Physiological signal processing method based on human intelligence, edge computing device and medium

Abstract

The invention provides a physiological signal processing method based on human intelligence, an edge computing device and a medium. The method comprises the steps that resting state physiological data, actually measured physiological data, real-time acceleration data and real-time angular velocity data of a testee are collected; creating a system template of adaptive unscented Kalman filtering according to the resting state physiological data; constructing a system equation model according to the system template and the actually measured physiological data; determining exercise intensity according to the real-time acceleration data and the real-time angular velocity data; and in combination with the system equation model and the motion intensity, performing unscented transformation on the state variable and the covariance matrix of the adaptive unscented Kalman filtering at the previous adjacent moments to obtain the updated state variable and covariance matrix at the current moment. According to the method provided by the invention, the nonlinear relationship between the physiological signal and the motion artifact can be accurately handled, and the artifact removal accuracy is improved.

Description

Technical Field

[0001] The present application relates to the technical field of ergonomics, and in particular to a physiological signal processing method based on human factors intelligence, an edge computing device and a medium. Background Art

[0002] Physiological signals can be collected through wearable devices such as smart bracelets and smart watches. For example, PPG (Photoplethysmography) signals can be collected to obtain human physiological parameters (such as heart rate, blood oxygen saturation, etc.) through PPG signals for human health monitoring. However, in actual application scenarios, people using wearable devices will inevitably be in motion, such as walking, running, hand shaking, etc. This kind of body movement makes the PPG signal susceptible to serious interference and produces motion artifacts. Traditional signal processing filtering methods, such as low-pass filtering, high-pass filtering, band-pass filtering, etc., as well as filtering technologies based on specific models, such as ordinary Kalman filtering algorithms, have poor filtering effects and are prone to filtering out useful physiological signal components at the same time or failing to completely remove artifacts. Summary of the invention

[0003] The present application provides a physiological signal processing method, edge computing device and medium based on human factors intelligence, which helps to accurately deal with the nonlinear relationship between physiological signals and motion artifacts and improve the accuracy of artifact removal.

[0004] In a first aspect, the present application provides a method for processing physiological signals based on human factor intelligence, comprising:

[0005] Collect the resting physiological data, measured physiological data, real-time acceleration data and real-time angular velocity data of the subjects;

[0006] Create a system template for adaptive unscented Kalman filtering based on resting-state physiological data;

[0007] Construct a system equation model based on the system template and measured physiological data;

[0008] Determine the exercise intensity according to the real-time acceleration data and the real-time angular velocity data;

[0009] Combined with the system equation model and motion intensity, the state variables and covariance matrix of the adaptive unscented Kalman filter at the previous adjacent moment are untraceably transformed to obtain the updated state variables and covariance matrix at the current moment; wherein, the untraceable transformation includes state prediction and state update, the state variables at the previous adjacent moment are used to characterize the change information of the physiological signal at the previous adjacent moment, the state variables at the current moment are used to characterize the change information of the physiological signal at the current moment, the covariance matrix at the previous adjacent moment is used to characterize the covariance between the various components in the state variables at the previous adjacent moment, and the covariance matrix at the current moment is used to characterize the covariance between the various components in the state variables at the current moment.

[0010] In one possible implementation, the state variables include signal strength and signal strength change rate, and a system equation model is constructed according to the system template and measured physiological data, including:

[0011] Counting the signal strength and change trend of the resting physiological data according to the system template to obtain a first signal strength set and a first change rate set;

[0012] Obtaining a second signal strength set and a second change rate set according to the measured physiological data statistics;

[0013] Perform similarity matching based on the first signal strength set, the first change rate set, the second signal strength set, and the second change rate set to obtain a best matching position, where the best matching position corresponds to the highest similarity in signal strength and change rate;

[0014] Obtaining the reference signal strength change rate at the current moment by indexing from the first change rate set according to the best matching position;

[0015] The reference signal strength at the next adjacent time is calculated based on the signal strength at the current time in the second signal strength set and the reference signal strength change rate at the current time.

[0016] In one possible implementation, the real-time acceleration data includes acceleration components of three coordinate axes at the current moment, and the real-time angular velocity data includes angular velocity components of three coordinate axes at the current moment;

[0017] Determine the intensity of exercise based on real-time acceleration data and real-time angular velocity data, including:

[0018] Calculate the acceleration amplitude according to the acceleration components of the three coordinate axes at the current moment, and calculate the angular velocity amplitude according to the angular velocity components of the three coordinate axes at the current moment;

[0019] The exercise intensity is obtained by performing weighted average calculation based on the acceleration amplitude and angular velocity amplitude.

[0020] In one possible implementation, the state variables and covariance matrix of the adaptive unscented Kalman filter at the previous adjacent time are unscented, combined with the system equation model and the motion intensity, to obtain the updated state variables and covariance matrix at the current time, including:

[0021] Using the adaptive unscented Kalman filter algorithm, a sampling point set is generated according to the state variables and covariance matrix at the previous adjacent time, and the sampling point weights are calculated;

[0022] Using the system equation model, a predicted sampling point set is generated based on the sampling point set, and the mean of the predicted state is calculated, and the predicted state covariance matrix is ​​calculated in combination with the exercise intensity;

[0023] Using the measurement equation, the theoretical measurement value is calculated according to the predicted sampling point set, and the mean of the theoretical measurement value and the measurement covariance matrix are calculated;

[0024] Calculate the cross covariance matrix based on the predicted sampling point set, the mean of the predicted state, the theoretical measurement value, and the mean of the theoretical measurement value;

[0025] Calculate the Kalman gain matrix based on the cross covariance matrix and the measurement covariance matrix;

[0026] Update the state variable according to the Kalman gain matrix, the mean of the predicted state, the mean of the theoretical measured value and the actual measured value at the current moment to obtain the state variable at the current moment;

[0027] The covariance matrix is ​​updated according to the predicted state covariance matrix, the Kalman gain matrix, and the measurement covariance matrix to obtain the covariance matrix at the current moment.

[0028] In one possible implementation, the predicted state covariance matrix is ​​calculated in combination with the exercise intensity, including:

[0029] Calculate the process noise adjustment factor according to the exercise intensity;

[0030] The predicted state covariance matrix is ​​calculated based on the sampling point set, the mean of the predicted state, the process noise adjustment coefficient and the process noise covariance matrix.

[0031] In one possible implementation manner, the method further includes performing outlier removal processing on the actual measurement value based on a comparison between a preset threshold and the actual measurement value.

[0032] In one possible implementation manner, the method further includes: if the value of the exercise intensity is less than a reference value, re-collecting the resting physiological data of the subject to update the system template.

[0033] In a second aspect, the present application provides a physiological signal processing device with human factors intelligence, comprising:

[0034] An acquisition module is used to collect the resting physiological data, measured physiological data, real-time acceleration data and real-time angular velocity data of the subject;

[0035] A creation module, used for creating a system template of an adaptive unscented Kalman filter according to the resting physiological data;

[0036] A construction module, used for constructing a system equation model according to the system template and the measured physiological data;

[0037] A determination module, configured to determine the exercise intensity according to the real-time acceleration data and the real-time angular velocity data;

[0038] The untraceable transformation module is used to combine the system equation model and the motion intensity to perform an untraceable transformation on the state variables and covariance matrix of the adaptive untraceable Kalman filter at the previous adjacent moment to obtain the updated state variables and covariance matrix at the current moment; wherein the untraceable transformation includes state prediction and state update, the state variables at the previous adjacent moment are used to characterize the change information of the physiological signal at the previous adjacent moment, the state variables at the current moment are used to characterize the change information of the physiological signal at the current moment, the covariance matrix at the previous adjacent moment is used to characterize the covariance between the various components in the state variables at the previous adjacent moment, and the covariance matrix at the current moment is used to characterize the covariance between the various components in the state variables at the current moment.

[0039] In a third aspect, the present application provides an edge computing device, comprising: a processor and a memory, the memory being used to store a computer program; the processor being used to run the computer program to implement the physiological signal processing method based on human factors intelligence as described in the first aspect.

[0040] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer-readable storage medium is executed on a computer, the computer implements the physiological signal processing method based on human factors intelligence as described in the first aspect.

[0041] Compared with the prior art, this application has at least the following technical effects:

[0042] The present application provides a physiological signal processing method based on human factors intelligence, by collecting the resting physiological data, measured physiological data, real-time acceleration data and real-time angular velocity data of the subject; creating a system template of an adaptive unscented Kalman filter based on the resting physiological data; building a system equation model based on the system template and the measured physiological data; determining the intensity of exercise based on the real-time acceleration data and the real-time angular velocity data; combining the system equation model and the intensity of exercise, performing an unscented transformation on the state variables and covariance matrix of the adaptive unscented Kalman filter at the previous adjacent moments, and obtaining the updated state variables and covariance matrix at the current moment. Using the adaptive unscented Kalman filter algorithm for unscented transformation can accurately process nonlinear functions, better fit the actual interaction relationship between physiological signals and motion artifacts, thereby more accurately estimating each state variable, effectively separating the state corresponding to the motion artifact from the overall signal, avoiding the error accumulation caused by linear approximation, and improving the accuracy of artifact removal. In addition, the motion intensity is continuously updated during the filtering process, so that the filtering algorithm can adapt to changes in different motion scenes, adapt to dynamic changes in signals in real time, and automatically optimize filtering parameters based on motion intensity, individual differences, etc., to ensure stable artifact removal effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flowchart of a method for processing physiological signals based on human intelligence provided in an embodiment of the present application;

[0044] Figure 2 A schematic diagram of a PPG template is provided for an embodiment of the present application;

[0045] Figure 3 A schematic diagram of the filtering results provided in the embodiment of the present application;

[0046] Figure 4 A schematic diagram of the structure of a physiological signal processing device based on human factors intelligence provided in an embodiment of the present application;

[0047] Figure 5 A schematic diagram of the structure of an edge computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] In the embodiments of the present application, unless otherwise specified, the character " / " indicates that the objects before and after the association are in an or relationship. For example, A / B can represent A or B. "And / or" describes the association relationship of the associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone.

[0049] It should be pointed out that the words "first", "second", etc. involved in the embodiments of the present application are only used to distinguish the description purpose, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated, nor can they be understood as indicating or implying order.

[0050] In the embodiments of the present application, "at least one" refers to one or more, and "plurality" refers to two or more. In addition, "at least one of the following" or similar expressions refers to any combination of these items, which may include any combination of single items or plural items. For example, at least one of A, B, or C may represent: A, B, C, A and B, A and C, B and C, or A, B and C. Among them, each of A, B, and C may be an element itself, or a set containing one or more elements.

[0051] In the embodiments of the present application, "exemplary", "in some embodiments", "in another embodiment", etc. are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" in the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present concepts in a concrete way.

[0052] In the embodiments of the present application, "of", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the distinction between them is not emphasized, the meanings to be expressed are consistent. In the embodiments of the present application, communication and transmission can sometimes be used interchangeably. It should be noted that when the distinction between them is not emphasized, the meanings to be expressed are consistent. For example, transmission can include sending and / or receiving, which can be a noun or a verb.

[0053] The equal to involved in the embodiments of the present application can be used in conjunction with greater than, and is applicable to the technical solution adopted when greater than, and can also be used in conjunction with less than, and is applicable to the technical solution adopted when less than. It should be noted that when equal to is used in conjunction with greater than, it cannot be used in conjunction with less than; when equal to is used in conjunction with less than, it cannot be used in conjunction with greater than.

[0054] With the continuous improvement of people's health awareness and the rapid development of wearable device technology, the application of obtaining human physiological parameters (such as heart rate, blood oxygen saturation, etc.) based on photoelectric volumetric pulse wave (PPG) signal detection is becoming more and more widespread. Wearable devices such as smart bracelets and smart watches have become common tools for health monitoring in people's daily lives due to their convenience and comfort. PPG technology, as the core sensing method, plays a key role. The PPG signal is a physiological signal that reflects the change in blood volume by detecting the change in light absorption by human tissue. When the heart beats, the blood volume in the peripheral blood vessels will change periodically, causing the light intensity passing through the tissue to change accordingly. After being collected and converted by the photoelectric sensor, the PPG waveform is obtained, which contains a wealth of information related to the cardiovascular system. It is of great value for non-invasive and continuous monitoring of heart rate and rhythm, as well as indirect evaluation of vascular health.

[0055] However, in actual application scenarios, users will inevitably be in motion, such as walking, running, hand shaking, etc. This kind of body movement makes the PPG signal susceptible to serious interference and produces motion artifacts. Motion artifacts usually manifest as signal waveform distortion, baseline drift, and mixing of frequency components. These interferences greatly affect the reliability of accurate extraction of physiological parameters based on PPG signals, resulting in increased measurement errors of key indicators such as heart rate, and even inability to obtain effective physiological data normally in some intense exercise situations, seriously limiting the effective application of PPG technology in sports scenarios and long-term continuous monitoring.

[0056] Traditional signal processing filtering methods, such as low-pass filtering, high-pass filtering, and band-pass filtering, can filter out some high-frequency or low-frequency noise components to a certain extent, but they are often ineffective for complex motion artifacts that are non-stationary and overlap with the physiological signal spectrum. They are prone to filtering out useful physiological signal components at the same time or are unable to completely remove artifacts, making it difficult for the quality of the processed PPG signal to meet the needs of high-precision physiological parameter monitoring.

[0057] Some filtering techniques based on specific models, such as ordinary Kalman filtering, have certain advantages when processing signal filtering of linear systems. However, the relationship between PPG signals and motion artifacts has strong nonlinear characteristics. When dealing with such nonlinear situations, ordinary Kalman filtering has limited filtering accuracy and ability to suppress motion artifacts due to its linearization assumption and other reasons. It cannot adapt well to the complex and changeable actual application environment of PPG signals, resulting in the accuracy of the final extracted physiological parameters still needing to be improved.

[0058] Based on the above problems, the embodiment of the present application proposes a physiological signal processing method based on human intelligence, which uses an adaptive unscented Kalman filtering algorithm to help accurately deal with the nonlinear relationship between physiological signals and motion artifacts, and improve the accuracy of removing artifacts.

[0059] Adaptive Unscented Kalman Filter (AUKF) is a filtering algorithm for state estimation, which is an improvement on the Unscented Kalman Filter (UKF). Based on the Unscented Kalman Filter, the Adaptive Unscented Kalman Filter introduces the idea of ​​adaptation. It dynamically adjusts the number and distribution of unscented points to adapt to the dynamic changes of the system. Specifically, it uses an adaptive method to select and update unscented points according to the dynamic characteristics of the system, thereby improving the estimation accuracy of the system.

[0060] Adaptive unscented Kalman filtering is suitable for nonlinear and non-Gaussian system state estimation problems and can be widely used in robot navigation, target tracking, spacecraft navigation and other fields. It can better adapt to the dynamic characteristics of the system and improve the estimation accuracy by dynamically adjusting the number and distribution of unscented points, while having lower computational complexity.

[0061] In the Unscented Kalman Filter, the word "unscented" mainly reflects a unique idea and method for dealing with nonlinear problems. Traditional Kalman filtering has good effects in dealing with problems such as state estimation of linear systems, but it has limitations when facing nonlinear systems because it relies on linearization to approximate nonlinear functions, which is prone to large errors. The "unscented" in the Unscented Kalman Filter means that it no longer uses conventional linearization methods, but cleverly selects a set of sample points called "Sigma Points". These sample points can accurately capture the probability distribution characteristics of nonlinear functions "without a trace" (that is, without relying on linearization), and use them to transmit and update the mean and covariance of the state, thereby achieving more accurate state estimation of nonlinear systems and other operations.

[0062] Figure 1 The flowchart of the method for processing physiological signals based on human intelligence provided in the embodiment of the present application specifically includes the following steps:

[0063] Step S11, collecting the subject's resting physiological data, measured physiological data, real-time acceleration data and real-time angular velocity data.

[0064] This application takes PPG signals as an example to illustrate the physiological signal processing method based on human intelligence provided by this application. Other physiological signals such as EEG signals, ECG signals, EMG signals, etc. are also applicable to the method provided by this application.

[0065] The subject wears a PPG signal acquisition device (such as a watch with a PPG acquisition function), places his arm flat on the table, and collects PPG data (i.e., resting physiological data) for a fixed time (such as one minute) in a static state. In the test state, the subject's PPG signal is collected in real time to obtain the measured PPG data (i.e., measured physiological data). The IMU (Inertial Measurement Unit) in the PPG signal acquisition device collects acceleration and angular velocity data in real time at a fixed sampling frequency to obtain real-time acceleration data and real-time angular velocity data.

[0066] Step S12: creating a system template of an adaptive unscented Kalman filter according to the resting-state physiological data.

[0067] Taking the PPG signal as an example, a PPG template (i.e., a system template of an adaptive unscented Kalman filter) is created based on the resting PPG data. Figure 2 As shown, Figure 2 The schematic diagram of the PPG template provided in the embodiment of the present application is: Figure 2 It shows multiple sample points in a sampling period, the horizontal axis represents the sampling time, the vertical axis represents the PPG signal strength, and the signal strength change rate can be calculated according to the difference in signal strength at adjacent times.

[0068] Step S13, constructing a system equation model according to the system template and the measured physiological data.

[0069] For a collection device with only one PPG sensor, the filtered state variable (hereinafter represented by x) includes the PPG signal strength (hereinafter represented by u) and the signal strength change rate (hereinafter represented by g), so the state variable of the filter in this application is a two-dimensional vector. The state variable x can be expressed as:

[0070]

[0071] In the related art, the state variable at time t is substituted into the system equation to calculate the state variable at time t+1. The related system equation is as follows:

[0072] x t+1 =f(x t )+v t

[0073] Among them, x t represents the state variable at time t, x t+1represents the predicted state variable at time t+1, v t represents the system noise at time t, and f represents the system equation. Since PPG is a time-varying system, the system equation is a set of partial differential equations, which are extremely difficult to model and solve.

[0074] Therefore, the present application proposes to construct a system equation model based on the system template and the measured physiological data of the subjects, which specifically includes the following steps:

[0075] Step S131 , counting the signal strength and change trend in the resting physiological data according to the system template to obtain a first signal strength set and a first change rate set.

[0076] Specifically, the first signal strength set (hereinafter represented by U) and the first change rate set (hereinafter represented by G) are expressed as:

[0077] U=[u0 u ... u n ]

[0078] G=[g1 g2 ... g n ]

[0079] Where n is the total number of points of the system template,

[0080]

[0081] g j represents the jth component in the set G, u j represents the jth component in the set U, u j-i Indicates u j The previous component of , the rate of change g is expressed as the difference in signal strength in adjacent time periods.

[0082] Step S132: Obtain a second signal strength set and a second change rate set according to the actually measured physiological data statistics.

[0083] Specifically, the first signal strength set (hereinafter referred to as U C denoted by) and the first rate of change set (hereinafter referred to as G C It is represented as:

[0084] U c =[u c_t-m ... u c_t u c_t+1 ]

[0085] G c =[g c_t-m ... g c_t-1 g c_t ]

[0086] Among them, t represents the current moment, and m represents the length of the matching transformation trend, that is, the length of the PPG data segment used to calculate the best matching position is m. C The last component in G is the actual measurement value, which has not been processed by state prediction and update. That is, all components before the last component are filtered values. c Indicates the changing trend of the PPG signal.

[0087] Step S133, performing similarity matching based on the first signal strength set, the first change rate set, the second signal strength set, and the second change rate set to obtain the best matching position, where the best matching position corresponds to the highest similarity in signal strength and change rate.

[0088] Specifically, the best matching position j is calculated according to formula (1), and the calculation formula of formula (1) is as follows:

[0089]

[0090] Among them, G c_r Represents G c The rth component in G j+r represents the j+rth component in G, U c_r Indicates U C The rth component in U j+r represents the j+rth component in U, L min_j Indicates the matching position j corresponding to the loss function value with the smallest return error. The signal strength and change rate corresponding to the best matching position have the highest similarity. In actual application, U C The last component in is the actual measurement value, which has not been processed by state prediction and update. Therefore, U C The last component in does not participate in the best match, and the value participating in the best match is the value after filtering.

[0091] Step S134: Obtain the reference signal strength change rate at the current moment by indexing from the first change rate set according to the best matching position.

[0092] Specifically, the reference signal strength change rate G at the current moment is obtained by indexing from the first change rate set G. j+m+1 .

[0093] Step S135: Calculate the reference signal strength at the next adjacent time based on the signal strength at the current time in the second signal strength set and the reference signal strength change rate at the current time.

[0094] The signal strength change rate is expressed as the difference in signal strength within adjacent time periods. Therefore, the reference signal strength u can be calculated according to formula (2): t+1, the calculation formula of formula (2) is as follows:

[0095] u t+1 =u c_t +G j+m+1 ×Δt

[0096] Among them, t represents the current time, t+1 represents the next adjacent time, and u c_t is the second signal strength set U C The signal strength at the current moment.

[0097] Therefore, the actual PPG signal strength number U c And the corresponding change trend G c , select m state variables (including signal strength and change rate) before time t+1, calculate the best matching position corresponding to the highest similarity with the state variables of the system template, and index to obtain the reference signal strength change rate G at time t j+m+1 , and then based on the second signal strength set U C The signal strength at time t and the reference signal strength change rate G at time t j+m+1 Calculate the reference signal strength u at time t+1 t+1 In addition, the reference signal strength at time t+1 is compared with the measured signal strength at the corresponding time. If the difference between the two is large, it means that the measured PPG data is greatly affected by motion artifacts. t It cannot be measured directly, so the system noise will be accounted for in the covariance matrix below.

[0098] Step S14, determining the exercise intensity according to the real-time acceleration data and the real-time angular velocity data.

[0099] In one embodiment, the real-time acceleration data includes the acceleration components a of the three coordinate axes at the current time (i.e., time t). x (t), a y (t), a z (t), the real-time angular velocity data includes the angular velocity components b of the three coordinate axes at the current moment (i.e., moment t) x (t), b y (t), b z (t), where x, y, and z represent the three coordinate axes. The exercise intensity (hereinafter referred to as E) is determined based on the real-time acceleration data and the real-time angular velocity data, including:

[0100] Step S141, calculating the acceleration amplitude according to the acceleration components of the three coordinate axes at the current moment, and calculating the angular velocity amplitude according to the angular velocity components of the three coordinate axes at the current moment.

[0101] In this application, acceleration amplitude refers to the square root of the sum of the squares of the components of the acceleration vector, which is used to indicate the magnitude of the acceleration vector, and angular velocity amplitude refers to the square root of the sum of the squares of the components of the angular velocity vector, which is used to indicate the magnitude of the angular velocity vector.

[0102] Therefore, the acceleration amplitude a can be calculated according to formula (3): total (t) and the magnitude of the angular velocity b total (t), the calculation formula of formula (3) is as follows:

[0103]

[0104] Step S142, performing weighted average calculation according to the acceleration amplitude and the angular velocity amplitude to obtain the exercise intensity.

[0105] Specifically, the calculation formula for exercise intensity E is as follows:

[0106] E=Aa total (t)+Bb total (t)

[0107] Among them, E is the exercise intensity, A and B are a total (t) and b total (t), E, ​​A and B are all scalars, and the amplitude involved in the calculation here is a dimensionless value.

[0108] Step S15, combining the system equation model and the motion intensity, performing an unscented transformation on the state variables and covariance matrix of the adaptive unscented Kalman filter at the previous adjacent time, to obtain the updated state variables and covariance matrix at the current time.

[0109] Among them, the unscented transformation includes state prediction and state update. The state variable (hereinafter referred to as x) at the previous adjacent time (i.e., time t-1) is t-1，t-1 The state variable at the current moment (i.e., moment t) (hereinafter referred to as x t，t denoted by) is used to characterize the change information of the physiological signal at the current moment, and the covariance matrix at the previous adjacent moment (hereinafter referred to as P t-1，t-1 denoted) is used to characterize the covariance between the components of the state variables at the previous adjacent moments, and the covariance matrix at the current moment (hereinafter referred to as P t，t It is used to characterize the covariance between the components of the state variable at the current moment.

[0110] Specifically, step S15 includes the following steps:

[0111] Step S151, using an adaptive unscented Kalman filter algorithm, generating a sampling point set according to the state variables and covariance matrix at the previous adjacent time, and calculating the sampling point weights.

[0112] Below, the sampling point set is represented as

[0113] The sampling points are also called sigma points, and the number of sigma points is 2N+1.

[0114] Sampling point set include:

[0115]

[0116] Where i is the state number, N is the dimension of the system equation, λ is the adjustment parameter (can be 1), represents (N+k)P t-1，t-1 Optionally, the system equation of the present application is a two-dimensional equation, N=2, and the number of sampling points is 2N+1.

[0117] The sampling point weight is expressed as:

[0118]

[0119] Among them, w0 represents the weight value of the first sigma point, w i Represents the weight value of the remaining sigma points in the sampling point set.

[0120] Step S152, using the system equation model, generating a predicted sampling point set based on the sampling point set, calculating the mean of the predicted state, and calculating the predicted state covariance matrix in combination with the motion intensity.

[0121] Below, the predicted sampling point set is represented as The mean of the predicted state is expressed as The predicted state covariance matrix is ​​denoted as P t，t-1 .

[0122] Specifically, the system equation model constructed in step S13 is used according to the sampling point set Estimate the sampling point set at time t, that is, generate the predicted sampling point set In this application, the dimension of the system equation is 2, so there are 5 sigma points, namely

[0123] And according to the formula Calculate the mean of the predicted state

[0124] Combined with the predicted sampling point set The mean of the predicted states and exercise intensity E to calculate the predicted state covariance matrix P t，t-1 , including:

[0125] The process noise adjustment coefficient α is calculated according to the motion intensity E, where α is calculated by the HardSwish activation function, which can be calculated according to formula (4). The calculation formula of formula (4) is as follows:

[0126]

[0127] Among them, E is the exercise intensity, RELU is the activation function, and the expression of RELU function is:

[0128]

[0129] Based on the sampling point set The mean of the predicted states The process noise adjustment coefficient α and the process noise covariance matrix Q are used to calculate the predicted state covariance matrix P according to formula (5): t，t-1 , the calculation formula of formula (5) is as follows:

[0130]

[0131] Step S153, using the measurement equation, calculate the theoretical measurement value according to the predicted sampling point set, and calculate the mean of the theoretical measurement value and the measurement covariance matrix.

[0132] Below, the theoretical measured value is expressed as The mean of the theoretical measured values ​​is expressed as The measurement covariance matrix is ​​expressed as

[0133] The measurement equation is of the form:

[0134] Z t =h(x t，t-1 )

[0135] Specifically expanded as follows:

[0136]

[0137] Therefore, according to the predicted sampling point set Calculate theoretical measured values It is expressed as:

[0138]

[0139] The mean of the theoretical measurements and the measurement covariance matrix They are:

[0140]

[0141] Among them, R t is the measurement noise covariance matrix, () T Represented as a matrix transpose operation.

[0142] Step S154, calculating the cross covariance matrix according to the predicted sampling point set, the mean of the predicted state, the theoretical measurement value, and the mean of the theoretical measurement value.

[0143] Below, the cross covariance matrix is ​​expressed as

[0144] Specifically, the cross covariance matrix is ​​calculated according to the following formula

[0145]

[0146] Step S155, calculating the Kalman gain matrix according to the cross covariance matrix and the measurement covariance matrix.

[0147] Below, the Kalman gain matrix is ​​represented as K t .

[0148] Specifically, the Kalman gain matrix K is calculated according to the following formula: t

[0149]

[0150] in,() -1 Represents the inverse operation of a matrix.

[0151] Step S156, updating the state variables according to the Kalman gain matrix, the mean of the predicted state, the mean of the theoretical measured values, and the actual measured values ​​at the current moment, to obtain the state variables at the current moment.

[0152] Below, the actual measured value at the current moment is represented as z t .

[0153] Specifically, the current state variable x is calculated according to the following formula: t，t

[0154]

[0155] The actual measured value z t Refers to the value actually measured by the PPG sensor at the current moment.

[0156] In an optional embodiment, the actual measured value z t Substitute the formula to calculate the current state variable x t，tBefore, according to the preset threshold (hereinafter represented by P) and the actual measured value z t The actual measured value z t Perform outlier removal processing. The details are as follows:

[0157]

[0158] If the actual measured value z t is less than or equal to the preset threshold value P, then the actual measured value z t remains unchanged, if the actual measured value z t If the actual measured value z is greater than the preset threshold value P, t Set to be equal to the preset threshold value P. In practical applications, it is necessary to limit the signal amplitude of the measured value or eliminate outliers. The above outlier removal process can remove excessive noise spikes and improve the stability and reliability of the signal.

[0159] Step S157, updating the covariance matrix according to the predicted state covariance matrix, the Kalman gain matrix, and the measured covariance matrix to obtain the covariance matrix at the current moment.

[0160] Specifically, the covariance matrix P at the current moment is calculated according to the following formula: t，t ,

[0161]

[0162] In the embodiment of the present application, the traceless transformation selects a set of sampling points (called sigma points), which can accurately capture the probability distribution characteristics of nonlinear functions "without a trace" (that is, without relying on linearization), and use them to transfer and update the mean and covariance of the state, thereby achieving more accurate state estimation of the nonlinear system and other operations. The sigma points are used to capture the nonlinear relationship between the PPG signal and the motion artifacts, and the motion artifacts in the PPG signal are effectively processed by propagating these sigma points and calculating their transformed statistical characteristics. The traceless transformation can avoid the errors that may be introduced in the linearization process and improve the accuracy of state estimation.

[0163] In an embodiment of the present application, the measured PPG signal (original PPG signal) is input, and in the state prediction step (steps S151-S152), the algorithm predicts the state at the next moment based on the current state and process noise. In the state update step (steps S153-S157), the algorithm uses the measured data (PPG signal) and the predicted state to update the state estimate, thereby reducing the impact of errors and noise. The updated state is the filtered PPG signal. Compared with the original PPG signal, the motion artifacts of the filtered PPG signal are reduced or removed. The filtered PPG signal can more accurately reflect physiological changes, improve the measurement accuracy of physiological parameters such as heart rate and blood oxygen, and facilitate long-term continuous monitoring.

[0164] In other optional embodiments, the method provided by the present application further includes: updating the motion intensity E, and iteratively performing the step of traceless transformation.

[0165] Specifically, the acceleration data and angular velocity data are collected again, the motion intensity E is updated according to step S14, and step S15 is iteratively performed. By updating the motion intensity, the filtering parameters can be optimized to ensure the stability of the artifact removal effect.

[0166] In an optional implementation, the present application can also update the system template according to the exercise intensity. Specifically, if the value of the exercise intensity E is less than the reference value, the resting physiological data of the subject is re-collected to update the system template to improve the accuracy of removing artifacts.

[0167] Figure 3 A schematic diagram of the filtering results provided in the embodiment of the present application, such as Figure 3 As shown, Figure 3 The upper waveform (red waveform) is the original PPG signal without filtering. Figure 3 The upper waveform (green waveform) in the figure is the PPG signal after being filtered by the algorithm provided by the present application.

[0168] There is a complex nonlinear relationship between the PPG signal and the motion artifact. Traditional linear filtering methods or filtering algorithms based on linearization assumptions often produce large deviations when dealing with such nonlinear situations. The adaptive unscented Kalman filtering algorithm uses unscented transformation to accurately process nonlinear functions, which is more in line with the actual interaction between the PPG signal and the motion artifact, thereby more accurately estimating each state variable, effectively separating the state corresponding to the motion artifact from the overall signal, avoiding the error accumulation caused by linear approximation, and improving the accuracy of artifact removal. It can also effectively reduce the computational complexity of computer vision.

[0169] Under different forms of exercise (such as walking, running, jumping, etc.) and different exercise intensities, the interference characteristics of motion artifacts on PPG signals vary greatly, and their nonlinear characteristics also change continuously. With its excellent adaptability to nonlinear systems, the adaptive unscented Kalman filter can flexibly respond to signal changes in various complex motion scenes. Whether it is artifacts generated by relatively slow daily activities or serious interference caused by intense exercise, it can be well processed to ensure the stability of the artifact removal effect.

[0170] This application uses an adaptive unscented Kalman filter algorithm to accurately deal with the nonlinear relationship between PPG signals and motion artifacts, avoid linear approximation errors, and accurately separate artifacts. And combining the system equation model and motion intensity E for filtering helps to adapt to complex and changeable motion scenes, adapt to dynamic changes in signals in real time, and automatically optimize filtering parameters based on motion intensity, individual differences, etc., to ensure stable artifact removal effects. The filtering method proposed in this application can also perform real-time filtering to meet real-time requirements, quickly process dynamic signals, avoid delays, and can integrate new data in real time and make online adaptive adjustments to ensure real-time output of high-quality filtered signals.

[0171] Based on the same idea, the embodiment of the present application also provides a physiological signal processing device based on human factor intelligence, Figure 4 A schematic diagram of the structure of a physiological signal processing device based on human intelligence provided in an embodiment of the present application is shown in FIG. Figure 4 As shown, the physiological signal processing device 40 based on human factors intelligence may include:

[0172] The acquisition module 41 is used to acquire the resting physiological data, measured physiological data, real-time acceleration data and real-time angular velocity data of the subject;

[0173] A creation module 42, for creating a system template of an adaptive unscented Kalman filter according to resting physiological data;

[0174] A construction module 43, for constructing a system equation model according to the system template and measured physiological data;

[0175] A determination module 44, configured to determine the exercise intensity according to the real-time acceleration data and the real-time angular velocity data;

[0176] The untraceable transformation module 45 is used to combine the system equation model and the motion intensity to perform an untraceable transformation on the state variables and covariance matrix of the adaptive untraceable Kalman filter at the previous adjacent moment to obtain the updated state variables and covariance matrix at the current moment; wherein the untraceable transformation includes state prediction and state update, the state variables at the previous adjacent moment are used to characterize the change information of the physiological signal at the previous adjacent moment, the state variables at the current moment are used to characterize the change information of the physiological signal at the current moment, the covariance matrix at the previous adjacent moment is used to characterize the covariance between the various components in the state variables at the previous adjacent moment, and the covariance matrix at the current moment is used to characterize the covariance between the various components in the state variables at the current moment.

[0177] Figure 4 The physiological signal processing device 40 based on human factors intelligence provided in the illustrated embodiment can be used to execute the technical solution of the method embodiment shown in the present application. Its implementation principle and technical effects can be further referred to the relevant description in the method embodiment.

[0178] It should be understood that the above Figure 4 The division of the various modules of the physiological signal processing device 40 based on human factors intelligence shown is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software calling through processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software calling through processing elements, and some modules can be implemented in the form of hardware. For example, the determination module can be a separately established processing element, or it can be integrated in a chip of the edge computing device. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together, or they can be implemented independently. In the implementation process, each step of the above method or each of the above modules can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0179] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more microprocessors (DSP), or one or more field programmable gate arrays (FPGA). For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0180] In the above embodiments, the processor involved may include, for example, a CPU, a DSP, a microcontroller or a digital signal processor, and may also include a GPU, an embedded neural network processor (Neural-network Process Units; hereinafter referred to as: NPU) and an image signal processor (Image Signal Processing; hereinafter referred to as: ISP). The processor may also include necessary hardware accelerators or logic processing hardware circuits, such as ASIC, or one or more integrated circuits for controlling the execution of the program of the technical solution of the present application. In addition, the processor may have the function of operating one or more software programs, and the software programs may be stored in a storage medium.

[0181] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer executes the method provided by the embodiment shown in the present application.

[0182] Combine the following Figure 5 The exemplary edge computing device provided in the embodiments of the present application is further introduced. Figure 5 A schematic diagram of the structure of an edge computing device 5000 is shown.

[0183] The above-mentioned edge computing device 5000 may include: at least one processor; and at least one memory communicatively connected to the above-mentioned processor, wherein: the above-mentioned memory stores program instructions that can be executed by the above-mentioned processor, and the processor calls the above-mentioned program instructions to execute the physiological signal processing method based on human factors intelligence provided in the embodiment shown in this application.

[0184] Figure 5 A block diagram of an exemplary edge computing device 5000 suitable for implementing the embodiments of the present application is shown. Figure 5 The edge computing device 5000 shown is merely an example and should not bring any limitations to the functionality and scope of use of the embodiments of the present application.

[0185] like Figure 5 As shown, the edge computing device 5000 is in the form of a general computing device. The components of the edge computing device 5000 may include, but are not limited to: one or more processors 5010, a memory 5020, a communication bus 5040 connecting different system components (including the memory 5020 and the processor 5010), and a communication interface 5030.

[0186] The communication bus 5040 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnection (PCI) bus.

[0187] The edge computing device 5000 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the edge computing device, including volatile and non-volatile media, removable and non-removable media.

[0188] The memory 5020 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The edge computing device may further include other removable / non-removable, volatile / non-volatile computer system storage media. Figure 5 Not shown in the figure, a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a compact disc read only memory (CD-ROM), a digital versatile disc read only memory (DVD-ROM), or other optical media) may be provided. In these cases, each drive may be connected to the communication bus 5040 via one or more data medium interfaces. The memory 5020 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present application.

[0189] A program / utility having a set (at least one) of program modules may be stored in memory 5020, such program modules including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules generally perform the functions and / or methods of the embodiments described herein.

[0190] The edge computing device 5000 may also communicate with one or more external devices (e.g., keyboards, pointing devices, displays, etc.), one or more devices that enable a user to interact with the edge computing device, and / or any device that enables the edge computing device to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed via the communication interface 5030. In addition, the edge computing device 5000 may also communicate with the network adapter ( Figure 5 The network adapter may communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via the communication bus 5040. It should be understood that although Figure 5 Not shown, other hardware and / or software modules may be used in conjunction with the edge computing device 5000, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, disk arrays (Redundant Arrays of Independent Drives; hereinafter referred to as: RAID) systems, tape drives, and data backup storage systems.

[0191] The processor 5010 executes various functional applications and data processing by running the programs stored in the memory 5020, such as implementing the method provided in the embodiment of the present application.

[0192] It is understandable that the interface connection relationship between the modules illustrated in the embodiment of the present application is only a schematic illustration and does not constitute a structural limitation on the edge computing device 5000. In other embodiments of the present application, the edge computing device 5000 may also adopt different interface connection methods in the above embodiments, or a combination of multiple interface connection methods.

[0193] In the above embodiments, the processor involved may include, for example, a CPU, a DSP, a microcontroller or a digital signal processor, and may also include a GPU, an embedded neural network processor (Neural-network Process Units; hereinafter referred to as: NPU) and an image signal processor (Image Signal Processing; hereinafter referred to as: ISP). The processor may also include necessary hardware accelerators or logic processing hardware circuits, such as ASIC, or one or more integrated circuits for controlling the execution of the program of the technical solution of the present application. In addition, the processor may have the function of operating one or more software programs, and the software programs may be stored in a storage medium.

[0194] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0195] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0196] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0197] The above is only a specific implementation of the present application. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. The protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A physiological signal processing method based on human intelligence, characterized in that: The method comprises: Collect the resting physiological data, measured physiological data, real-time acceleration data and real-time angular velocity data of the subjects; Creating a system template of an adaptive unscented Kalman filter based on the resting-state physiological data; Constructing a system equation model according to the system template and the measured physiological data; determining exercise intensity according to the real-time acceleration data and the real-time angular velocity data; In combination with the system equation model and the motion intensity, the state variables and covariance matrix of the adaptive unscented Kalman filter at the previous adjacent moment are untraceably transformed to obtain the updated state variables and covariance matrix at the current moment; wherein the untraceable transformation includes state prediction and state update, the state variables at the previous adjacent moment are used to characterize the change information of the physiological signal at the previous adjacent moment, the state variables at the current moment are used to characterize the change information of the physiological signal at the current moment, the covariance matrix at the previous adjacent moment is used to characterize the covariance between the components in the state variables at the previous adjacent moment, and the covariance matrix at the current moment is used to characterize the covariance between the components in the state variables at the current moment.

2. The method according to claim 1, characterized in that The state variables include signal strength and signal strength change rate, and the system equation model is constructed according to the system template and the measured physiological data, including: Counting the signal strength and change trend of the resting physiological data according to the system template to obtain a first signal strength set and a first change rate set; Obtaining a second signal strength set and a second change rate set according to the measured physiological data statistics; Perform similarity matching based on the first signal strength set, the first change rate set, the second signal strength set, and the second change rate set to obtain a best matching position, wherein the best matching position corresponds to the highest similarity in signal strength and change rate; Obtaining the reference signal strength change rate at the current moment by indexing from the first change rate set according to the best matching position; The reference signal strength at the next adjacent time is calculated based on the signal strength at the current time in the second signal strength set and the reference signal strength change rate at the current time.

3. The method according to claim 1, characterized in that The real-time acceleration data includes the acceleration components of the three coordinate axes at the current moment, and the real-time angular velocity data includes the angular velocity components of the three coordinate axes at the current moment; The determining the exercise intensity according to the real-time acceleration data and the real-time angular velocity data comprises: Calculating the acceleration amplitude according to the acceleration components of the three coordinate axes at the current moment, and calculating the angular velocity amplitude according to the angular velocity components of the three coordinate axes at the current moment; The exercise intensity is obtained by performing a weighted average calculation based on the acceleration amplitude and the angular velocity amplitude.

4. The method according to claim 1, characterized in that: The step of combining the system equation model and the motion intensity to perform an unscented transformation on the state variables and covariance matrix of the adaptive unscented Kalman filter at the previous adjacent time to obtain the updated state variables and covariance matrix at the current time includes: Using the adaptive unscented Kalman filter algorithm, a sampling point set is generated according to the state variables and covariance matrix at the previous adjacent time, and the sampling point weights are calculated; Using the system equation model, generating a predicted sampling point set based on the sampling point set, calculating a mean of the predicted state, and calculating a predicted state covariance matrix in combination with the motion intensity; Using the measurement equation, calculating theoretical measurement values ​​according to the predicted sampling point set, and calculating the mean of the theoretical measurement values ​​and the measurement covariance matrix; Calculating a cross covariance matrix according to the predicted sampling point set, the mean of the predicted state, the theoretical measurement value, and the mean of the theoretical measurement value; Calculating a Kalman gain matrix based on the cross covariance matrix and the measurement covariance matrix; Update the state variable according to the Kalman gain matrix, the mean of the predicted state, the mean of the theoretical measurement value, and the actual measurement value at the current moment to obtain the state variable at the current moment; The covariance matrix is ​​updated according to the predicted state covariance matrix, the Kalman gain matrix, and the measurement covariance matrix to obtain the covariance matrix at the current moment.

5. The method according to claim 4, characterized in that The step of calculating the predicted state covariance matrix in combination with the exercise intensity includes: Calculating a process noise adjustment factor according to the exercise intensity; A predicted state covariance matrix is ​​calculated based on the sampling point set, the mean of the predicted state, the process noise adjustment coefficient and the process noise covariance matrix.

6. The method according to claim 4, characterized in that The method further includes performing an outlier removal process on the actual measurement value based on a comparison between a preset threshold and the actual measurement value.

7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: If the value of the exercise intensity is less than the reference value, the resting physiological data of the subject is collected again to update the system template.

8. A physiological signal processing device based on human intelligence, characterized in that: include: An acquisition module is used to collect the resting physiological data, measured physiological data, real-time acceleration data and real-time angular velocity data of the subject; A creation module, used for creating a system template of an adaptive unscented Kalman filter according to the resting physiological data; A construction module, used for constructing a system equation model according to the system template and the measured physiological data; A determination module, configured to determine the exercise intensity according to the real-time acceleration data and the real-time angular velocity data; The untraceable transformation module is used to combine the system equation model and the motion intensity to perform an untraceable transformation on the state variables and covariance matrix of the adaptive untraceable Kalman filter at the previous adjacent moment to obtain the updated state variables and covariance matrix at the current moment; wherein the untraceable transformation includes state prediction and state update, the state variables at the previous adjacent moment are used to characterize the change information of the physiological signal at the previous adjacent moment, the state variables at the current moment are used to characterize the change information of the physiological signal at the current moment, the covariance matrix at the previous adjacent moment is used to characterize the covariance between the various components in the state variables at the previous adjacent moment, and the covariance matrix at the current moment is used to characterize the covariance between the various components in the state variables at the current moment.

9. An edge computing device, characterized in that: include: A processor and a memory, wherein the memory is used to store a computer program; the processor is used to run the computer program to implement the physiological signal processing method based on human factors intelligence as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed on a computer, the method for processing physiological signals based on human factors intelligence as described in any one of claims 1 to 7 is implemented.

Images