Prediction master cylinder pressure signal noise reduction method based on leg myoelectricity activity of driver

Abstract

The invention provides a noise reduction method for predicting a master cylinder pressure signal based on leg electromyographic activity of a driver, and relates to the technical field of intelligent vehicle collision avoidance, the method introduces an adaptive unscented Kalman filter (AUKF) method, constructs an automobile braking inverse dynamics model, adaptively adjusts the parameters of the AUKF according to the change of sEMG noise intensity, and improves the noise reduction performance of the vehicle. And the interference of the sEMG noise on the predicted master cylinder pressure signal is weakened. When the predicted master cylinder pressure signal subjected to noise reduction is used as a target value to control a drive-by-wire system (BBWs), the stability and the reliability of the BBWs can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent vehicle collision avoidance, and in particular to a method for predicting master cylinder pressure signal noise reduction based on driver's leg electromyographic activity. Background Art

[0002] A driver's leg muscle activity can indicate braking behavior 100-300 milliseconds in advance. Using leg electromyography (sEMG) signals to predict brake master cylinder pressure and use this as a target value to control brake-by-wire systems (BBWs) can shorten braking distances for electric vehicles and significantly improve the vehicle's collision avoidance success rate.

[0003] Noise in the sEMG signal acquisition process is inevitable. Figure 1 As shown in the figure, due to the noise interference of the sEMG signal, the master cylinder pressure signal is predicted to fluctuate within the shaded area. Using this target value to control BBWs will cause system oscillation, long braking response time, and large fluctuations in braking force output, seriously reducing the smart car's collision avoidance success rate.

[0004] The pressure in a vehicle's brake master cylinder is related to the driver's braking behavior. Emergency braking requires the driver to depress the brake pedal as quickly as possible, resulting in large calf movements, rapid speed, and strong leg muscle contraction. Conversely, during light braking, the driver's calf movements are smaller, slower, and less forceful. sEMG signal noise is uncertain, and its level can vary due to factors such as electrode contact variations, muscle fatigue, and environmental interference. Traditional "model-free" filtering methods, such as Butterworth filters and bandpass filters, while simple to implement and offering good real-time performance, have fixed filtering characteristics and cannot adaptively adjust to changes in noise intensity in the sEMG signal.

[0005] Compared to traditional "model-free" methods, model-driven approaches based on Kalman filtering (KF) can integrate the system's dynamic characteristics with observational information, resulting in superior state estimation capabilities when processing noisy, non-stationary sEMG signals. However, the highly complex and nonlinear musculoskeletal dynamics of drivers, coupled with significant individual variability, make it difficult to establish an accurate state-space model, and KFs struggle to directly filter sEMG signals. Summary of the Invention

[0006] In light of this, the present invention provides a method for denoising the predicted master cylinder pressure signal based on the driver's leg electromyographic activity. This method directly performs inverse dynamics modeling of the vehicle braking process and uses a series of KF methods to directly filter the target value of BBWs control, namely the predicted master cylinder pressure, to achieve accurate estimation of braking intention and interference suppression. Furthermore, considering that the unscented Kalman filter (UKF) method, which uses a fixed measurement noise covariance matrix R, is clearly unable to adapt to dynamic changes in noise levels, it produces large errors when dealing with abnormal fluctuations in the predicted master cylinder pressure signal caused by sEMG signal noise, which can lead to severe deterioration of the filtering effect. Therefore, an adaptive unscented Kalman filter (AUKF) optimization algorithm is proposed to prevent filter divergence.

[0007] Specifically, the present invention provides a method for reducing noise of a predicted master cylinder pressure signal based on the driver's leg electromyographic activity, comprising: Construct an inverse dynamics model of the vehicle braking process and derive the relationship between the vehicle's speed, acceleration and the brake master cylinder pressure; Define the state vector , control input , the outputs are Based on the inverse dynamics model, a state space model of Kalman filtering is established, wherein the state space model includes a state equation and an observation equation; wherein, express The speed of the vehicle at that moment, express The vehicle's acceleration at that moment, express The predicted value of the brake master cylinder pressure output by the LaTAS-F model at time t; Generate sigma points based on the state equation, propagate the sigma points through the state transfer equation, calculate the state prediction value and the state prediction error covariance; based on the observation equation, adaptively adjust the observation covariance matrix ; According to the state prediction error covariance, state prediction value and adaptively adjusted observation covariance matrix Performing state update, wherein the state update includes Kalman gain update, state vector update, and state estimation covariance update; The updated unscented Kalman filter model is used for filtering to obtain the filtered master cylinder pressure prediction value.

[0008] As a further illustration of the present invention, the construction of the inverse dynamics model during vehicle braking specifically includes: Construct the relationship between the vehicle acceleration and the resultant force of its resistance source during vehicle braking: (5) Where, is the mass of the vehicle, a is the acceleration of the vehicle, is the proportional coefficient of braking torque and master cylinder pressure, is the drag coefficient, is the windward area, is the air density, is the rolling resistance coefficient, is the vehicle speed, is the acceleration due to gravity; The brake master cylinder pressure can be deduced from the above formula Expressed as (6).

[0009] As a further illustration of the present invention, the relationship between the acceleration of the vehicle and the resultant force of its resistance source during the vehicle braking process is constructed, specifically including: During vehicle braking, the vehicle acceleration a mainly comes from the combined force of three resistances: (1); ( ); ( ); ( ); After sorting out the above announcement, we can get: ( ).

[0010] As a further illustration of the present invention, the state space model of the Kalman filter is established based on the inverse dynamics model, specifically including: Use the forward Euler method to discretize the continuous system and convert the continuous-time differential equation into a discrete-time difference equation. Let the sampling period be t, then: ( ) The state vector, control input and output of the discrete-time state space model are defined as , , , then the state equation of the system is: ( ) Where, , , is the system noise; The observation equation of the system is: ( ) ( ) Where, is the observation matrix, is the observation noise, represents the predicted observed value of the brake master cylinder pressure obtained according to the inverse dynamics model.

[0011] As a further illustration of the present invention, generating a sigma point based on a state equation, propagating the sigma point through a state transfer equation, and calculating a state prediction value and a state prediction error covariance specifically include: The state vector and the state error covariance matrix Initialization, select the process noise covariance matrix Q and the observation noise covariance matrix R; Generate a set of sigma points and capture the first-order and second-order statistical characteristics of the sigma points through unscented transformation; Propagate the sigma point through the state transfer equation and calculate the state prediction value and state prediction error covariance.

[0012] As a further illustration of the present invention, generating a set of sigma points and capturing the first-order and second-order statistical characteristics of the sigma points by an unscented transformation includes: In each sampling period ( ), for the state dimension n, define the parameter : ( ) To determine the scaling parameter for the sigma point distribution range, To adjust the quadratic scaling parameter of the sigma point distribution, Controls the number of generated sigma points.

[0013] The sigma point is generated as follows: ( ) Where, represents the i-th sigma point, is the i-th column of the state covariance matrix.

[0014] As a further illustration of the present invention, the propagation of the sigma point through the state transfer equation and the calculation of the state prediction value and the state prediction error covariance include: The state transition equation is: ( ) Calculate the state prediction value and state prediction error covariance: ( ) ( ) ( ) ( ) ( ) Where, represents the state prediction value at time k, Represents the state prediction error covariance matrix at time k , and are the weights of the mean and covariance, is the shape parameter of the prior distribution, which represents the prior knowledge of the shape of the posterior distribution. is the process noise covariance matrix.

[0015] As a further illustration of the present invention, the observation covariance matrix is ​​adaptively adjusted based on the observation equation. , specifically including: According to the observation equation, the observation noise is obtained The calculation formula is: (19) Using a sliding window Estimate the actual covariance: ( ) Where, is the sliding window length; Compute the theoretical observation covariance matrix: ( ) Where, is the observation matrix, for The state prediction error covariance matrix at time , is the observation noise covariance matrix of the previous moment; Compute the actual observation covariance matrix: ( ) Where, is the sliding window length; set up The adjustment factor is , the observation noise covariance matrix generated after adaptive adjustment is: ( ) By comparing the actual covariance matrix and the theoretical covariance matrix To dynamically adjust the observation noise covariance matrix ,when Time decrease , otherwise it increases ,Right now: (twenty four) ( ).

[0016] As a further explanation of the present invention, the observation covariance matrix after the state prediction error covariance, the state prediction value and the adaptive adjustment is Perform status updates, including: According to the observation covariance matrix after adaptive adjustment , perform the standard unscented Kalman filter state update steps: (26) Where, for The Kalman gain at time t, for The state prediction error covariance at time , is the observation matrix, is the observation noise covariance matrix after adaptive update; Update the state vector of the inverse dynamics model: ( ) Where, for The final state estimate at time t, for The state prediction value at the moment; ( ) Where, for The state estimation covariance at time t.

[0017] Compared with the prior art, the present invention has the following beneficial technical effects: The present invention introduces an adaptive unscented Kalman filtering method and constructs an inverse dynamics model of vehicle braking. The AUKF filter parameters are adaptively adjusted according to the changes in the sEMG noise intensity to reduce the interference of sEMG noise on the predicted master cylinder pressure signal. When the noise-reduced predicted master cylinder pressure signal is used as the target value for control, the stability and reliability of the brake-by-wire system (BBWs) can be improved.

[0018] Other features and advantages of this technical solution will be described in the subsequent description, and in part will become apparent from the description, or will be understood by practicing this technical solution. The objectives and other advantages of this technical solution can be achieved and obtained through the structures specifically pointed out in the written description and the accompanying drawings.

[0019] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are used to provide a further understanding of the present technical solution and constitute a part of the specification. Together with the embodiments of the present technical solution, they are used to explain the present technical solution and do not constitute a limitation of the present technical solution. In the accompanying drawings: Figure 1 Comparison results of the predicted and actual values ​​of the brake master cylinder pressure after being affected by the noise of the sEMG signal.

[0021] Figure 2 An adaptive unscented Kalman filter model for the brake master cylinder pressure signal noise reduction method provided by the present invention.

[0022] Figure 3 The comparison results of the filtering performance of EKF, UKF and AUKF models under different noise types, where (a) is the SNR indicator comparison result, (b) is the MSE indicator comparison result, and (c) is the CC indicator comparison result.

[0023] Figure 4 This is the filtering result of the AUKF model under the interference of the noise signal whose power varies with time.

[0024] Figure 5 This is the filtering result of the AUKF model under the interference of power random variation noise signal.

[0025] Figure 6 is the filtering result of the AUKF model when the noise variance changes with time. DETAILED DESCRIPTION

[0026] The preferred embodiments of the present technical solution are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present technical solution and are not used to limit the present technical solution.

[0027] The pressure in a vehicle's brake master cylinder is related to the driver's braking behavior. Emergency braking requires the driver to depress the brake pedal as quickly as possible, resulting in large calf movements, rapid speed, and strong leg muscle contraction. Conversely, during light braking, the driver's calf movements are smaller, slower, and less forceful. sEMG signal noise is uncertain, and its level can vary due to factors such as electrode contact variations, muscle fatigue, and environmental interference. Traditional "model-free" filtering methods, such as Butterworth filters and bandpass filters, while simple to implement and offering good real-time performance, have fixed filtering characteristics and cannot adaptively adjust to changes in noise intensity in the sEMG signal.

[0028] Compared to traditional "model-free" approaches, model-driven methods based on Kalman filtering (KF) can integrate the system's dynamic characteristics with observational information, providing superior state estimation capabilities when processing high-noise, non-stationary sEMG signals. However, the driver's musculoskeletal dynamics are highly complex, highly nonlinear, and exhibit significant individual variability, making it difficult to establish an accurate state-space model. Furthermore, KFs struggle to directly filter sEMG signals. Therefore, the present invention directly performs inverse dynamics modeling of the vehicle braking process and uses a series of KF methods to directly filter the target value of BBWs control, namely the predicted master cylinder pressure, to achieve accurate estimation of braking intent and interference suppression.

[0029] Considering that the Unscented Kalman Filter (UKF) method, which uses a fixed measurement noise covariance matrix R, is clearly unable to adapt to dynamic changes in noise levels, it can produce large errors when dealing with abnormal fluctuations in the predicted master cylinder pressure signal caused by sEMG signal noise, which can severely degrade the filtering effect. Therefore, this invention provides an Adaptive Unscented Kalman Filter (AUKF) optimization algorithm to prevent filter divergence.

[0030] Specifically, the present invention provides a method for predicting master cylinder pressure signal noise reduction based on the driver's leg electromyographic activity, comprising the following steps: S1: Construct an inverse dynamics model of the vehicle during braking and derive the relationship between the vehicle's speed, acceleration, and brake master cylinder pressure.

[0031] Step S1 specifically includes the following process: Construct the relationship between the vehicle acceleration and the resultant force of its resistance source during vehicle braking: During vehicle braking, the vehicle acceleration a mainly comes from the combined force of three resistances: ( ) (2) ( ) ( ) After sorting out the above announcement, we can get: (5).

[0032] Where, is the mass of the vehicle, a is the acceleration of the vehicle, is the proportional coefficient of braking torque and master cylinder pressure, is the drag coefficient, is the windward area, is the air density, is the rolling resistance coefficient, is the vehicle speed, is the acceleration due to gravity; The brake master cylinder pressure can be deduced from the above formula Expressed as (6).

[0033] S2: Define the state vector , control input , the outputs are Based on the inverse dynamics model, a state space model of Kalman filtering is established. The state space model includes the state equation and the observation equation. express The speed of the vehicle at that moment, express The vehicle's acceleration at that moment, express The predicted brake master cylinder pressure value output by the LaTAS-F model at this moment. The LaTAS-F (Locality-aware Transformer Architecture Search with Multi-source Fusion) model here is an enhanced local perception Transformer model based on neural network structure search under multi-source information fusion. It is a known model that can predict the driver's continuous braking intention by detecting changes in muscle activity approximately 100 to 300 ms before the driver's braking operation behavior, thereby controlling the brake system master cylinder pressure for braking, which can significantly improve the success rate of collision avoidance. The present invention only uses this model to output the predicted brake master cylinder pressure value, and the specific construction and training process of the model are not repeated here.

[0034] Step S2 specifically includes the following process: Without considering the energy recovery during braking, the forward Euler method is used to discretize the continuous system, converting the continuous-time differential equation into a discrete-time difference equation. Assuming the sampling period is t, then: ( ) The state vector, control input and output of the discrete-time state space model are defined as , , , then the state equation of the system is: ( ) Where, , , is the system noise, which represents the uncertainty and error of the system model, which comes from model inaccuracy, parameter changes, external disturbances and numerical errors; The observation equation of the system is: (9) ( ) Where, is the observation matrix, is the observation noise, used to consider the sEMG signal noise, Represents the predicted observed value of the brake master cylinder pressure obtained according to the inverse dynamics model.

[0035] S3: Generate sigma points based on the state equation, propagate the sigma points through the state transfer equation, and calculate the state prediction value and state prediction error covariance.

[0036] The Kalman filter (KF) algorithm based on the recursive minimum variance estimation principle can optimally estimate the state of linear or weak nonlinear systems. However, for strongly nonlinear systems, the estimation error of KF is large and may even cause the filter to diverge. The present invention adopts the unscented Kalman filter algorithm to extend this function to nonlinear systems, approximating the state distribution by introducing sigma points, and capturing the distribution characteristics after nonlinear transformation through the propagation of these sigma points.

[0037] Specifically, the above step S3 includes the following process: (1) The state vector and the state error covariance matrix Initialization: select the process noise covariance matrix Q and the observation noise covariance matrix R according to actual conditions.

[0038] (2) In order to represent the mean and covariance of the state, a set of sigma points are generated, and the first-order and second-order statistical characteristics of the sigma points are captured by unscented transformation: In each sampling period ( ), for the state dimension n, define the parameter : ( ) To determine the scaling parameter for the sigma point distribution range, To adjust the quadratic scaling parameter of the sigma point distribution, Controls the number of generated sigma points.

[0039] The sigma point is generated as follows: (12) Where, represents the i-th sigma point, is the i-th column of the state covariance matrix.

[0040] (3) Propagate the sigma point through the state transfer equation and calculate the state prediction value and state prediction error covariance: The state transition equation is: ( ) Calculate the state prediction value and state prediction error covariance: (14) ( ) ( ) ( ) ( ) Where, represents the state prediction value at time k, Represents the state prediction error covariance matrix at time k , and are the weights of the mean and covariance, is the shape parameter of the prior distribution, which represents the prior knowledge of the shape of the posterior distribution. is the process noise covariance matrix.

[0041] S4: Adaptively adjust the observation covariance matrix based on the observation equation .

[0042] Specifically, the above step S4 includes the following process: Considering the large error in predicting abnormal fluctuations of the master cylinder pressure signal caused by the uncertainty of the sEMG signal, the predicted observed values ​​of the brake master cylinder pressure are compared. and the actual observed value Adaptive Adjustment .

[0043] According to the observation equation, the observation noise is obtained The calculation formula is: (19) Using a sliding window Estimate the actual covariance: ( ) Where, is the sliding window length.

[0044] Compute the theoretical observation covariance matrix: ( ) Where, is the observation matrix, for The state prediction error covariance matrix at time t represents the system's estimate of the uncertainty of the current state. is the observation noise covariance matrix at the previous moment, which is used to model the effect of sEMG sensor noise on the predicted master cylinder pressure interference.

[0045] Compute the actual observation covariance matrix: (twenty two) Where, is the sliding window length; set up The adjustment factor is , the observation noise covariance matrix generated after adaptive adjustment is: ( ) By comparing the actual covariance matrix and the theoretical covariance matrix To dynamically adjust the observation noise covariance matrix ,when Time decrease , otherwise it increases ,Right now: ( ) ( ).

[0046] S5: Based on the state prediction error covariance, state prediction value and adaptively adjusted observation covariance matrix Perform state update, which includes Kalman gain update, state vector update and state estimation covariance update.

[0047] Specifically, the above step S5 includes the following process: According to the observation covariance matrix after adaptive adjustment , perform the standard unscented Kalman filter state update steps: ( ) Where, for The Kalman gain at time t, for The state prediction error covariance at time , is the observation matrix, is the observation noise covariance matrix after adaptive update; Update the state vector of the inverse dynamics model of the vehicle braking system: ( ) Where, for The final state estimate at time t, for The state prediction value at the moment, for The predicted value of the brake master cylinder pressure output by the LaTAS-F model at time t (actual observation value) ( ) Where, for The state estimation covariance at time t.

[0048] S6: Filtering is performed using the updated unscented Kalman filter model to obtain a filtered master cylinder pressure prediction value.

[0049] Specifically, the final filtered output: ( ) Where, The filtered master cylinder pressure prediction value is the final output of this method.

[0050] Algorithm performance comparison: The signal-to-noise ratio (SNR), mean square error (MSE), and correlation coefficient (CC) are used as metrics to evaluate the filtering performance of the unscented Kalman filter (AUKF) model proposed in this paper. Table 1 compares the performance of several noise reduction algorithms under different noise levels. Denoising methods based on state-space equations generally outperform those without state-space equations. The performance advantages of the state-space-based KF (Kalman filter), EKF (extended Kalman filter), UKF (unscented Kalman filter), and AUKF (adaptive unscented Kalman filter, proposed in this paper) are particularly pronounced for low SNR conditions, such as SNRs of -10 dB, -5 dB, and 0 dB. Describing the dynamic characteristics of the signal through a state-space model can better adapt to the dynamic changes in the driver's braking behavior, significantly improving signal quality while preserving signal details and providing more accurate estimates of the master cylinder pressure in noisy environments.

[0051] Table 1 Comparison of filter performance under different noise levels

[0052] Figure 3 A comparison of the filtering performance of the EKF, UKF, and AUKF algorithms under different noise types revealed that the UKF generally outperformed the EKF. In terms of SNR, the UKF achieved 5.27%, 11.1%, and 5.27% better noise reduction than the EKF. Because the EKF relies heavily on the accuracy of the system model and noise characteristics, errors can accumulate, affecting filtering performance.

[0053] Under the interference of noise signals with randomly varying power, the noise reduction performance of EKF shows more obvious differences compared with UKF. Figures 4-6 The following are the filtering results of three filters under different noise properties. Under the interference of randomly varying noise, the brake master cylinder pressure signal exhibits sudden, short-term, high-amplitude abnormal fluctuations. The EKF linearizes the nonlinear function using a first-order Taylor expansion. However, the random variations in noise power cause the noise signal to have characteristics such as suddenness and discontinuity. The linearization process cannot fully capture the nonlinear characteristics of the randomly varying power noise signal, resulting in large errors when using Taylor expansion for linearization. Especially for operating conditions with large instantaneous noise fluctuations, the EKF may not be able to adjust and adapt to the changes in the noise model in a timely manner.

[0054] In contrast, the UKF handles nonlinear problems through an unscented transform, independent of the linearization assumptions about the system. Unscented transforms can more accurately capture the distributional characteristics of nonlinear systems. When dealing with non-Gaussian noise interference characterized by burstiness and discontinuity, the UKF often exhibits greater robustness and higher accuracy. By capturing the first- and second-order statistical characteristics of the mean and covariance of the predicted master cylinder pressure through an unscented transform, the UKF is more sensitive to the dynamic changes in the predicted master cylinder pressure state distribution and can therefore more effectively cope with noise signal interference with random power variations. Under three different noise attributes, the SNR of the AUKF improves by 18.87%, 37.21%, and 37.22% compared to the UKF, demonstrating that the AUKF can dynamically adjust the filter parameters based on noise variations. Therefore, when dealing with uncertain noise, the proposed AUKF generally provides better noise reduction.

[0055] Obviously, those skilled in the art may make various changes and modifications to this technical solution without departing from the spirit and scope of this technical solution. Thus, if these modifications and variations of this technical solution fall within the scope of the claims of this technical solution and their equivalents, this technical solution is intended to include these modifications and variations.

Claims

1. A method for predicting master cylinder pressure signal noise reduction based on driver's leg electromyographic activity, characterized in that: include: Construct an inverse dynamics model of the vehicle braking process and derive the relationship between the vehicle's speed, acceleration and the brake master cylinder pressure; Define the state vector , control input , the outputs are Based on the inverse dynamics model, a state space model of Kalman filtering is established, wherein the state space model includes a state equation and an observation equation; wherein, express The speed of the vehicle at that moment, express The vehicle's acceleration at that moment, express The predicted value of the brake master cylinder pressure output by the LaTAS-F model at time t; Generate sigma points based on the state equation, propagate the sigma points through the state transfer equation, calculate the state prediction value and the state prediction error covariance; based on the observation equation, adaptively adjust the observation covariance matrix ; According to the state prediction error covariance, state prediction value and adaptively adjusted observation covariance matrix Performing state update, wherein the state update includes Kalman gain update, state vector update, and state estimation covariance update; The updated unscented Kalman filter model is used for filtering to obtain the filtered master cylinder pressure prediction value.

2. The method for predicting master cylinder pressure signal noise reduction based on driver's leg electromyographic activity according to claim 1, characterized in that: The construction of the inverse dynamics model during the vehicle braking process specifically includes: Construct the relationship between the vehicle acceleration and the resultant force of its resistance source during vehicle braking: （5）； Where, is the mass of the vehicle, a is the acceleration of the vehicle, is the proportional coefficient of braking torque and master cylinder pressure, is the drag coefficient, is the windward area, is the air density, is the rolling resistance coefficient, is the vehicle speed, is the acceleration due to gravity; The brake master cylinder pressure can be deduced from the above formula Expressed as (6)。 3. The method for predicting master cylinder pressure signal noise reduction based on driver's leg electromyographic activity according to claim 2, characterized in that: The constructing of the relationship between the vehicle acceleration and the resultant force of its resistance source during the vehicle braking process specifically includes: During vehicle braking, the vehicle acceleration a mainly comes from the combined force of three resistances: (1)； ( )； ( )； ( )； After sorting out the above announcement, we can get: ( )。 4. The method for predicting master cylinder pressure signal noise reduction based on driver's leg electromyographic activity according to claim 2, characterized in that: The state space model of the Kalman filter is established based on the inverse dynamics model, specifically including: Use the forward Euler method to discretize the continuous system and convert the continuous-time differential equation into a discrete-time difference equation. Let the sampling period be t, then: ( )； The state vector, control input and output of the discrete-time state space model are defined as , , , then the state equation of the system is: ( )； Where, , , is the system noise; The observation equation of the system is: ( )； ( )； Where, is the observation matrix, is the observation noise, represents the predicted observed value of the brake master cylinder pressure obtained according to the inverse dynamics model.

5. The method for predicting master cylinder pressure signal noise reduction based on driver's leg electromyographic activity according to claim 1, characterized in that: The method of generating a sigma point based on the state equation, propagating the sigma point through the state transfer equation, and calculating the state prediction value and the state prediction error covariance specifically includes: The state vector and the state error covariance matrix Initialization, select the process noise covariance matrix Q and the observation noise covariance matrix R; Generate a set of sigma points and capture the first-order and second-order statistical characteristics of the sigma points through unscented transformation; Propagate the sigma point through the state transfer equation and calculate the state prediction value and state prediction error covariance.

6. The method for predicting master cylinder pressure signal noise reduction based on driver's leg electromyographic activity according to claim 5, characterized in that: The method of generating a set of sigma points and capturing the first-order and second-order statistical characteristics of the sigma points through an unscented transformation includes: In each sampling period ( ), for the state dimension n, define the parameter : ( )； To determine the scaling parameter for the sigma point distribution range, To adjust the quadratic scaling parameter of the sigma point distribution, Control the number of generated sigma points; The sigma point is generated as follows: (12)； Where, represents the i-th sigma point, is the i-th column of the state covariance matrix.

7. The method for reducing noise of a master cylinder pressure signal based on driver's leg electromyographic activity prediction according to claim 5, characterized in that: The sigma point is propagated through the state transfer equation, and the state prediction value and the state prediction error covariance are calculated, including: The state transition equation is: ( )； Calculate the state prediction value and state prediction error covariance: ( )； ( )； ( )； ( )； ( )； Where, represents the state prediction value at time k, Represents the state prediction error covariance matrix at time k , and are the weights of the mean and covariance, is the shape parameter of the prior distribution, which represents the prior knowledge of the shape of the posterior distribution. is the process noise covariance matrix.

8. The method for reducing noise of a master cylinder pressure signal based on driver's leg electromyographic activity prediction according to claim 1, characterized in that: Based on the observation equation, the observation covariance matrix is ​​adaptively adjusted , specifically including: According to the observation equation, the observation noise is obtained The calculation formula is: ( )； Using a sliding window Estimate the actual covariance: ( )； Where, is the sliding window length; Compute the theoretical observation covariance matrix: ( )； Where, is the observation matrix, for The state prediction error covariance matrix at time , is the observation noise covariance matrix of the previous moment; Compute the actual observation covariance matrix: ( )； Where, is the sliding window length; set up The adjustment factor is , the observation noise covariance matrix generated after adaptive adjustment is: ( )； By comparing the actual covariance matrix and the theoretical covariance matrix To dynamically adjust the observation noise covariance matrix ,when Time decrease , otherwise it increases ,Right now: ( )； ( )。 9. The method for predicting master cylinder pressure signal noise reduction based on driver's leg electromyographic activity according to claim 1, characterized in that: The observation covariance matrix after the state prediction error covariance, state prediction value and adaptive adjustment Perform status updates, including: According to the observation covariance matrix after adaptive adjustment , perform the standard unscented Kalman filter state update steps: ( )； Where, for The Kalman gain at time t, for The state prediction error covariance at time , is the observation matrix, is the observation noise covariance matrix after adaptive update; Update the state vector of the inverse dynamics model: ( )； Where, for The final state estimate at time t, for The state prediction value at the moment; ( )； Where, for The state estimation covariance at time t.

Images