three-axis accelerometer and a strain sensor integrated in the wearable underwear can be used. For example, the accelerometer data can be used to identify the motion pattern and posture of the user's arm, while the strain sensor can measure the tightness of the underwear in contact with the skin. According to these wearing state information, the reference range of the pulsation feature parameters can be dynamically adjusted. For example, a lookup table can be preset, which stores the normal reference interval of the peak amplitude, waveform width and rising slope corresponding to different wearing states (e.g. tight, moderate, loose) and motion intensity (e.g. stationary, light activity, strenuous exercise). When the system identifies that the current wearing state is "loose" and the user is in a "stationary" state, the reference lower limit of the peak amplitude can be appropriately relaxed to avoid misjudging normal but low-amplitude pulsations as distorted. Subsequently, the currently extracted pulsation feature parameters are compared with the adjusted reference range obtained from the lookup table. For example, if the peak amplitude is lower than the adjusted reference lower limit, or the waveform width exceeds the adjusted reference interval, it is considered that there is potential waveform distortion. Finally, a weighted score-based decision mechanism can be used to determine the integrity of the pulsation waveform in the purified physiological signal. For example, a weight is assigned to each pulsation feature parameter that exceeds the reference range, and the total score is calculated. If the total score exceeds a pre-set distortion threshold, it is determined that the pulsation waveform is distorted, and the operation of reducing the filter suppression strength is triggered.

[0130] The present application further proposes that the step of dynamically adjusting the reference range of the pulsation feature parameters according to the current wearing state information of the user comprises:

[0131] Collecting inertial measurement data and strain data at multiple positions in the wearable underwear;

[0132] Judging the current wearing state of the user based on a pre-set wearing state recognition model;

[0133] Outputting a corresponding wearing state label according to the current wearing state of the user, the wearing state label being used for reference range adjustment of the pulsation feature parameters.

[0134] wherein the inertial measurement data refers to data reflecting the motion state of the object in space, which can specifically include acceleration, angular velocity, and attitude, etc. information, and the purpose is to capture the dynamic behavior of the underwear, such as shaking, shifting or relative motion; the strain data refers to data reflecting the deformation degree of the object under force, which can specifically include deformation information such as stretching, compression or bending, and the purpose is to quantify the close fitting degree between the underwear and the body, such as whether it is too loose or too tight; the preset wearing state recognition model refers to a trained and optimized algorithm or system, which can specifically use machine learning models such as support vector machines, neural networks or decision trees, and the purpose is to automatically identify and classify different wearing states according to the input inertial measurement data and strain data, such as good wearing, loose wearing or tight wearing; the wearing state label refers to a classification result that summarizes the description of the identified wearing state, which can specifically be a string, an enumeration value or a numerical coding, and the purpose is to simplify and standardize the complex wearing state information, so that the subsequent module can directly use the label to adjust the reference range of the pulsation feature parameter.

[0135] The scheme of the present application comprehensively acquires multi-dimensional information related to the wearing state by collecting inertial measurement data and strain data at multiple positions in the underwear. Inertial measurement data can reflect the motion state of the underwear, such as whether there is looseness or displacement, while strain data can reflect the fit between the underwear and the body, such as whether it is too tight or too loose. By collecting data at multiple positions, the wearing state of the underwear can be more completely described, avoiding the limitations of a single data source. Based on these multi-source data, the system uses a pre-set wearing state recognition model to judge the current wearing state of the user. The model is pre-trained and can accurately identify different wearing states, such as good wearing, loose wearing, or tight wearing, according to the input inertial measurement data and strain data. This model-based judgment avoids errors caused by subjective judgment and improves the accuracy of wearing state recognition. Subsequently, the system outputs the corresponding wearing state label according to the judged current wearing state of the user. These labels are a general description of the recognition result of the wearing state, such as "good wearing", "loose wearing" or "tight wearing". These labels are provided as key information to the reference range adjustment module of the pulsation feature parameter. Because the current wearing state information of the user can be accurately obtained and converted into a wearing state label that can be used by the system, the scheme can achieve dynamic and accurate adjustment of the reference range of the pulsation feature parameter. For example, when the wearing state is "loose wearing", the system can appropriately relax the reference range of the pulsation feature parameter to avoid misjudging normal pulsation as abnormal; conversely, when the wearing state is "tight wearing", appropriate adjustments can also be made. This accurate wearing state recognition and reference range adjustment mechanism is closely combined with the step of monitoring the integrity of the pulsation waveform in the physiological signal. When monitoring the integrity of the pulsation waveform, the extracted pulsation feature parameter needs to be compared with a reference range. If the reference range does not accurately reflect the current wearing state of the user, it may lead to misjudgment of the integrity of the pulsation waveform, and thus affect the adjustment of the filter suppression intensity and the final judgment of heart rate abnormalities. Through the present scheme, the reference range of the pulsation feature parameter can be dynamically optimized according to the actual wearing state, making the judgment of the integrity of the pulsation waveform more reliable. For example, when the underwear is worn loosely, the pulsation signal may have reduced amplitude or distorted waveform. If the reference range is not adjusted, it may be misjudged as a distorted pulsation waveform, thus unnecessarily reducing the suppression intensity of the filter, resulting in the failure to effectively suppress the persistent fluctuation signal component. Conversely, if it is too tight, it may cause the pulsation signal to be suppressed, and if the reference range is not adjusted, it may also lead to misjudgment.Therefore, the present scheme provides accurate wearing state information, so that the reference range of the pulsation feature parameter is adjusted more reasonably, thereby improving the accuracy of the completeness judgment of the pulsation waveform, and further optimizing the suppression effect of the continuous fluctuation signal component, ultimately improving the accuracy and reliability of the entire heart rate detection method for heart rate anomaly judgment, and effectively solving the problem of inaccurate heart rate anomaly judgment caused by wearing state uncertainty.

[0136] In some preferred embodiments, the wearable underwear can integrate multiple miniature inertial measurement units and flexible strain sensors. For example, inertial measurement units can be deployed at multiple key positions of the underwear, such as the chest, side wings, and shoulder straps. These inertial measurement units can use a combination of three-axis accelerometers and three-axis gyroscopes to collect linear acceleration and angular velocity data of the underwear in different directions. At the same time, flexible strain sensors can be embedded in key areas of the underwear in contact with the skin, such as under the sternum, on both sides of the ribs, and on the shoulders. These sensors can use piezoresistive or capacitive principles to monitor the deformation of the underwear material under stress in real time, thereby reflecting the fit of the underwear. The collected inertial measurement data and strain data can be transmitted to an embedded processor, such as a low-power microcontroller or a dedicated signal processing chip. The processor can pre-store a wearing state recognition model inside, which can be a convolutional neural network or a recurrent neural network based on deep learning. The model is trained in the offline stage through a large amount of inertial measurement data and strain data under different wearing states. For example, data under various states such as good wearing, loose wearing, tight wearing, and underwear displacement can be collected and labeled to train the model to recognize these states. When real-time data is input, the model extracts features and classifies the data, and outputs corresponding wearing state labels, such as "good wearing", "slight looseness", "severe looseness", "tight wearing", or "local displacement". These wearing state labels can then be used as input parameters and transmitted to the reference range adjustment module of the pulsation feature parameter. The module can pre-set a lookup table or a parameter adjustment algorithm to dynamically adjust the upper and lower reference values of the pulsation feature parameters (such as peak amplitude, waveform width, and rising slope) according to the received wearing state label. For example, when the label is "severe looseness", the amplitude of the pulsation waveform may be generally low, so the reference lower limit of the peak amplitude can be appropriately lowered to avoid misjudging normal low-amplitude pulsation as distortion.

[0137] Reference Figure 2 The present application further proposes a wearable underwear heart rate detection system, which is applied to a wearable underwear heart rate detection method. The system comprises:

[0138] An information processing module is configured to obtain motion information of a user and identify a specific body posture change based on the motion information of the user.

[0139] The physiological artifact signal generation module generates a physiological artifact signal related to the specific body posture change based on the specific body posture change;

[0140] The physiological signal purification module is used to remove the physiological artifact signal from the original physiological signal to obtain a purified physiological signal.

[0141] The heart rate anomaly judgment module performs heart rate anomaly judgment on the purified physiological signal.

[0142] Among them, the information processing module refers to a unit responsible for receiving and analyzing user motion data, which can be implemented by a microcontroller integrated with an accelerometer, a gyroscope and other inertial sensors or a special digital signal processor, and its purpose is to perceive the activity state of the user and identify specific body posture changes to provide context information for subsequent signal processing; the physiological artifact signal generation module refers to a unit that constructs or predicts the corresponding interference signal according to the identified body posture change, which can be implemented by a software module based on a pre-trained model or algorithm, which can generate simulated physiological artifact waveforms according to posture change parameters, and its purpose is to provide accurate artifact reference for subsequent signal purification; the physiological signal purification module refers to a unit that separates and removes interference components from the original physiological signal, which can be implemented by a digital filter, an adaptive noise cancellation algorithm or a blind source separation algorithm, and its purpose is to improve the purity of the heart rate signal and ensure the accuracy of subsequent analysis; the heart rate anomaly judgment module refers to a unit that analyzes the processed physiological signal to identify potential heart rate anomalies, which can be implemented by a software module based on a rule engine, a machine learning classifier or a statistical analysis model, and its purpose is to provide real-time health warnings for users.

[0143] The scheme of the present application constructs a system capable of stably and reliably performing heart rate detection tasks by embodying key steps in the heart rate detection method as independent system modules. Specifically, the information processing module first acquires the user's motion information and identifies specific body posture changes based on these information. These posture change information is then passed to the physiological artifact signal generation module, which uses this information to generate physiological artifact signals related to specific body posture changes. These generated artifact signals are not directly used for heart rate determination, but rather serve as a reference or model to guide the work of the physiological signal purification module. The physiological signal purification module receives the original physiological signal and, in combination with the artifact information provided by the physiological artifact signal generation module, accurately removes these physiological artifact signals from the original physiological signal, thereby obtaining a more pure physiological signal. Finally, the heart rate abnormality judgment module analyzes these purified physiological signals to identify whether there is a heart rate abnormality. It is due to this modular design that each functional unit has a clear responsibility and can be optimized for its respective task, for example, the information processing module can focus on the accuracy of motion data acquisition and posture recognition, the physiological artifact signal generation module can focus on the accurate construction of artifact models, the physiological signal purification module can focus on efficient artifact removal, and the heart rate abnormality judgment module can focus on accurate analysis of the purified signal. This division of labor and cooperation of the system architecture makes the entire heart rate detection process in the dynamic and complex application scenario of wearing underwear able to effectively cope with the interference brought by body activity and wearing state changes, ensuring the stability and accuracy of heart rate detection. Compared with merely describing the method steps, the present system provides specific execution carriers to ensure the effective landing of the method steps and the close cooperation between the functions, thereby overcoming the problem that a simple method cannot guarantee the execution effect, and significantly improving the reliability of heart rate monitoring.

[0144] In some preferred embodiments, the present application is implemented as follows: the wearable undergarment heart rate detection system can be integrated on a microcircuit board and embedded in a specific location of the wearable undergarment. The information processing module can be a low-power microcontroller, such as one based on the ARM Cortex-M series, which integrates multi-axis accelerometer and gyroscope sensors. The microcontroller continuously collects user motion data and analyzes these data in real time through built-in algorithms to identify the current body posture, such as the transition from a sitting to a standing position, the lifting of arms, or the twisting of the torso. The physiological artifact signal generation module can be a software program running on the microcontroller, which contains a lightweight artifact prediction model. When the information processing module identifies specific body posture changes, the model generates a simulated physiological artifact signal based on the type and intensity of the posture change. For example, if a rapid arm lift is identified, the module can generate a simulated artifact waveform with a specific frequency and amplitude that simulates the PPG signal fluctuations caused by changes in venous return. The physiological signal purification module can be a digital signal processing algorithm running on the same microcontroller. The algorithm receives the raw physiological signal from the PPG sensor and combines the simulated artifact signal provided by the physiological artifact signal generation module to remove artifact components from the original signal in real time using adaptive filtering techniques such as the least mean squares (LMS) algorithm. In this way, motion artifacts, respiratory artifacts, and other non-cardiac interference can be effectively suppressed, resulting in a clear heart rate pulsation signal. The heart rate abnormality judgment module can be a rule-based or machine learning-based judgment logic running on the same microcontroller. The module analyzes the waveform of the purified physiological signal, extracts heart rate pulsation period, heart rate variability, and other parameters, and compares them with pre-set or dynamically updated user health baselines. Once the heart rate is detected to be outside the normal range or an abnormal pattern is detected, the system can trigger an alarm or record the abnormal event, and send information to the user's smartphone or other display devices through a wireless communication module (such as a Bluetooth Low Energy module).

[0145] The above disclosed content is only the preferred feasible embodiment of the present application, and does not limit the protection scope of the present application, so any equivalent technical changes made according to the content of the present application specification and drawings are included in the protection scope of the present application, and in addition, the elements can be updated as technology develops.

Claims

1. A method of detecting heart rate by wearing an undergarment, characterized by, The method comprises the following steps: Obtain motion information of the user, and identify a specific body posture change based on the motion information of the user; Generate a physiological artifact signal related to the specific body posture change based on the specific body posture change, specifically, extract a dynamic parameter related to the specific body posture change from the motion information of the user, generate the physiological artifact signal through an artifact prediction model based on the specific body posture change and the corresponding dynamic parameter, wherein at least one of the shape, amplitude and duration of the physiological artifact signal is dynamically adjusted according to the dynamic parameter, the specific body posture change refers to a body posture or action transition that is pre-defined or identified by a learning model and is associated with the generation of a physiological artifact, which can be implemented by using a threshold-based accelerometer signal analysis, a pattern recognition algorithm or a machine learning model; Remove the physiological artifact signal from the original physiological signal to obtain a purified physiological signal, identify a sustained fluctuation signal component in the purified physiological signal, the sustained fluctuation signal component is a non-pulsatile signal, and the frequency of the sustained fluctuation signal component partially overlaps with the heart rate frequency range, and perform suppression processing on the sustained fluctuation signal component; specifically, the step of performing suppression processing on the sustained fluctuation signal component comprises: performing frequency analysis on the purified physiological signal, extracting dominant frequency characteristics and energy distribution intervals of the sustained fluctuation signal component, configuring parameters of a filter based on the dominant frequency characteristics and energy distribution intervals of the sustained fluctuation signal component, and applying the configured filter to the purified physiological signal to suppress the sustained fluctuation signal component, wherein during the suppression of the sustained fluctuation signal component, the integrity of the pulsatile waveform in the purified physiological signal is monitored, and when the pulsatile waveform distortion is detected, the suppression strength of the filter is reduced; specifically, the step of monitoring the integrity of the pulsatile waveform in the purified physiological signal comprises: identifying a heart rate beat from the purified physiological signal, extracting pulsatile feature parameters from the heart rate beat, the pulsatile feature parameters comprising peak amplitude, waveform width and rising slope, obtaining current wearing state information of the user, dynamically adjusting the reference range of the pulsatile feature parameters according to the current wearing state information of the user, comparing the pulsatile feature parameters with the adjusted reference range of the pulsatile feature parameters, and determining the integrity of the pulsatile waveform in the purified physiological signal according to the comparison result. Perform heart rate abnormality judgment on the purified physiological signal.

2. The method of claim 1, wherein the method further comprises: The step of performing heart rate abnormality judgment on the purified physiological signal comprises: Obtain and update heart rate ranges and heart rate variability parameters of the user in different activity states to form a parameter set; Identify the current activity state of the user; According to the current activity state of the user, select the heart rate range and the heart rate variability parameter corresponding to the current activity state of the user from the parameter set as the heart rate abnormality judgment standard; Identify a heart rate beat from the purified physiological signal; Compare the heart rate beat with the heart rate abnormality judgment standard; Determine whether there is a heart rate abnormality according to the comparison result.

3. The method of claim 2, wherein the method further comprises: The step of obtaining and updating the heart rate range and the heart rate variability parameter of the user in different activity states to form a parameter set comprises: The original mechanical impact signal of the user is acquired, and the original mechanical impact signal is preprocessed to obtain a preprocessed mechanical impact signal. The original mechanical impact signal refers to a raw data stream generated by user body movement or external physical contact and collected by a mechanical sensor. The original mechanical impact signal can be realized by using a three-axis accelerometer, a MEMS gyroscope, or a piezoelectric force sensor. A heart rate beat in the purified physiological signal is identified. Taking a time point of occurrence of the heart rate beat as a starting point, a characteristic event synchronized with the heart rate beat is identified from the preprocessed mechanical impact signal within a preset time window after the starting point. A time difference between the time point of occurrence of the heart rate beat and a time point of occurrence of the characteristic event is calculated. It is judged whether the time difference falls within a time threshold range to confirm the authenticity of the heart rate beat. The heart rate beat confirmed to be authentic is used to acquire and update a heart rate range and a heart rate variability parameter of the user in different activity states to form a parameter set. The parameter set is limited to data corresponding to a case where the purified physiological signal is not determined to be abnormal.

4. The method of claim 1, wherein the method further comprises: The step of dynamically adjusting the reference range of the beat characteristic parameter according to the current wearing state information of the user includes: Inertial measurement data and strain data at multiple positions in the underwear are collected. A current wearing state of the user is judged based on a preset wearing state recognition model. A corresponding wearing state label is output according to the current wearing state of the user, and the wearing state label is used for reference range adjustment of the beat characteristic parameter.

5. A wearable undergarment heart rate detection system applied to the wearable undergarment heart rate detection method of claim 1, characterized in that, The system includes: An information processing module is configured to acquire motion information of a user and identify a specific body posture change based on the motion information of the user. A physiological artifact signal generation module is configured to generate a physiological artifact signal related to the specific body posture change based on the specific body posture change. Specifically, dynamic parameters related to the specific body posture change are extracted from the motion information of the user, and the physiological artifact signal is generated by an artifact prediction model based on the specific body posture change and the corresponding dynamic parameters. At least one of the shape, amplitude, and duration of the physiological artifact signal is dynamically adjusted according to the dynamic parameters. The specific body posture change refers to a predefined or learned model-identified body posture or action transition associated with the generation of a physiological artifact. The specific body posture change can be realized by using a threshold-based accelerometer signal analysis, a pattern recognition algorithm, or a machine learning model. The physiological signal purification module is used for removing physiological artifact signals from the original physiological signals to obtain purified physiological signals, identifying a sustained fluctuation signal component in the purified physiological signals, performing suppression processing on the sustained fluctuation signal component, wherein the sustained fluctuation signal component is a non-pulsatile signal, and the frequency of the sustained fluctuation signal component partially overlaps with the heart rate frequency range. Specifically, the step of performing suppression processing on the sustained fluctuation signal component includes: performing frequency analysis on the purified physiological signals, extracting dominant frequency characteristics and energy distribution intervals of the sustained fluctuation signal component, configuring parameters of a filter based on the dominant frequency characteristics and the energy distribution intervals of the sustained fluctuation signal component, and applying the configured filter to the purified physiological signals to suppress the sustained fluctuation signal component. In the process of suppressing the sustained fluctuation signal component, the integrity of the pulsatile waveform in the purified physiological signals is monitored, and when the pulsatile waveform distortion is detected, the suppression strength of the filter is reduced. Specifically, the step of monitoring the integrity of the pulsatile waveform in the purified physiological signals includes: identifying a heart rate pulse from the purified physiological signals, extracting pulsatile feature parameters from the heart rate pulse, the pulsatile feature parameters including a peak amplitude, a waveform width, and a rising edge slope, obtaining current wearing state information of a user, dynamically adjusting a reference range of the pulsatile feature parameters according to the current wearing state information of the user, comparing the pulsatile feature parameters with the adjusted reference range of the pulsatile feature parameters, and determining the integrity of the pulsatile waveform in the purified physiological signals according to a comparison result. The heart rate anomaly judgment module is used for performing heart rate anomaly judgment on the purified physiological signals.

Images